refactor: simplify chat query construction by extracting sort_by from debouncedQueryParams

Co-authored-by: iamjoel <2120155+iamjoel@users.noreply.github.com>

.github/workflows/style.yml

@@ -106,7 +106,7 @@ jobs:
       - name: Web type check
         if: steps.changed-files.outputs.any_changed == 'true'
         working-directory: ./web
-        run: pnpm run type-check
+        run: pnpm run type-check:tsgo
 
   docker-compose-template:
     name: Docker Compose Template

AGENTS.md

@@ -24,8 +24,8 @@ The codebase is split into:
 
 ```bash
 cd web
-pnpm lint
 pnpm lint:fix
+pnpm type-check:tsgo
 pnpm test
 ```
 
@@ -39,7 +39,7 @@ pnpm test
 ## Language Style
 
 - **Python**: Keep type hints on functions and attributes, and implement relevant special methods (e.g., `__repr__`, `__str__`).
-- **TypeScript**: Use the strict config, lean on ESLint + Prettier workflows, and avoid `any` types.
+- **TypeScript**: Use the strict config, rely on ESLint (`pnpm lint:fix` preferred) plus `pnpm type-check:tsgo`, and avoid `any` types.
 
 ## General Practices
 

README.md

@@ -139,6 +139,19 @@ Star Dify on GitHub and be instantly notified of new releases.
 
 If you need to customize the configuration, please refer to the comments in our [.env.example](docker/.env.example) file and update the corresponding values in your `.env` file. Additionally, you might need to make adjustments to the `docker-compose.yaml` file itself, such as changing image versions, port mappings, or volume mounts, based on your specific deployment environment and requirements. After making any changes, please re-run `docker-compose up -d`. You can find the full list of available environment variables [here](https://docs.dify.ai/getting-started/install-self-hosted/environments).
 
+#### Customizing Suggested Questions
+
+You can now customize the "Suggested Questions After Answer" feature to better fit your use case. For example, to generate longer, more technical questions:
+
+```bash
+# In your .env file
+SUGGESTED_QUESTIONS_PROMPT='Please help me predict the five most likely technical follow-up questions a developer would ask. Focus on implementation details, best practices, and architecture considerations. Keep each question between 40-60 characters. Output must be JSON array: ["question1","question2","question3","question4","question5"]'
+SUGGESTED_QUESTIONS_MAX_TOKENS=512
+SUGGESTED_QUESTIONS_TEMPERATURE=0.3
+```
+
+See the [Suggested Questions Configuration Guide](docs/suggested-questions-configuration.md) for detailed examples and usage instructions.
+
 ### Metrics Monitoring with Grafana
 
 Import the dashboard to Grafana, using Dify's PostgreSQL database as data source, to monitor metrics in granularity of apps, tenants, messages, and more.

api/.dockerignore

@@ -1,26 +1,6 @@
 .env
 *.env.*
 
-# VCS and editor noise
-.git
-.DS_Store
-__pycache__/
-*.pyc
-*.pyo
-*.swp
-.vscode
-
-# Large or irrelevant project files
-docs/
-dev/
-images/
-
-
-# Local caches
-.pytest_cache
-.ruff_cache
-.mypy_cache
-
 storage/generate_files/*
 storage/privkeys/*
 storage/tools/*
@@ -32,6 +12,8 @@ logs
 
 # jetbrains
 .idea
+.mypy_cache
+.ruff_cache
 
 # venv
 .venv

api/.env.example

@@ -633,8 +633,30 @@ SWAGGER_UI_PATH=/swagger-ui.html
 # Set to false to export dataset IDs as plain text for easier cross-environment import
 DSL_EXPORT_ENCRYPT_DATASET_ID=true
 
+# Suggested Questions After Answer Configuration
+# These environment variables allow customization of the suggested questions feature
+#
+# Custom prompt for generating suggested questions (optional)
+# If not set, uses the default prompt that generates 3 questions under 20 characters each
+# Example: "Please help me predict the five most likely technical follow-up questions a developer would ask. Focus on implementation details, best practices, and architecture considerations. Keep each question between 40-60 characters. Output must be JSON array: [\"question1\",\"question2\",\"question3\",\"question4\",\"question5\"]"
+# SUGGESTED_QUESTIONS_PROMPT=
+
+# Maximum number of tokens for suggested questions generation (default: 256)
+# Adjust this value for longer questions or more questions
+# SUGGESTED_QUESTIONS_MAX_TOKENS=256
+
+# Temperature for suggested questions generation (default: 0.0)
+# Higher values (0.5-1.0) produce more creative questions, lower values (0.0-0.3) produce more focused questions
+# SUGGESTED_QUESTIONS_TEMPERATURE=0
+
 # Tenant isolated task queue configuration
 TENANT_ISOLATED_TASK_CONCURRENCY=1
 
 # Maximum number of segments for dataset segments API (0 for unlimited)
 DATASET_MAX_SEGMENTS_PER_REQUEST=0
+
+# Multimodal knowledgebase limit
+SINGLE_CHUNK_ATTACHMENT_LIMIT=10
+ATTACHMENT_IMAGE_FILE_SIZE_LIMIT=2
+ATTACHMENT_IMAGE_DOWNLOAD_TIMEOUT=60
+IMAGE_FILE_BATCH_LIMIT=10

api/.ruff.toml

@@ -36,17 +36,20 @@ select = [
     "UP",      # pyupgrade rules
     "W191",    # tab-indentation
     "W605",    # invalid-escape-sequence
+    "G001",    # don't use str format to logging messages
+    "G003",    # don't use + in logging messages
+    "G004",    # don't use f-strings to format logging messages
+    "UP042",   # use StrEnum,
+    "S110",    # disallow the try-except-pass pattern.
+
     # security related linting rules
     # RCE proctection (sort of)
     "S102", # exec-builtin, disallow use of `exec`
     "S307", # suspicious-eval-usage, disallow use of `eval` and `ast.literal_eval`
     "S301", # suspicious-pickle-usage, disallow use of `pickle` and its wrappers.
     "S302", # suspicious-marshal-usage, disallow use of `marshal` module
-    "S311", # suspicious-non-cryptographic-random-usage
-    "G001", # don't use str format to logging messages
-    "G003", # don't use + in logging messages
-    "G004", # don't use f-strings to format logging messages
-    "UP042", # use StrEnum
+    "S311", # suspicious-non-cryptographic-random-usage,
+
 ]
 
 ignore = [
@@ -91,18 +94,16 @@ ignore = [
 "configs/*" = [
     "N802", # invalid-function-name
 ]
-"core/model_runtime/callbacks/base_callback.py" = [
-    "T201",
-]
-"core/workflow/callbacks/workflow_logging_callback.py" = [
-    "T201",
-]
+"core/model_runtime/callbacks/base_callback.py" = ["T201"]
+"core/workflow/callbacks/workflow_logging_callback.py" = ["T201"]
 "libs/gmpy2_pkcs10aep_cipher.py" = [
     "N803", # invalid-argument-name
 ]
 "tests/*" = [
     "F811", # redefined-while-unused
-    "T201", # allow print in tests
+    "T201", # allow print in tests,
+    "S110", # allow ignoring exceptions in tests code (currently)
+
 ]
 
 [lint.pyflakes]

api/Dockerfile

@@ -19,15 +19,11 @@ RUN apt-get update \
           # basic environment
           g++ \
           # for building gmpy2
-          libmpfr-dev libmpc-dev \
-    && rm -rf /var/lib/apt/lists/*
+          libmpfr-dev libmpc-dev
 
 # Install Python dependencies
 COPY pyproject.toml uv.lock ./
-# Export without hashes because we'll build local wheels (hashes would mismatch)
-RUN uv export --locked --no-dev --format requirements.txt --no-hashes --output-file /tmp/requirements.txt \
-    && pip wheel --no-cache-dir -r /tmp/requirements.txt -w /wheels \
-    && uv cache prune --ci
+RUN uv sync --locked --no-dev
 
 # production stage
 FROM base AS production
@@ -58,10 +54,10 @@ RUN groupadd -r -g ${dify_uid} dify && \
     useradd -r -u ${dify_uid} -g ${dify_uid} -s /bin/bash dify && \
     chown -R dify:dify /app
 
-RUN set -eux; \
-    apt-get update; \
+RUN \
+    apt-get update \
     # Install dependencies
-    apt-get install -y --no-install-recommends \
+    && apt-get install -y --no-install-recommends \
         # basic environment
         curl nodejs \
         # for gmpy2 \
@@ -71,33 +67,33 @@ RUN set -eux; \
         # install fonts to support the use of tools like pypdfium2
         fonts-noto-cjk \
         # install a package to improve the accuracy of guessing mime type and file extension
-    media-types \
-    # install libmagic to support the use of python-magic guess MIMETYPE
-    libmagic1; \
-    apt-get autoremove -y; \
-    rm -rf /var/lib/apt/lists/*
+        media-types \
+        # install libmagic to support the use of python-magic guess MIMETYPE
+        libmagic1 \
+    && apt-get autoremove -y \
+    && rm -rf /var/lib/apt/lists/*
 
-# Install Python packages from prebuilt wheels (no virtualenv to avoid copying ~1.8GB layer)
-COPY --from=packages /tmp/requirements.txt /tmp/requirements.txt
-COPY --from=packages /wheels /wheels
-RUN pip install --no-cache-dir --no-index --find-links /wheels -r /tmp/requirements.txt \
-    && rm -rf /wheels /tmp/requirements.txt
+# Copy Python environment and packages
+ENV VIRTUAL_ENV=/app/api/.venv
+COPY --from=packages --chown=dify:dify ${VIRTUAL_ENV} ${VIRTUAL_ENV}
+ENV PATH="${VIRTUAL_ENV}/bin:${PATH}"
+
+# Download nltk data
+RUN mkdir -p /usr/local/share/nltk_data && NLTK_DATA=/usr/local/share/nltk_data python -c "import nltk; nltk.download('punkt'); nltk.download('averaged_perceptron_tagger'); nltk.download('stopwords')" \
+    && chmod -R 755 /usr/local/share/nltk_data
 
 ENV TIKTOKEN_CACHE_DIR=/app/api/.tiktoken_cache
 
-# Pre-fetch NLTK data and warm tiktoken cache before copying source to maximize layer reuse
-RUN set -eux; \
-    mkdir -p /usr/local/share/nltk_data; \
-    NLTK_DATA=/usr/local/share/nltk_data python -c "import nltk; nltk.download('punkt'); nltk.download('averaged_perceptron_tagger'); nltk.download('stopwords')"; \
-    chmod -R 755 /usr/local/share/nltk_data; \
-    python -c "import tiktoken; tiktoken.encoding_for_model('gpt2')" \
+RUN python -c "import tiktoken; tiktoken.encoding_for_model('gpt2')" \
     && chown -R dify:dify ${TIKTOKEN_CACHE_DIR}
 
 # Copy source code
 COPY --chown=dify:dify . /app/api/
+
 # Prepare entrypoint script
 COPY --chown=dify:dify --chmod=755 docker/entrypoint.sh /entrypoint.sh
 
+
 ARG COMMIT_SHA
 ENV COMMIT_SHA=${COMMIT_SHA}
 ENV NLTK_DATA=/usr/local/share/nltk_data

api/commands.py

@@ -1139,6 +1139,7 @@ def remove_orphaned_files_on_storage(force: bool):
         click.echo(click.style(f"Found {len(all_files_in_tables)} files in tables.", fg="white"))
     except Exception as e:
         click.echo(click.style(f"Error fetching keys: {str(e)}", fg="red"))
+        return
 
     all_files_on_storage = []
     for storage_path in storage_paths:

api/configs/feature/__init__.py

@@ -360,6 +360,26 @@ class FileUploadConfig(BaseSettings):
         default=10,
     )
 
+    IMAGE_FILE_BATCH_LIMIT: PositiveInt = Field(
+        description="Maximum number of files allowed in a image batch upload operation",
+        default=10,
+    )
+
+    SINGLE_CHUNK_ATTACHMENT_LIMIT: PositiveInt = Field(
+        description="Maximum number of files allowed in a single chunk attachment",
+        default=10,
+    )
+
+    ATTACHMENT_IMAGE_FILE_SIZE_LIMIT: NonNegativeInt = Field(
+        description="Maximum allowed image file size for attachments in megabytes",
+        default=2,
+    )
+
+    ATTACHMENT_IMAGE_DOWNLOAD_TIMEOUT: NonNegativeInt = Field(
+        description="Timeout for downloading image attachments in seconds",
+        default=60,
+    )
+
     inner_UPLOAD_FILE_EXTENSION_BLACKLIST: str = Field(
         description=(
             "Comma-separated list of file extensions that are blocked from upload. "

api/controllers/common/schema.py

@@ -0,0 +1,26 @@
+"""Helpers for registering Pydantic models with Flask-RESTX namespaces."""
+
+from flask_restx import Namespace
+from pydantic import BaseModel
+
+DEFAULT_REF_TEMPLATE_SWAGGER_2_0 = "#/definitions/{model}"
+
+
+def register_schema_model(namespace: Namespace, model: type[BaseModel]) -> None:
+    """Register a single BaseModel with a namespace for Swagger documentation."""
+
+    namespace.schema_model(model.__name__, model.model_json_schema(ref_template=DEFAULT_REF_TEMPLATE_SWAGGER_2_0))
+
+
+def register_schema_models(namespace: Namespace, *models: type[BaseModel]) -> None:
+    """Register multiple BaseModels with a namespace."""
+
+    for model in models:
+        register_schema_model(namespace, model)
+
+
+__all__ = [
+    "DEFAULT_REF_TEMPLATE_SWAGGER_2_0",
+    "register_schema_model",
+    "register_schema_models",
+]

api/controllers/console/app/app.py

@@ -31,7 +31,6 @@ from fields.app_fields import (
 from fields.workflow_fields import workflow_partial_fields as _workflow_partial_fields_dict
 from libs.helper import AppIconUrlField, TimestampField
 from libs.login import current_account_with_tenant, login_required
-from libs.validators import validate_description_length
 from models import App, Workflow
 from services.app_dsl_service import AppDslService, ImportMode
 from services.app_service import AppService
@@ -76,51 +75,30 @@ class AppListQuery(BaseModel):
 
 class CreateAppPayload(BaseModel):
     name: str = Field(..., min_length=1, description="App name")
-    description: str | None = Field(default=None, description="App description (max 400 chars)")
+    description: str | None = Field(default=None, description="App description (max 400 chars)", max_length=400)
     mode: Literal["chat", "agent-chat", "advanced-chat", "workflow", "completion"] = Field(..., description="App mode")
     icon_type: str | None = Field(default=None, description="Icon type")
     icon: str | None = Field(default=None, description="Icon")
     icon_background: str | None = Field(default=None, description="Icon background color")
 
-    @field_validator("description")
-    @classmethod
-    def validate_description(cls, value: str | None) -> str | None:
-        if value is None:
-            return value
-        return validate_description_length(value)
-
 
 class UpdateAppPayload(BaseModel):
     name: str = Field(..., min_length=1, description="App name")
-    description: str | None = Field(default=None, description="App description (max 400 chars)")
+    description: str | None = Field(default=None, description="App description (max 400 chars)", max_length=400)
     icon_type: str | None = Field(default=None, description="Icon type")
     icon: str | None = Field(default=None, description="Icon")
     icon_background: str | None = Field(default=None, description="Icon background color")
     use_icon_as_answer_icon: bool | None = Field(default=None, description="Use icon as answer icon")
     max_active_requests: int | None = Field(default=None, description="Maximum active requests")
 
-    @field_validator("description")
-    @classmethod
-    def validate_description(cls, value: str | None) -> str | None:
-        if value is None:
-            return value
-        return validate_description_length(value)
-
 
 class CopyAppPayload(BaseModel):
     name: str | None = Field(default=None, description="Name for the copied app")
-    description: str | None = Field(default=None, description="Description for the copied app")
+    description: str | None = Field(default=None, description="Description for the copied app", max_length=400)
     icon_type: str | None = Field(default=None, description="Icon type")
     icon: str | None = Field(default=None, description="Icon")
     icon_background: str | None = Field(default=None, description="Icon background color")
 
-    @field_validator("description")
-    @classmethod
-    def validate_description(cls, value: str | None) -> str | None:
-        if value is None:
-            return value
-        return validate_description_length(value)
-
 
 class AppExportQuery(BaseModel):
     include_secret: bool = Field(default=False, description="Include secrets in export")

api/controllers/console/app/message.py

@@ -61,6 +61,7 @@ class ChatMessagesQuery(BaseModel):
 class MessageFeedbackPayload(BaseModel):
     message_id: str = Field(..., description="Message ID")
     rating: Literal["like", "dislike"] | None = Field(default=None, description="Feedback rating")
+    content: str | None = Field(default=None, description="Feedback content")
 
     @field_validator("message_id")
     @classmethod
@@ -324,6 +325,7 @@ class MessageFeedbackApi(Resource):
             db.session.delete(feedback)
         elif args.rating and feedback:
             feedback.rating = args.rating
+            feedback.content = args.content
         elif not args.rating and not feedback:
             raise ValueError("rating cannot be None when feedback not exists")
         else:
@@ -335,6 +337,7 @@ class MessageFeedbackApi(Resource):
                 conversation_id=message.conversation_id,
                 message_id=message.id,
                 rating=rating_value,
+                content=args.content,
                 from_source="admin",
                 from_account_id=current_user.id,
             )

api/controllers/console/datasets/data_source.py

@@ -1,15 +1,15 @@
 import json
 from collections.abc import Generator
-from typing import cast
+from typing import Any, cast
 
 from flask import request
-from flask_restx import Resource, marshal_with, reqparse
+from flask_restx import Resource, marshal_with
+from pydantic import BaseModel, Field
 from sqlalchemy import select
 from sqlalchemy.orm import Session
 from werkzeug.exceptions import NotFound
 
-from controllers.console import console_ns
-from controllers.console.wraps import account_initialization_required, setup_required
+from controllers.common.schema import register_schema_model
 from core.datasource.entities.datasource_entities import DatasourceProviderType, OnlineDocumentPagesMessage
 from core.datasource.online_document.online_document_plugin import OnlineDocumentDatasourcePlugin
 from core.indexing_runner import IndexingRunner
@@ -25,6 +25,19 @@ from services.dataset_service import DatasetService, DocumentService
 from services.datasource_provider_service import DatasourceProviderService
 from tasks.document_indexing_sync_task import document_indexing_sync_task
 
+from .. import console_ns
+from ..wraps import account_initialization_required, setup_required
+
+
+class NotionEstimatePayload(BaseModel):
+    notion_info_list: list[dict[str, Any]]
+    process_rule: dict[str, Any]
+    doc_form: str = Field(default="text_model")
+    doc_language: str = Field(default="English")
+
+
+register_schema_model(console_ns, NotionEstimatePayload)
+
 
 @console_ns.route(
     "/data-source/integrates",
@@ -243,20 +256,15 @@ class DataSourceNotionApi(Resource):
     @setup_required
     @login_required
     @account_initialization_required
+    @console_ns.expect(console_ns.models[NotionEstimatePayload.__name__])
     def post(self):
         _, current_tenant_id = current_account_with_tenant()
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("notion_info_list", type=list, required=True, nullable=True, location="json")
-            .add_argument("process_rule", type=dict, required=True, nullable=True, location="json")
-            .add_argument("doc_form", type=str, default="text_model", required=False, nullable=False, location="json")
-            .add_argument("doc_language", type=str, default="English", required=False, nullable=False, location="json")
-        )
-        args = parser.parse_args()
+        payload = NotionEstimatePayload.model_validate(console_ns.payload or {})
+        args = payload.model_dump()
         # validate args
         DocumentService.estimate_args_validate(args)
-        notion_info_list = args["notion_info_list"]
+        notion_info_list = payload.notion_info_list
         extract_settings = []
         for notion_info in notion_info_list:
             workspace_id = notion_info["workspace_id"]

api/controllers/console/datasets/datasets.py

@@ -1,12 +1,14 @@
 from typing import Any, cast
 
 from flask import request
-from flask_restx import Resource, fields, marshal, marshal_with, reqparse
+from flask_restx import Resource, fields, marshal, marshal_with
+from pydantic import BaseModel, Field, field_validator
 from sqlalchemy import select
 from werkzeug.exceptions import Forbidden, NotFound
 
 import services
 from configs import dify_config
+from controllers.common.schema import register_schema_models
 from controllers.console import console_ns
 from controllers.console.apikey import (
     api_key_item_model,
@@ -48,7 +50,6 @@ from fields.dataset_fields import (
 )
 from fields.document_fields import document_status_fields
 from libs.login import current_account_with_tenant, login_required
-from libs.validators import validate_description_length
 from models import ApiToken, Dataset, Document, DocumentSegment, UploadFile
 from models.dataset import DatasetPermissionEnum
 from models.provider_ids import ModelProviderID
@@ -107,10 +108,75 @@ related_app_list_copy["data"] = fields.List(fields.Nested(app_detail_kernel_mode
 related_app_list_model = _get_or_create_model("RelatedAppList", related_app_list_copy)
 
 
-def _validate_name(name: str) -> str:
-    if not name or len(name) < 1 or len(name) > 40:
-        raise ValueError("Name must be between 1 to 40 characters.")
-    return name
+def _validate_indexing_technique(value: str | None) -> str | None:
+    if value is None:
+        return value
+    if value not in Dataset.INDEXING_TECHNIQUE_LIST:
+        raise ValueError("Invalid indexing technique.")
+    return value
+
+
+class DatasetCreatePayload(BaseModel):
+    name: str = Field(..., min_length=1, max_length=40)
+    description: str = Field("", max_length=400)
+    indexing_technique: str | None = None
+    permission: DatasetPermissionEnum | None = DatasetPermissionEnum.ONLY_ME
+    provider: str = "vendor"
+    external_knowledge_api_id: str | None = None
+    external_knowledge_id: str | None = None
+
+    @field_validator("indexing_technique")
+    @classmethod
+    def validate_indexing(cls, value: str | None) -> str | None:
+        return _validate_indexing_technique(value)
+
+    @field_validator("provider")
+    @classmethod
+    def validate_provider(cls, value: str) -> str:
+        if value not in Dataset.PROVIDER_LIST:
+            raise ValueError("Invalid provider.")
+        return value
+
+
+class DatasetUpdatePayload(BaseModel):
+    name: str | None = Field(None, min_length=1, max_length=40)
+    description: str | None = Field(None, max_length=400)
+    permission: DatasetPermissionEnum | None = None
+    indexing_technique: str | None = None
+    embedding_model: str | None = None
+    embedding_model_provider: str | None = None
+    retrieval_model: dict[str, Any] | None = None
+    partial_member_list: list[str] | None = None
+    external_retrieval_model: dict[str, Any] | None = None
+    external_knowledge_id: str | None = None
+    external_knowledge_api_id: str | None = None
+    icon_info: dict[str, Any] | None = None
+    is_multimodal: bool | None = False
+
+    @field_validator("indexing_technique")
+    @classmethod
+    def validate_indexing(cls, value: str | None) -> str | None:
+        return _validate_indexing_technique(value)
+
+
+class IndexingEstimatePayload(BaseModel):
+    info_list: dict[str, Any]
+    process_rule: dict[str, Any]
+    indexing_technique: str
+    doc_form: str = "text_model"
+    dataset_id: str | None = None
+    doc_language: str = "English"
+
+    @field_validator("indexing_technique")
+    @classmethod
+    def validate_indexing(cls, value: str) -> str:
+        result = _validate_indexing_technique(value)
+        if result is None:
+            raise ValueError("indexing_technique is required.")
+        return result
+
+
+register_schema_models(console_ns, DatasetCreatePayload, DatasetUpdatePayload, IndexingEstimatePayload)
 
 
 def _get_retrieval_methods_by_vector_type(vector_type: str | None, is_mock: bool = False) -> dict[str, list[str]]:
@@ -255,20 +321,7 @@ class DatasetListApi(Resource):
 
     @console_ns.doc("create_dataset")
     @console_ns.doc(description="Create a new dataset")
-    @console_ns.expect(
-        console_ns.model(
-            "CreateDatasetRequest",
-            {
-                "name": fields.String(required=True, description="Dataset name (1-40 characters)"),
-                "description": fields.String(description="Dataset description (max 400 characters)"),
-                "indexing_technique": fields.String(description="Indexing technique"),
-                "permission": fields.String(description="Dataset permission"),
-                "provider": fields.String(description="Provider"),
-                "external_knowledge_api_id": fields.String(description="External knowledge API ID"),
-                "external_knowledge_id": fields.String(description="External knowledge ID"),
-            },
-        )
-    )
+    @console_ns.expect(console_ns.models[DatasetCreatePayload.__name__])
     @console_ns.response(201, "Dataset created successfully")
     @console_ns.response(400, "Invalid request parameters")
     @setup_required
@@ -276,52 +329,7 @@ class DatasetListApi(Resource):
     @account_initialization_required
     @cloud_edition_billing_rate_limit_check("knowledge")
     def post(self):
-        parser = (
-            reqparse.RequestParser()
-            .add_argument(
-                "name",
-                nullable=False,
-                required=True,
-                help="type is required. Name must be between 1 to 40 characters.",
-                type=_validate_name,
-            )
-            .add_argument(
-                "description",
-                type=validate_description_length,
-                nullable=True,
-                required=False,
-                default="",
-            )
-            .add_argument(
-                "indexing_technique",
-                type=str,
-                location="json",
-                choices=Dataset.INDEXING_TECHNIQUE_LIST,
-                nullable=True,
-                help="Invalid indexing technique.",
-            )
-            .add_argument(
-                "external_knowledge_api_id",
-                type=str,
-                nullable=True,
-                required=False,
-            )
-            .add_argument(
-                "provider",
-                type=str,
-                nullable=True,
-                choices=Dataset.PROVIDER_LIST,
-                required=False,
-                default="vendor",
-            )
-            .add_argument(
-                "external_knowledge_id",
-                type=str,
-                nullable=True,
-                required=False,
-            )
-        )
-        args = parser.parse_args()
+        payload = DatasetCreatePayload.model_validate(console_ns.payload or {})
         current_user, current_tenant_id = current_account_with_tenant()
 
         # The role of the current user in the ta table must be admin, owner, or editor, or dataset_operator
@@ -331,14 +339,14 @@ class DatasetListApi(Resource):
         try:
             dataset = DatasetService.create_empty_dataset(
                 tenant_id=current_tenant_id,
-                name=args["name"],
-                description=args["description"],
-                indexing_technique=args["indexing_technique"],
+                name=payload.name,
+                description=payload.description,
+                indexing_technique=payload.indexing_technique,
                 account=current_user,
-                permission=DatasetPermissionEnum.ONLY_ME,
-                provider=args["provider"],
-                external_knowledge_api_id=args["external_knowledge_api_id"],
-                external_knowledge_id=args["external_knowledge_id"],
+                permission=payload.permission or DatasetPermissionEnum.ONLY_ME,
+                provider=payload.provider,
+                external_knowledge_api_id=payload.external_knowledge_api_id,
+                external_knowledge_id=payload.external_knowledge_id,
             )
         except services.errors.dataset.DatasetNameDuplicateError:
             raise DatasetNameDuplicateError()
@@ -399,18 +407,7 @@ class DatasetApi(Resource):
 
     @console_ns.doc("update_dataset")
     @console_ns.doc(description="Update dataset details")
-    @console_ns.expect(
-        console_ns.model(
-            "UpdateDatasetRequest",
-            {
-                "name": fields.String(description="Dataset name"),
-                "description": fields.String(description="Dataset description"),
-                "permission": fields.String(description="Dataset permission"),
-                "indexing_technique": fields.String(description="Indexing technique"),
-                "external_retrieval_model": fields.Raw(description="External retrieval model settings"),
-            },
-        )
-    )
+    @console_ns.expect(console_ns.models[DatasetUpdatePayload.__name__])
     @console_ns.response(200, "Dataset updated successfully", dataset_detail_model)
     @console_ns.response(404, "Dataset not found")
     @console_ns.response(403, "Permission denied")
@@ -424,93 +421,25 @@ class DatasetApi(Resource):
         if dataset is None:
             raise NotFound("Dataset not found.")
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument(
-                "name",
-                nullable=False,
-                help="type is required. Name must be between 1 to 40 characters.",
-                type=_validate_name,
-            )
-            .add_argument("description", location="json", store_missing=False, type=validate_description_length)
-            .add_argument(
-                "indexing_technique",
-                type=str,
-                location="json",
-                choices=Dataset.INDEXING_TECHNIQUE_LIST,
-                nullable=True,
-                help="Invalid indexing technique.",
-            )
-            .add_argument(
-                "permission",
-                type=str,
-                location="json",
-                choices=(
-                    DatasetPermissionEnum.ONLY_ME,
-                    DatasetPermissionEnum.ALL_TEAM,
-                    DatasetPermissionEnum.PARTIAL_TEAM,
-                ),
-                help="Invalid permission.",
-            )
-            .add_argument("embedding_model", type=str, location="json", help="Invalid embedding model.")
-            .add_argument(
-                "embedding_model_provider", type=str, location="json", help="Invalid embedding model provider."
-            )
-            .add_argument("retrieval_model", type=dict, location="json", help="Invalid retrieval model.")
-            .add_argument("partial_member_list", type=list, location="json", help="Invalid parent user list.")
-            .add_argument(
-                "external_retrieval_model",
-                type=dict,
-                required=False,
-                nullable=True,
-                location="json",
-                help="Invalid external retrieval model.",
-            )
-            .add_argument(
-                "external_knowledge_id",
-                type=str,
-                required=False,
-                nullable=True,
-                location="json",
-                help="Invalid external knowledge id.",
-            )
-            .add_argument(
-                "external_knowledge_api_id",
-                type=str,
-                required=False,
-                nullable=True,
-                location="json",
-                help="Invalid external knowledge api id.",
-            )
-            .add_argument(
-                "icon_info",
-                type=dict,
-                required=False,
-                nullable=True,
-                location="json",
-                help="Invalid icon info.",
-            )
-        )
-        args = parser.parse_args()
-        data = request.get_json()
+        payload = DatasetUpdatePayload.model_validate(console_ns.payload or {})
+        payload_data = payload.model_dump(exclude_unset=True)
         current_user, current_tenant_id = current_account_with_tenant()
-
         # check embedding model setting
         if (
-            data.get("indexing_technique") == "high_quality"
-            and data.get("embedding_model_provider") is not None
-            and data.get("embedding_model") is not None
+            payload.indexing_technique == "high_quality"
+            and payload.embedding_model_provider is not None
+            and payload.embedding_model is not None
         ):
-            DatasetService.check_embedding_model_setting(
-                dataset.tenant_id, data.get("embedding_model_provider"), data.get("embedding_model")
+            is_multimodal = DatasetService.check_is_multimodal_model(
+                dataset.tenant_id, payload.embedding_model_provider, payload.embedding_model
             )
-
+            payload.is_multimodal = is_multimodal
         # The role of the current user in the ta table must be admin, owner, editor, or dataset_operator
         DatasetPermissionService.check_permission(
-            current_user, dataset, data.get("permission"), data.get("partial_member_list")
+            current_user, dataset, payload.permission, payload.partial_member_list
         )
 
-        dataset = DatasetService.update_dataset(dataset_id_str, args, current_user)
+        dataset = DatasetService.update_dataset(dataset_id_str, payload_data, current_user)
 
         if dataset is None:
             raise NotFound("Dataset not found.")
@@ -518,15 +447,10 @@ class DatasetApi(Resource):
         result_data = cast(dict[str, Any], marshal(dataset, dataset_detail_fields))
         tenant_id = current_tenant_id
 
-        if data.get("partial_member_list") and data.get("permission") == "partial_members":
-            DatasetPermissionService.update_partial_member_list(
-                tenant_id, dataset_id_str, data.get("partial_member_list")
-            )
+        if payload.partial_member_list is not None and payload.permission == DatasetPermissionEnum.PARTIAL_TEAM:
+            DatasetPermissionService.update_partial_member_list(tenant_id, dataset_id_str, payload.partial_member_list)
         # clear partial member list when permission is only_me or all_team_members
-        elif (
-            data.get("permission") == DatasetPermissionEnum.ONLY_ME
-            or data.get("permission") == DatasetPermissionEnum.ALL_TEAM
-        ):
+        elif payload.permission in {DatasetPermissionEnum.ONLY_ME, DatasetPermissionEnum.ALL_TEAM}:
             DatasetPermissionService.clear_partial_member_list(dataset_id_str)
 
         partial_member_list = DatasetPermissionService.get_dataset_partial_member_list(dataset_id_str)
@@ -615,24 +539,10 @@ class DatasetIndexingEstimateApi(Resource):
     @setup_required
     @login_required
     @account_initialization_required
+    @console_ns.expect(console_ns.models[IndexingEstimatePayload.__name__])
     def post(self):
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("info_list", type=dict, required=True, nullable=True, location="json")
-            .add_argument("process_rule", type=dict, required=True, nullable=True, location="json")
-            .add_argument(
-                "indexing_technique",
-                type=str,
-                required=True,
-                choices=Dataset.INDEXING_TECHNIQUE_LIST,
-                nullable=True,
-                location="json",
-            )
-            .add_argument("doc_form", type=str, default="text_model", required=False, nullable=False, location="json")
-            .add_argument("dataset_id", type=str, required=False, nullable=False, location="json")
-            .add_argument("doc_language", type=str, default="English", required=False, nullable=False, location="json")
-        )
-        args = parser.parse_args()
+        payload = IndexingEstimatePayload.model_validate(console_ns.payload or {})
+        args = payload.model_dump()
         _, current_tenant_id = current_account_with_tenant()
         # validate args
         DocumentService.estimate_args_validate(args)

api/controllers/console/datasets/datasets_document.py

@@ -6,31 +6,14 @@ from typing import Literal, cast
 
 import sqlalchemy as sa
 from flask import request
-from flask_restx import Resource, fields, marshal, marshal_with, reqparse
+from flask_restx import Resource, fields, marshal, marshal_with
+from pydantic import BaseModel
 from sqlalchemy import asc, desc, select
 from werkzeug.exceptions import Forbidden, NotFound
 
 import services
+from controllers.common.schema import register_schema_models
 from controllers.console import console_ns
-from controllers.console.app.error import (
-    ProviderModelCurrentlyNotSupportError,
-    ProviderNotInitializeError,
-    ProviderQuotaExceededError,
-)
-from controllers.console.datasets.error import (
-    ArchivedDocumentImmutableError,
-    DocumentAlreadyFinishedError,
-    DocumentIndexingError,
-    IndexingEstimateError,
-    InvalidActionError,
-    InvalidMetadataError,
-)
-from controllers.console.wraps import (
-    account_initialization_required,
-    cloud_edition_billing_rate_limit_check,
-    cloud_edition_billing_resource_check,
-    setup_required,
-)
 from core.errors.error import (
     LLMBadRequestError,
     ModelCurrentlyNotSupportError,
@@ -55,10 +38,30 @@ from fields.document_fields import (
 )
 from libs.datetime_utils import naive_utc_now
 from libs.login import current_account_with_tenant, login_required
-from models import Dataset, DatasetProcessRule, Document, DocumentSegment, UploadFile
+from models import DatasetProcessRule, Document, DocumentSegment, UploadFile
 from models.dataset import DocumentPipelineExecutionLog
 from services.dataset_service import DatasetService, DocumentService
-from services.entities.knowledge_entities.knowledge_entities import KnowledgeConfig
+from services.entities.knowledge_entities.knowledge_entities import KnowledgeConfig, ProcessRule, RetrievalModel
+
+from ..app.error import (
+    ProviderModelCurrentlyNotSupportError,
+    ProviderNotInitializeError,
+    ProviderQuotaExceededError,
+)
+from ..datasets.error import (
+    ArchivedDocumentImmutableError,
+    DocumentAlreadyFinishedError,
+    DocumentIndexingError,
+    IndexingEstimateError,
+    InvalidActionError,
+    InvalidMetadataError,
+)
+from ..wraps import (
+    account_initialization_required,
+    cloud_edition_billing_rate_limit_check,
+    cloud_edition_billing_resource_check,
+    setup_required,
+)
 
 logger = logging.getLogger(__name__)
 
@@ -93,6 +96,24 @@ dataset_and_document_fields_copy["documents"] = fields.List(fields.Nested(docume
 dataset_and_document_model = _get_or_create_model("DatasetAndDocument", dataset_and_document_fields_copy)
 
 
+class DocumentRetryPayload(BaseModel):
+    document_ids: list[str]
+
+
+class DocumentRenamePayload(BaseModel):
+    name: str
+
+
+register_schema_models(
+    console_ns,
+    KnowledgeConfig,
+    ProcessRule,
+    RetrievalModel,
+    DocumentRetryPayload,
+    DocumentRenamePayload,
+)
+
+
 class DocumentResource(Resource):
     def get_document(self, dataset_id: str, document_id: str) -> Document:
         current_user, current_tenant_id = current_account_with_tenant()
@@ -201,8 +222,9 @@ class DatasetDocumentListApi(Resource):
     @setup_required
     @login_required
     @account_initialization_required
-    def get(self, dataset_id: str):
+    def get(self, dataset_id):
         current_user, current_tenant_id = current_account_with_tenant()
+        dataset_id = str(dataset_id)
         page = request.args.get("page", default=1, type=int)
         limit = request.args.get("limit", default=20, type=int)
         search = request.args.get("keyword", default=None, type=str)
@@ -310,6 +332,7 @@ class DatasetDocumentListApi(Resource):
     @marshal_with(dataset_and_document_model)
     @cloud_edition_billing_resource_check("vector_space")
     @cloud_edition_billing_rate_limit_check("knowledge")
+    @console_ns.expect(console_ns.models[KnowledgeConfig.__name__])
     def post(self, dataset_id):
         current_user, _ = current_account_with_tenant()
         dataset_id = str(dataset_id)
@@ -328,23 +351,7 @@ class DatasetDocumentListApi(Resource):
         except services.errors.account.NoPermissionError as e:
             raise Forbidden(str(e))
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument(
-                "indexing_technique", type=str, choices=Dataset.INDEXING_TECHNIQUE_LIST, nullable=False, location="json"
-            )
-            .add_argument("data_source", type=dict, required=False, location="json")
-            .add_argument("process_rule", type=dict, required=False, location="json")
-            .add_argument("duplicate", type=bool, default=True, nullable=False, location="json")
-            .add_argument("original_document_id", type=str, required=False, location="json")
-            .add_argument("doc_form", type=str, default="text_model", required=False, nullable=False, location="json")
-            .add_argument("retrieval_model", type=dict, required=False, nullable=False, location="json")
-            .add_argument("embedding_model", type=str, required=False, nullable=True, location="json")
-            .add_argument("embedding_model_provider", type=str, required=False, nullable=True, location="json")
-            .add_argument("doc_language", type=str, default="English", required=False, nullable=False, location="json")
-        )
-        args = parser.parse_args()
-        knowledge_config = KnowledgeConfig.model_validate(args)
+        knowledge_config = KnowledgeConfig.model_validate(console_ns.payload or {})
 
         if not dataset.indexing_technique and not knowledge_config.indexing_technique:
             raise ValueError("indexing_technique is required.")
@@ -390,17 +397,7 @@ class DatasetDocumentListApi(Resource):
 class DatasetInitApi(Resource):
     @console_ns.doc("init_dataset")
     @console_ns.doc(description="Initialize dataset with documents")
-    @console_ns.expect(
-        console_ns.model(
-            "DatasetInitRequest",
-            {
-                "upload_file_id": fields.String(required=True, description="Upload file ID"),
-                "indexing_technique": fields.String(description="Indexing technique"),
-                "process_rule": fields.Raw(description="Processing rules"),
-                "data_source": fields.Raw(description="Data source configuration"),
-            },
-        )
-    )
+    @console_ns.expect(console_ns.models[KnowledgeConfig.__name__])
     @console_ns.response(201, "Dataset initialized successfully", dataset_and_document_model)
     @console_ns.response(400, "Invalid request parameters")
     @setup_required
@@ -415,27 +412,7 @@ class DatasetInitApi(Resource):
         if not current_user.is_dataset_editor:
             raise Forbidden()
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument(
-                "indexing_technique",
-                type=str,
-                choices=Dataset.INDEXING_TECHNIQUE_LIST,
-                required=True,
-                nullable=False,
-                location="json",
-            )
-            .add_argument("data_source", type=dict, required=True, nullable=True, location="json")
-            .add_argument("process_rule", type=dict, required=True, nullable=True, location="json")
-            .add_argument("doc_form", type=str, default="text_model", required=False, nullable=False, location="json")
-            .add_argument("doc_language", type=str, default="English", required=False, nullable=False, location="json")
-            .add_argument("retrieval_model", type=dict, required=False, nullable=False, location="json")
-            .add_argument("embedding_model", type=str, required=False, nullable=True, location="json")
-            .add_argument("embedding_model_provider", type=str, required=False, nullable=True, location="json")
-        )
-        args = parser.parse_args()
-
-        knowledge_config = KnowledgeConfig.model_validate(args)
+        knowledge_config = KnowledgeConfig.model_validate(console_ns.payload or {})
         if knowledge_config.indexing_technique == "high_quality":
             if knowledge_config.embedding_model is None or knowledge_config.embedding_model_provider is None:
                 raise ValueError("embedding model and embedding model provider are required for high quality indexing.")
@@ -443,10 +420,14 @@ class DatasetInitApi(Resource):
                 model_manager = ModelManager()
                 model_manager.get_model_instance(
                     tenant_id=current_tenant_id,
-                    provider=args["embedding_model_provider"],
+                    provider=knowledge_config.embedding_model_provider,
                     model_type=ModelType.TEXT_EMBEDDING,
-                    model=args["embedding_model"],
+                    model=knowledge_config.embedding_model,
                 )
+                is_multimodal = DatasetService.check_is_multimodal_model(
+                    current_tenant_id, knowledge_config.embedding_model_provider, knowledge_config.embedding_model
+                )
+                knowledge_config.is_multimodal = is_multimodal
             except InvokeAuthorizationError:
                 raise ProviderNotInitializeError(
                     "No Embedding Model available. Please configure a valid provider in the Settings -> Model Provider."
@@ -1076,19 +1057,16 @@ class DocumentRetryApi(DocumentResource):
     @login_required
     @account_initialization_required
     @cloud_edition_billing_rate_limit_check("knowledge")
+    @console_ns.expect(console_ns.models[DocumentRetryPayload.__name__])
     def post(self, dataset_id):
         """retry document."""
-
-        parser = reqparse.RequestParser().add_argument(
-            "document_ids", type=list, required=True, nullable=False, location="json"
-        )
-        args = parser.parse_args()
+        payload = DocumentRetryPayload.model_validate(console_ns.payload or {})
         dataset_id = str(dataset_id)
         dataset = DatasetService.get_dataset(dataset_id)
         retry_documents = []
         if not dataset:
             raise NotFound("Dataset not found.")
-        for document_id in args["document_ids"]:
+        for document_id in payload.document_ids:
             try:
                 document_id = str(document_id)
 
@@ -1121,6 +1099,7 @@ class DocumentRenameApi(DocumentResource):
     @login_required
     @account_initialization_required
     @marshal_with(document_fields)
+    @console_ns.expect(console_ns.models[DocumentRenamePayload.__name__])
     def post(self, dataset_id, document_id):
         # The role of the current user in the ta table must be admin, owner, editor, or dataset_operator
         current_user, _ = current_account_with_tenant()
@@ -1130,11 +1109,10 @@ class DocumentRenameApi(DocumentResource):
         if not dataset:
             raise NotFound("Dataset not found.")
         DatasetService.check_dataset_operator_permission(current_user, dataset)
-        parser = reqparse.RequestParser().add_argument("name", type=str, required=True, nullable=False, location="json")
-        args = parser.parse_args()
+        payload = DocumentRenamePayload.model_validate(console_ns.payload or {})
 
         try:
-            document = DocumentService.rename_document(dataset_id, document_id, args["name"])
+            document = DocumentService.rename_document(dataset_id, document_id, payload.name)
         except services.errors.document.DocumentIndexingError:
             raise DocumentIndexingError("Cannot delete document during indexing.")
 

api/controllers/console/datasets/datasets_segments.py

@@ -1,11 +1,13 @@
 import uuid
 
 from flask import request
-from flask_restx import Resource, marshal, reqparse
+from flask_restx import Resource, marshal
+from pydantic import BaseModel, Field
 from sqlalchemy import select
 from werkzeug.exceptions import Forbidden, NotFound
 
 import services
+from controllers.common.schema import register_schema_models
 from controllers.console import console_ns
 from controllers.console.app.error import ProviderNotInitializeError
 from controllers.console.datasets.error import (
@@ -36,6 +38,58 @@ from services.errors.chunk import ChildChunkIndexingError as ChildChunkIndexingS
 from tasks.batch_create_segment_to_index_task import batch_create_segment_to_index_task
 
 
+class SegmentListQuery(BaseModel):
+    limit: int = Field(default=20, ge=1, le=100)
+    status: list[str] = Field(default_factory=list)
+    hit_count_gte: int | None = None
+    enabled: str = Field(default="all")
+    keyword: str | None = None
+    page: int = Field(default=1, ge=1)
+
+
+class SegmentCreatePayload(BaseModel):
+    content: str
+    answer: str | None = None
+    keywords: list[str] | None = None
+    attachment_ids: list[str] | None = None
+
+
+class SegmentUpdatePayload(BaseModel):
+    content: str
+    answer: str | None = None
+    keywords: list[str] | None = None
+    regenerate_child_chunks: bool = False
+    attachment_ids: list[str] | None = None
+
+
+class BatchImportPayload(BaseModel):
+    upload_file_id: str
+
+
+class ChildChunkCreatePayload(BaseModel):
+    content: str
+
+
+class ChildChunkUpdatePayload(BaseModel):
+    content: str
+
+
+class ChildChunkBatchUpdatePayload(BaseModel):
+    chunks: list[ChildChunkUpdateArgs]
+
+
+register_schema_models(
+    console_ns,
+    SegmentListQuery,
+    SegmentCreatePayload,
+    SegmentUpdatePayload,
+    BatchImportPayload,
+    ChildChunkCreatePayload,
+    ChildChunkUpdatePayload,
+    ChildChunkBatchUpdatePayload,
+)
+
+
 @console_ns.route("/datasets/<uuid:dataset_id>/documents/<uuid:document_id>/segments")
 class DatasetDocumentSegmentListApi(Resource):
     @setup_required
@@ -60,23 +114,18 @@ class DatasetDocumentSegmentListApi(Resource):
         if not document:
             raise NotFound("Document not found.")
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("limit", type=int, default=20, location="args")
-            .add_argument("status", type=str, action="append", default=[], location="args")
-            .add_argument("hit_count_gte", type=int, default=None, location="args")
-            .add_argument("enabled", type=str, default="all", location="args")
-            .add_argument("keyword", type=str, default=None, location="args")
-            .add_argument("page", type=int, default=1, location="args")
+        args = SegmentListQuery.model_validate(
+            {
+                **request.args.to_dict(),
+                "status": request.args.getlist("status"),
+            }
         )
 
-        args = parser.parse_args()
-
-        page = args["page"]
-        limit = min(args["limit"], 100)
-        status_list = args["status"]
-        hit_count_gte = args["hit_count_gte"]
-        keyword = args["keyword"]
+        page = args.page
+        limit = min(args.limit, 100)
+        status_list = args.status
+        hit_count_gte = args.hit_count_gte
+        keyword = args.keyword
 
         query = (
             select(DocumentSegment)
@@ -96,10 +145,10 @@ class DatasetDocumentSegmentListApi(Resource):
         if keyword:
             query = query.where(DocumentSegment.content.ilike(f"%{keyword}%"))
 
-        if args["enabled"].lower() != "all":
-            if args["enabled"].lower() == "true":
+        if args.enabled.lower() != "all":
+            if args.enabled.lower() == "true":
                 query = query.where(DocumentSegment.enabled == True)
-            elif args["enabled"].lower() == "false":
+            elif args.enabled.lower() == "false":
                 query = query.where(DocumentSegment.enabled == False)
 
         segments = db.paginate(select=query, page=page, per_page=limit, max_per_page=100, error_out=False)
@@ -210,6 +259,7 @@ class DatasetDocumentSegmentAddApi(Resource):
     @cloud_edition_billing_resource_check("vector_space")
     @cloud_edition_billing_knowledge_limit_check("add_segment")
     @cloud_edition_billing_rate_limit_check("knowledge")
+    @console_ns.expect(console_ns.models[SegmentCreatePayload.__name__])
     def post(self, dataset_id, document_id):
         current_user, current_tenant_id = current_account_with_tenant()
 
@@ -246,15 +296,10 @@ class DatasetDocumentSegmentAddApi(Resource):
         except services.errors.account.NoPermissionError as e:
             raise Forbidden(str(e))
         # validate args
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("content", type=str, required=True, nullable=False, location="json")
-            .add_argument("answer", type=str, required=False, nullable=True, location="json")
-            .add_argument("keywords", type=list, required=False, nullable=True, location="json")
-        )
-        args = parser.parse_args()
-        SegmentService.segment_create_args_validate(args, document)
-        segment = SegmentService.create_segment(args, document, dataset)
+        payload = SegmentCreatePayload.model_validate(console_ns.payload or {})
+        payload_dict = payload.model_dump(exclude_none=True)
+        SegmentService.segment_create_args_validate(payload_dict, document)
+        segment = SegmentService.create_segment(payload_dict, document, dataset)
         return {"data": marshal(segment, segment_fields), "doc_form": document.doc_form}, 200
 
 
@@ -265,6 +310,7 @@ class DatasetDocumentSegmentUpdateApi(Resource):
     @account_initialization_required
     @cloud_edition_billing_resource_check("vector_space")
     @cloud_edition_billing_rate_limit_check("knowledge")
+    @console_ns.expect(console_ns.models[SegmentUpdatePayload.__name__])
     def patch(self, dataset_id, document_id, segment_id):
         current_user, current_tenant_id = current_account_with_tenant()
 
@@ -313,18 +359,12 @@ class DatasetDocumentSegmentUpdateApi(Resource):
         except services.errors.account.NoPermissionError as e:
             raise Forbidden(str(e))
         # validate args
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("content", type=str, required=True, nullable=False, location="json")
-            .add_argument("answer", type=str, required=False, nullable=True, location="json")
-            .add_argument("keywords", type=list, required=False, nullable=True, location="json")
-            .add_argument(
-                "regenerate_child_chunks", type=bool, required=False, nullable=True, default=False, location="json"
-            )
+        payload = SegmentUpdatePayload.model_validate(console_ns.payload or {})
+        payload_dict = payload.model_dump(exclude_none=True)
+        SegmentService.segment_create_args_validate(payload_dict, document)
+        segment = SegmentService.update_segment(
+            SegmentUpdateArgs.model_validate(payload.model_dump(exclude_none=True)), segment, document, dataset
         )
-        args = parser.parse_args()
-        SegmentService.segment_create_args_validate(args, document)
-        segment = SegmentService.update_segment(SegmentUpdateArgs.model_validate(args), segment, document, dataset)
         return {"data": marshal(segment, segment_fields), "doc_form": document.doc_form}, 200
 
     @setup_required
@@ -377,6 +417,7 @@ class DatasetDocumentSegmentBatchImportApi(Resource):
     @cloud_edition_billing_resource_check("vector_space")
     @cloud_edition_billing_knowledge_limit_check("add_segment")
     @cloud_edition_billing_rate_limit_check("knowledge")
+    @console_ns.expect(console_ns.models[BatchImportPayload.__name__])
     def post(self, dataset_id, document_id):
         current_user, current_tenant_id = current_account_with_tenant()
 
@@ -391,11 +432,8 @@ class DatasetDocumentSegmentBatchImportApi(Resource):
         if not document:
             raise NotFound("Document not found.")
 
-        parser = reqparse.RequestParser().add_argument(
-            "upload_file_id", type=str, required=True, nullable=False, location="json"
-        )
-        args = parser.parse_args()
-        upload_file_id = args["upload_file_id"]
+        payload = BatchImportPayload.model_validate(console_ns.payload or {})
+        upload_file_id = payload.upload_file_id
 
         upload_file = db.session.query(UploadFile).where(UploadFile.id == upload_file_id).first()
         if not upload_file:
@@ -446,6 +484,7 @@ class ChildChunkAddApi(Resource):
     @cloud_edition_billing_resource_check("vector_space")
     @cloud_edition_billing_knowledge_limit_check("add_segment")
     @cloud_edition_billing_rate_limit_check("knowledge")
+    @console_ns.expect(console_ns.models[ChildChunkCreatePayload.__name__])
     def post(self, dataset_id, document_id, segment_id):
         current_user, current_tenant_id = current_account_with_tenant()
 
@@ -491,13 +530,9 @@ class ChildChunkAddApi(Resource):
         except services.errors.account.NoPermissionError as e:
             raise Forbidden(str(e))
         # validate args
-        parser = reqparse.RequestParser().add_argument(
-            "content", type=str, required=True, nullable=False, location="json"
-        )
-        args = parser.parse_args()
         try:
-            content = args["content"]
-            child_chunk = SegmentService.create_child_chunk(content, segment, document, dataset)
+            payload = ChildChunkCreatePayload.model_validate(console_ns.payload or {})
+            child_chunk = SegmentService.create_child_chunk(payload.content, segment, document, dataset)
         except ChildChunkIndexingServiceError as e:
             raise ChildChunkIndexingError(str(e))
         return {"data": marshal(child_chunk, child_chunk_fields)}, 200
@@ -529,18 +564,17 @@ class ChildChunkAddApi(Resource):
         )
         if not segment:
             raise NotFound("Segment not found.")
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("limit", type=int, default=20, location="args")
-            .add_argument("keyword", type=str, default=None, location="args")
-            .add_argument("page", type=int, default=1, location="args")
+        args = SegmentListQuery.model_validate(
+            {
+                "limit": request.args.get("limit", default=20, type=int),
+                "keyword": request.args.get("keyword"),
+                "page": request.args.get("page", default=1, type=int),
+            }
         )
 
-        args = parser.parse_args()
-
-        page = args["page"]
-        limit = min(args["limit"], 100)
-        keyword = args["keyword"]
+        page = args.page
+        limit = min(args.limit, 100)
+        keyword = args.keyword
 
         child_chunks = SegmentService.get_child_chunks(segment_id, document_id, dataset_id, page, limit, keyword)
         return {
@@ -588,14 +622,9 @@ class ChildChunkAddApi(Resource):
         except services.errors.account.NoPermissionError as e:
             raise Forbidden(str(e))
         # validate args
-        parser = reqparse.RequestParser().add_argument(
-            "chunks", type=list, required=True, nullable=False, location="json"
-        )
-        args = parser.parse_args()
+        payload = ChildChunkBatchUpdatePayload.model_validate(console_ns.payload or {})
         try:
-            chunks_data = args["chunks"]
-            chunks = [ChildChunkUpdateArgs.model_validate(chunk) for chunk in chunks_data]
-            child_chunks = SegmentService.update_child_chunks(chunks, segment, document, dataset)
+            child_chunks = SegmentService.update_child_chunks(payload.chunks, segment, document, dataset)
         except ChildChunkIndexingServiceError as e:
             raise ChildChunkIndexingError(str(e))
         return {"data": marshal(child_chunks, child_chunk_fields)}, 200
@@ -665,6 +694,7 @@ class ChildChunkUpdateApi(Resource):
     @account_initialization_required
     @cloud_edition_billing_resource_check("vector_space")
     @cloud_edition_billing_rate_limit_check("knowledge")
+    @console_ns.expect(console_ns.models[ChildChunkUpdatePayload.__name__])
     def patch(self, dataset_id, document_id, segment_id, child_chunk_id):
         current_user, current_tenant_id = current_account_with_tenant()
 
@@ -711,13 +741,9 @@ class ChildChunkUpdateApi(Resource):
         except services.errors.account.NoPermissionError as e:
             raise Forbidden(str(e))
         # validate args
-        parser = reqparse.RequestParser().add_argument(
-            "content", type=str, required=True, nullable=False, location="json"
-        )
-        args = parser.parse_args()
         try:
-            content = args["content"]
-            child_chunk = SegmentService.update_child_chunk(content, child_chunk, segment, document, dataset)
+            payload = ChildChunkUpdatePayload.model_validate(console_ns.payload or {})
+            child_chunk = SegmentService.update_child_chunk(payload.content, child_chunk, segment, document, dataset)
         except ChildChunkIndexingServiceError as e:
             raise ChildChunkIndexingError(str(e))
         return {"data": marshal(child_chunk, child_chunk_fields)}, 200

api/controllers/console/datasets/external.py

@@ -1,8 +1,10 @@
 from flask import request
-from flask_restx import Resource, fields, marshal, reqparse
+from flask_restx import Resource, fields, marshal
+from pydantic import BaseModel, Field
 from werkzeug.exceptions import Forbidden, InternalServerError, NotFound
 
 import services
+from controllers.common.schema import register_schema_models
 from controllers.console import console_ns
 from controllers.console.datasets.error import DatasetNameDuplicateError
 from controllers.console.wraps import account_initialization_required, edit_permission_required, setup_required
@@ -71,10 +73,38 @@ except KeyError:
     dataset_detail_model = _build_dataset_detail_model()
 
 
-def _validate_name(name: str) -> str:
-    if not name or len(name) < 1 or len(name) > 100:
-        raise ValueError("Name must be between 1 to 100 characters.")
-    return name
+class ExternalKnowledgeApiPayload(BaseModel):
+    name: str = Field(..., min_length=1, max_length=40)
+    settings: dict[str, object]
+
+
+class ExternalDatasetCreatePayload(BaseModel):
+    external_knowledge_api_id: str
+    external_knowledge_id: str
+    name: str = Field(..., min_length=1, max_length=40)
+    description: str | None = Field(None, max_length=400)
+    external_retrieval_model: dict[str, object] | None = None
+
+
+class ExternalHitTestingPayload(BaseModel):
+    query: str
+    external_retrieval_model: dict[str, object] | None = None
+    metadata_filtering_conditions: dict[str, object] | None = None
+
+
+class BedrockRetrievalPayload(BaseModel):
+    retrieval_setting: dict[str, object]
+    query: str
+    knowledge_id: str
+
+
+register_schema_models(
+    console_ns,
+    ExternalKnowledgeApiPayload,
+    ExternalDatasetCreatePayload,
+    ExternalHitTestingPayload,
+    BedrockRetrievalPayload,
+)
 
 
 @console_ns.route("/datasets/external-knowledge-api")
@@ -113,28 +143,12 @@ class ExternalApiTemplateListApi(Resource):
     @setup_required
     @login_required
     @account_initialization_required
+    @console_ns.expect(console_ns.models[ExternalKnowledgeApiPayload.__name__])
     def post(self):
         current_user, current_tenant_id = current_account_with_tenant()
-        parser = (
-            reqparse.RequestParser()
-            .add_argument(
-                "name",
-                nullable=False,
-                required=True,
-                help="Name is required. Name must be between 1 to 100 characters.",
-                type=_validate_name,
-            )
-            .add_argument(
-                "settings",
-                type=dict,
-                location="json",
-                nullable=False,
-                required=True,
-            )
-        )
-        args = parser.parse_args()
+        payload = ExternalKnowledgeApiPayload.model_validate(console_ns.payload or {})
 
-        ExternalDatasetService.validate_api_list(args["settings"])
+        ExternalDatasetService.validate_api_list(payload.settings)
 
         # The role of the current user in the ta table must be admin, owner, or editor, or dataset_operator
         if not current_user.is_dataset_editor:
@@ -142,7 +156,7 @@ class ExternalApiTemplateListApi(Resource):
 
         try:
             external_knowledge_api = ExternalDatasetService.create_external_knowledge_api(
-                tenant_id=current_tenant_id, user_id=current_user.id, args=args
+                tenant_id=current_tenant_id, user_id=current_user.id, args=payload.model_dump()
             )
         except services.errors.dataset.DatasetNameDuplicateError:
             raise DatasetNameDuplicateError()
@@ -171,35 +185,19 @@ class ExternalApiTemplateApi(Resource):
     @setup_required
     @login_required
     @account_initialization_required
+    @console_ns.expect(console_ns.models[ExternalKnowledgeApiPayload.__name__])
     def patch(self, external_knowledge_api_id):
         current_user, current_tenant_id = current_account_with_tenant()
         external_knowledge_api_id = str(external_knowledge_api_id)
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument(
-                "name",
-                nullable=False,
-                required=True,
-                help="type is required. Name must be between 1 to 100 characters.",
-                type=_validate_name,
-            )
-            .add_argument(
-                "settings",
-                type=dict,
-                location="json",
-                nullable=False,
-                required=True,
-            )
-        )
-        args = parser.parse_args()
-        ExternalDatasetService.validate_api_list(args["settings"])
+        payload = ExternalKnowledgeApiPayload.model_validate(console_ns.payload or {})
+        ExternalDatasetService.validate_api_list(payload.settings)
 
         external_knowledge_api = ExternalDatasetService.update_external_knowledge_api(
             tenant_id=current_tenant_id,
             user_id=current_user.id,
             external_knowledge_api_id=external_knowledge_api_id,
-            args=args,
+            args=payload.model_dump(),
         )
 
         return external_knowledge_api.to_dict(), 200
@@ -240,17 +238,7 @@ class ExternalApiUseCheckApi(Resource):
 class ExternalDatasetCreateApi(Resource):
     @console_ns.doc("create_external_dataset")
     @console_ns.doc(description="Create external knowledge dataset")
-    @console_ns.expect(
-        console_ns.model(
-            "CreateExternalDatasetRequest",
-            {
-                "external_knowledge_api_id": fields.String(required=True, description="External knowledge API ID"),
-                "external_knowledge_id": fields.String(required=True, description="External knowledge ID"),
-                "name": fields.String(required=True, description="Dataset name"),
-                "description": fields.String(description="Dataset description"),
-            },
-        )
-    )
+    @console_ns.expect(console_ns.models[ExternalDatasetCreatePayload.__name__])
     @console_ns.response(201, "External dataset created successfully", dataset_detail_model)
     @console_ns.response(400, "Invalid parameters")
     @console_ns.response(403, "Permission denied")
@@ -261,22 +249,8 @@ class ExternalDatasetCreateApi(Resource):
     def post(self):
         # The role of the current user in the ta table must be admin, owner, or editor
         current_user, current_tenant_id = current_account_with_tenant()
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("external_knowledge_api_id", type=str, required=True, nullable=False, location="json")
-            .add_argument("external_knowledge_id", type=str, required=True, nullable=False, location="json")
-            .add_argument(
-                "name",
-                nullable=False,
-                required=True,
-                help="name is required. Name must be between 1 to 100 characters.",
-                type=_validate_name,
-            )
-            .add_argument("description", type=str, required=False, nullable=True, location="json")
-            .add_argument("external_retrieval_model", type=dict, required=False, location="json")
-        )
-
-        args = parser.parse_args()
+        payload = ExternalDatasetCreatePayload.model_validate(console_ns.payload or {})
+        args = payload.model_dump(exclude_none=True)
 
         # The role of the current user in the ta table must be admin, owner, or editor, or dataset_operator
         if not current_user.is_dataset_editor:
@@ -299,16 +273,7 @@ class ExternalKnowledgeHitTestingApi(Resource):
     @console_ns.doc("test_external_knowledge_retrieval")
     @console_ns.doc(description="Test external knowledge retrieval for dataset")
     @console_ns.doc(params={"dataset_id": "Dataset ID"})
-    @console_ns.expect(
-        console_ns.model(
-            "ExternalHitTestingRequest",
-            {
-                "query": fields.String(required=True, description="Query text for testing"),
-                "retrieval_model": fields.Raw(description="Retrieval model configuration"),
-                "external_retrieval_model": fields.Raw(description="External retrieval model configuration"),
-            },
-        )
-    )
+    @console_ns.expect(console_ns.models[ExternalHitTestingPayload.__name__])
     @console_ns.response(200, "External hit testing completed successfully")
     @console_ns.response(404, "Dataset not found")
     @console_ns.response(400, "Invalid parameters")
@@ -327,23 +292,16 @@ class ExternalKnowledgeHitTestingApi(Resource):
         except services.errors.account.NoPermissionError as e:
             raise Forbidden(str(e))
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("query", type=str, location="json")
-            .add_argument("external_retrieval_model", type=dict, required=False, location="json")
-            .add_argument("metadata_filtering_conditions", type=dict, required=False, location="json")
-        )
-        args = parser.parse_args()
-
-        HitTestingService.hit_testing_args_check(args)
+        payload = ExternalHitTestingPayload.model_validate(console_ns.payload or {})
+        HitTestingService.hit_testing_args_check(payload.model_dump())
 
         try:
             response = HitTestingService.external_retrieve(
                 dataset=dataset,
-                query=args["query"],
+                query=payload.query,
                 account=current_user,
-                external_retrieval_model=args["external_retrieval_model"],
-                metadata_filtering_conditions=args["metadata_filtering_conditions"],
+                external_retrieval_model=payload.external_retrieval_model,
+                metadata_filtering_conditions=payload.metadata_filtering_conditions,
             )
 
             return response
@@ -356,33 +314,13 @@ class BedrockRetrievalApi(Resource):
     # this api is only for internal testing
     @console_ns.doc("bedrock_retrieval_test")
     @console_ns.doc(description="Bedrock retrieval test (internal use only)")
-    @console_ns.expect(
-        console_ns.model(
-            "BedrockRetrievalTestRequest",
-            {
-                "retrieval_setting": fields.Raw(required=True, description="Retrieval settings"),
-                "query": fields.String(required=True, description="Query text"),
-                "knowledge_id": fields.String(required=True, description="Knowledge ID"),
-            },
-        )
-    )
+    @console_ns.expect(console_ns.models[BedrockRetrievalPayload.__name__])
     @console_ns.response(200, "Bedrock retrieval test completed")
     def post(self):
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("retrieval_setting", nullable=False, required=True, type=dict, location="json")
-            .add_argument(
-                "query",
-                nullable=False,
-                required=True,
-                type=str,
-            )
-            .add_argument("knowledge_id", nullable=False, required=True, type=str)
-        )
-        args = parser.parse_args()
+        payload = BedrockRetrievalPayload.model_validate(console_ns.payload or {})
 
         # Call the knowledge retrieval service
         result = ExternalDatasetTestService.knowledge_retrieval(
-            args["retrieval_setting"], args["query"], args["knowledge_id"]
+            payload.retrieval_setting, payload.query, payload.knowledge_id
         )
         return result, 200

api/controllers/console/datasets/hit_testing.py

@@ -1,13 +1,17 @@
-from flask_restx import Resource, fields
+from flask_restx import Resource
 
-from controllers.console import console_ns
-from controllers.console.datasets.hit_testing_base import DatasetsHitTestingBase
-from controllers.console.wraps import (
+from controllers.common.schema import register_schema_model
+from libs.login import login_required
+
+from .. import console_ns
+from ..datasets.hit_testing_base import DatasetsHitTestingBase, HitTestingPayload
+from ..wraps import (
     account_initialization_required,
     cloud_edition_billing_rate_limit_check,
     setup_required,
 )
-from libs.login import login_required
+
+register_schema_model(console_ns, HitTestingPayload)
 
 
 @console_ns.route("/datasets/<uuid:dataset_id>/hit-testing")
@@ -15,17 +19,7 @@ class HitTestingApi(Resource, DatasetsHitTestingBase):
     @console_ns.doc("test_dataset_retrieval")
     @console_ns.doc(description="Test dataset knowledge retrieval")
     @console_ns.doc(params={"dataset_id": "Dataset ID"})
-    @console_ns.expect(
-        console_ns.model(
-            "HitTestingRequest",
-            {
-                "query": fields.String(required=True, description="Query text for testing"),
-                "retrieval_model": fields.Raw(description="Retrieval model configuration"),
-                "top_k": fields.Integer(description="Number of top results to return"),
-                "score_threshold": fields.Float(description="Score threshold for filtering results"),
-            },
-        )
-    )
+    @console_ns.expect(console_ns.models[HitTestingPayload.__name__])
     @console_ns.response(200, "Hit testing completed successfully")
     @console_ns.response(404, "Dataset not found")
     @console_ns.response(400, "Invalid parameters")
@@ -37,7 +31,8 @@ class HitTestingApi(Resource, DatasetsHitTestingBase):
         dataset_id_str = str(dataset_id)
 
         dataset = self.get_and_validate_dataset(dataset_id_str)
-        args = self.parse_args()
+        payload = HitTestingPayload.model_validate(console_ns.payload or {})
+        args = payload.model_dump(exclude_none=True)
         self.hit_testing_args_check(args)
 
         return self.perform_hit_testing(dataset, args)

api/controllers/console/datasets/hit_testing_base.py

@@ -1,6 +1,8 @@
 import logging
+from typing import Any
 
 from flask_restx import marshal, reqparse
+from pydantic import BaseModel, Field
 from werkzeug.exceptions import Forbidden, InternalServerError, NotFound
 
 import services
@@ -27,6 +29,13 @@ from services.hit_testing_service import HitTestingService
 logger = logging.getLogger(__name__)
 
 
+class HitTestingPayload(BaseModel):
+    query: str = Field(max_length=250)
+    retrieval_model: dict[str, Any] | None = None
+    external_retrieval_model: dict[str, Any] | None = None
+    attachment_ids: list[str] | None = None
+
+
 class DatasetsHitTestingBase:
     @staticmethod
     def get_and_validate_dataset(dataset_id: str):
@@ -43,14 +52,15 @@ class DatasetsHitTestingBase:
         return dataset
 
     @staticmethod
-    def hit_testing_args_check(args):
+    def hit_testing_args_check(args: dict[str, Any]):
         HitTestingService.hit_testing_args_check(args)
 
     @staticmethod
     def parse_args():
         parser = (
             reqparse.RequestParser()
-            .add_argument("query", type=str, location="json")
+            .add_argument("query", type=str, required=False, location="json")
+            .add_argument("attachment_ids", type=list, required=False, location="json")
             .add_argument("retrieval_model", type=dict, required=False, location="json")
             .add_argument("external_retrieval_model", type=dict, required=False, location="json")
         )
@@ -62,10 +72,11 @@ class DatasetsHitTestingBase:
         try:
             response = HitTestingService.retrieve(
                 dataset=dataset,
-                query=args["query"],
+                query=args.get("query"),
                 account=current_user,
-                retrieval_model=args["retrieval_model"],
-                external_retrieval_model=args["external_retrieval_model"],
+                retrieval_model=args.get("retrieval_model"),
+                external_retrieval_model=args.get("external_retrieval_model"),
+                attachment_ids=args.get("attachment_ids"),
                 limit=10,
             )
             return {"query": response["query"], "records": marshal(response["records"], hit_testing_record_fields)}

api/controllers/console/datasets/metadata.py

@@ -1,8 +1,10 @@
 from typing import Literal
 
-from flask_restx import Resource, marshal_with, reqparse
+from flask_restx import Resource, marshal_with
+from pydantic import BaseModel
 from werkzeug.exceptions import NotFound
 
+from controllers.common.schema import register_schema_model, register_schema_models
 from controllers.console import console_ns
 from controllers.console.wraps import account_initialization_required, enterprise_license_required, setup_required
 from fields.dataset_fields import dataset_metadata_fields
@@ -15,6 +17,14 @@ from services.entities.knowledge_entities.knowledge_entities import (
 from services.metadata_service import MetadataService
 
 
+class MetadataUpdatePayload(BaseModel):
+    name: str
+
+
+register_schema_models(console_ns, MetadataArgs, MetadataOperationData)
+register_schema_model(console_ns, MetadataUpdatePayload)
+
+
 @console_ns.route("/datasets/<uuid:dataset_id>/metadata")
 class DatasetMetadataCreateApi(Resource):
     @setup_required
@@ -22,15 +32,10 @@ class DatasetMetadataCreateApi(Resource):
     @account_initialization_required
     @enterprise_license_required
     @marshal_with(dataset_metadata_fields)
+    @console_ns.expect(console_ns.models[MetadataArgs.__name__])
     def post(self, dataset_id):
         current_user, _ = current_account_with_tenant()
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("type", type=str, required=True, nullable=False, location="json")
-            .add_argument("name", type=str, required=True, nullable=False, location="json")
-        )
-        args = parser.parse_args()
-        metadata_args = MetadataArgs.model_validate(args)
+        metadata_args = MetadataArgs.model_validate(console_ns.payload or {})
 
         dataset_id_str = str(dataset_id)
         dataset = DatasetService.get_dataset(dataset_id_str)
@@ -60,11 +65,11 @@ class DatasetMetadataApi(Resource):
     @account_initialization_required
     @enterprise_license_required
     @marshal_with(dataset_metadata_fields)
+    @console_ns.expect(console_ns.models[MetadataUpdatePayload.__name__])
     def patch(self, dataset_id, metadata_id):
         current_user, _ = current_account_with_tenant()
-        parser = reqparse.RequestParser().add_argument("name", type=str, required=True, nullable=False, location="json")
-        args = parser.parse_args()
-        name = args["name"]
+        payload = MetadataUpdatePayload.model_validate(console_ns.payload or {})
+        name = payload.name
 
         dataset_id_str = str(dataset_id)
         metadata_id_str = str(metadata_id)
@@ -131,6 +136,7 @@ class DocumentMetadataEditApi(Resource):
     @login_required
     @account_initialization_required
     @enterprise_license_required
+    @console_ns.expect(console_ns.models[MetadataOperationData.__name__])
     def post(self, dataset_id):
         current_user, _ = current_account_with_tenant()
         dataset_id_str = str(dataset_id)
@@ -139,11 +145,7 @@ class DocumentMetadataEditApi(Resource):
             raise NotFound("Dataset not found.")
         DatasetService.check_dataset_permission(dataset, current_user)
 
-        parser = reqparse.RequestParser().add_argument(
-            "operation_data", type=list, required=True, nullable=False, location="json"
-        )
-        args = parser.parse_args()
-        metadata_args = MetadataOperationData.model_validate(args)
+        metadata_args = MetadataOperationData.model_validate(console_ns.payload or {})
 
         MetadataService.update_documents_metadata(dataset, metadata_args)
 

api/controllers/console/datasets/rag_pipeline/datasource_auth.py

@@ -1,20 +1,63 @@
+from typing import Any
+
 from flask import make_response, redirect, request
-from flask_restx import Resource, reqparse
+from flask_restx import Resource
+from pydantic import BaseModel, Field
 from werkzeug.exceptions import Forbidden, NotFound
 
 from configs import dify_config
+from controllers.common.schema import register_schema_models
 from controllers.console import console_ns
 from controllers.console.wraps import account_initialization_required, edit_permission_required, setup_required
 from core.model_runtime.errors.validate import CredentialsValidateFailedError
 from core.model_runtime.utils.encoders import jsonable_encoder
 from core.plugin.impl.oauth import OAuthHandler
-from libs.helper import StrLen
 from libs.login import current_account_with_tenant, login_required
 from models.provider_ids import DatasourceProviderID
 from services.datasource_provider_service import DatasourceProviderService
 from services.plugin.oauth_service import OAuthProxyService
 
 
+class DatasourceCredentialPayload(BaseModel):
+    name: str | None = Field(default=None, max_length=100)
+    credentials: dict[str, Any]
+
+
+class DatasourceCredentialDeletePayload(BaseModel):
+    credential_id: str
+
+
+class DatasourceCredentialUpdatePayload(BaseModel):
+    credential_id: str
+    name: str | None = Field(default=None, max_length=100)
+    credentials: dict[str, Any] | None = None
+
+
+class DatasourceCustomClientPayload(BaseModel):
+    client_params: dict[str, Any] | None = None
+    enable_oauth_custom_client: bool | None = None
+
+
+class DatasourceDefaultPayload(BaseModel):
+    id: str
+
+
+class DatasourceUpdateNamePayload(BaseModel):
+    credential_id: str
+    name: str = Field(max_length=100)
+
+
+register_schema_models(
+    console_ns,
+    DatasourceCredentialPayload,
+    DatasourceCredentialDeletePayload,
+    DatasourceCredentialUpdatePayload,
+    DatasourceCustomClientPayload,
+    DatasourceDefaultPayload,
+    DatasourceUpdateNamePayload,
+)
+
+
 @console_ns.route("/oauth/plugin/<path:provider_id>/datasource/get-authorization-url")
 class DatasourcePluginOAuthAuthorizationUrl(Resource):
     @setup_required
@@ -121,16 +164,9 @@ class DatasourceOAuthCallback(Resource):
         return redirect(f"{dify_config.CONSOLE_WEB_URL}/oauth-callback")
 
 
-parser_datasource = (
-    reqparse.RequestParser()
-    .add_argument("name", type=StrLen(max_length=100), required=False, nullable=True, location="json", default=None)
-    .add_argument("credentials", type=dict, required=True, nullable=False, location="json")
-)
-
-
 @console_ns.route("/auth/plugin/datasource/<path:provider_id>")
 class DatasourceAuth(Resource):
-    @console_ns.expect(parser_datasource)
+    @console_ns.expect(console_ns.models[DatasourceCredentialPayload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
@@ -138,7 +174,7 @@ class DatasourceAuth(Resource):
     def post(self, provider_id: str):
         _, current_tenant_id = current_account_with_tenant()
 
-        args = parser_datasource.parse_args()
+        payload = DatasourceCredentialPayload.model_validate(console_ns.payload or {})
         datasource_provider_id = DatasourceProviderID(provider_id)
         datasource_provider_service = DatasourceProviderService()
 
@@ -146,8 +182,8 @@ class DatasourceAuth(Resource):
             datasource_provider_service.add_datasource_api_key_provider(
                 tenant_id=current_tenant_id,
                 provider_id=datasource_provider_id,
-                credentials=args["credentials"],
-                name=args["name"],
+                credentials=payload.credentials,
+                name=payload.name,
             )
         except CredentialsValidateFailedError as ex:
             raise ValueError(str(ex))
@@ -169,14 +205,9 @@ class DatasourceAuth(Resource):
         return {"result": datasources}, 200
 
 
-parser_datasource_delete = reqparse.RequestParser().add_argument(
-    "credential_id", type=str, required=True, nullable=False, location="json"
-)
-
-
 @console_ns.route("/auth/plugin/datasource/<path:provider_id>/delete")
 class DatasourceAuthDeleteApi(Resource):
-    @console_ns.expect(parser_datasource_delete)
+    @console_ns.expect(console_ns.models[DatasourceCredentialDeletePayload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
@@ -188,28 +219,20 @@ class DatasourceAuthDeleteApi(Resource):
         plugin_id = datasource_provider_id.plugin_id
         provider_name = datasource_provider_id.provider_name
 
-        args = parser_datasource_delete.parse_args()
+        payload = DatasourceCredentialDeletePayload.model_validate(console_ns.payload or {})
         datasource_provider_service = DatasourceProviderService()
         datasource_provider_service.remove_datasource_credentials(
             tenant_id=current_tenant_id,
-            auth_id=args["credential_id"],
+            auth_id=payload.credential_id,
             provider=provider_name,
             plugin_id=plugin_id,
         )
         return {"result": "success"}, 200
 
 
-parser_datasource_update = (
-    reqparse.RequestParser()
-    .add_argument("credentials", type=dict, required=False, nullable=True, location="json")
-    .add_argument("name", type=StrLen(max_length=100), required=False, nullable=True, location="json")
-    .add_argument("credential_id", type=str, required=True, nullable=False, location="json")
-)
-
-
 @console_ns.route("/auth/plugin/datasource/<path:provider_id>/update")
 class DatasourceAuthUpdateApi(Resource):
-    @console_ns.expect(parser_datasource_update)
+    @console_ns.expect(console_ns.models[DatasourceCredentialUpdatePayload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
@@ -218,16 +241,16 @@ class DatasourceAuthUpdateApi(Resource):
         _, current_tenant_id = current_account_with_tenant()
 
         datasource_provider_id = DatasourceProviderID(provider_id)
-        args = parser_datasource_update.parse_args()
+        payload = DatasourceCredentialUpdatePayload.model_validate(console_ns.payload or {})
 
         datasource_provider_service = DatasourceProviderService()
         datasource_provider_service.update_datasource_credentials(
             tenant_id=current_tenant_id,
-            auth_id=args["credential_id"],
+            auth_id=payload.credential_id,
             provider=datasource_provider_id.provider_name,
             plugin_id=datasource_provider_id.plugin_id,
-            credentials=args.get("credentials", {}),
-            name=args.get("name", None),
+            credentials=payload.credentials or {},
+            name=payload.name,
         )
         return {"result": "success"}, 201
 
@@ -258,16 +281,9 @@ class DatasourceHardCodeAuthListApi(Resource):
         return {"result": jsonable_encoder(datasources)}, 200
 
 
-parser_datasource_custom = (
-    reqparse.RequestParser()
-    .add_argument("client_params", type=dict, required=False, nullable=True, location="json")
-    .add_argument("enable_oauth_custom_client", type=bool, required=False, nullable=True, location="json")
-)
-
-
 @console_ns.route("/auth/plugin/datasource/<path:provider_id>/custom-client")
 class DatasourceAuthOauthCustomClient(Resource):
-    @console_ns.expect(parser_datasource_custom)
+    @console_ns.expect(console_ns.models[DatasourceCustomClientPayload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
@@ -275,14 +291,14 @@ class DatasourceAuthOauthCustomClient(Resource):
     def post(self, provider_id: str):
         _, current_tenant_id = current_account_with_tenant()
 
-        args = parser_datasource_custom.parse_args()
+        payload = DatasourceCustomClientPayload.model_validate(console_ns.payload or {})
         datasource_provider_id = DatasourceProviderID(provider_id)
         datasource_provider_service = DatasourceProviderService()
         datasource_provider_service.setup_oauth_custom_client_params(
             tenant_id=current_tenant_id,
             datasource_provider_id=datasource_provider_id,
-            client_params=args.get("client_params", {}),
-            enabled=args.get("enable_oauth_custom_client", False),
+            client_params=payload.client_params or {},
+            enabled=payload.enable_oauth_custom_client or False,
         )
         return {"result": "success"}, 200
 
@@ -301,12 +317,9 @@ class DatasourceAuthOauthCustomClient(Resource):
         return {"result": "success"}, 200
 
 
-parser_default = reqparse.RequestParser().add_argument("id", type=str, required=True, nullable=False, location="json")
-
-
 @console_ns.route("/auth/plugin/datasource/<path:provider_id>/default")
 class DatasourceAuthDefaultApi(Resource):
-    @console_ns.expect(parser_default)
+    @console_ns.expect(console_ns.models[DatasourceDefaultPayload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
@@ -314,27 +327,20 @@ class DatasourceAuthDefaultApi(Resource):
     def post(self, provider_id: str):
         _, current_tenant_id = current_account_with_tenant()
 
-        args = parser_default.parse_args()
+        payload = DatasourceDefaultPayload.model_validate(console_ns.payload or {})
         datasource_provider_id = DatasourceProviderID(provider_id)
         datasource_provider_service = DatasourceProviderService()
         datasource_provider_service.set_default_datasource_provider(
             tenant_id=current_tenant_id,
             datasource_provider_id=datasource_provider_id,
-            credential_id=args["id"],
+            credential_id=payload.id,
         )
         return {"result": "success"}, 200
 
 
-parser_update_name = (
-    reqparse.RequestParser()
-    .add_argument("name", type=StrLen(max_length=100), required=True, nullable=False, location="json")
-    .add_argument("credential_id", type=str, required=True, nullable=False, location="json")
-)
-
-
 @console_ns.route("/auth/plugin/datasource/<path:provider_id>/update-name")
 class DatasourceUpdateProviderNameApi(Resource):
-    @console_ns.expect(parser_update_name)
+    @console_ns.expect(console_ns.models[DatasourceUpdateNamePayload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
@@ -342,13 +348,13 @@ class DatasourceUpdateProviderNameApi(Resource):
     def post(self, provider_id: str):
         _, current_tenant_id = current_account_with_tenant()
 
-        args = parser_update_name.parse_args()
+        payload = DatasourceUpdateNamePayload.model_validate(console_ns.payload or {})
         datasource_provider_id = DatasourceProviderID(provider_id)
         datasource_provider_service = DatasourceProviderService()
         datasource_provider_service.update_datasource_provider_name(
             tenant_id=current_tenant_id,
             datasource_provider_id=datasource_provider_id,
-            name=args["name"],
-            credential_id=args["credential_id"],
+            name=payload.name,
+            credential_id=payload.credential_id,
         )
         return {"result": "success"}, 200

api/controllers/console/datasets/rag_pipeline/rag_pipeline.py

@@ -1,9 +1,11 @@
 import logging
 
 from flask import request
-from flask_restx import Resource, reqparse
+from flask_restx import Resource
+from pydantic import BaseModel, Field
 from sqlalchemy.orm import Session
 
+from controllers.common.schema import register_schema_models
 from controllers.console import console_ns
 from controllers.console.wraps import (
     account_initialization_required,
@@ -20,18 +22,6 @@ from services.rag_pipeline.rag_pipeline import RagPipelineService
 logger = logging.getLogger(__name__)
 
 
-def _validate_name(name: str) -> str:
-    if not name or len(name) < 1 or len(name) > 40:
-        raise ValueError("Name must be between 1 to 40 characters.")
-    return name
-
-
-def _validate_description_length(description: str) -> str:
-    if len(description) > 400:
-        raise ValueError("Description cannot exceed 400 characters.")
-    return description
-
-
 @console_ns.route("/rag/pipeline/templates")
 class PipelineTemplateListApi(Resource):
     @setup_required
@@ -59,6 +49,15 @@ class PipelineTemplateDetailApi(Resource):
         return pipeline_template, 200
 
 
+class Payload(BaseModel):
+    name: str = Field(..., min_length=1, max_length=40)
+    description: str = Field(default="", max_length=400)
+    icon_info: dict[str, object] | None = None
+
+
+register_schema_models(console_ns, Payload)
+
+
 @console_ns.route("/rag/pipeline/customized/templates/<string:template_id>")
 class CustomizedPipelineTemplateApi(Resource):
     @setup_required
@@ -66,31 +65,8 @@ class CustomizedPipelineTemplateApi(Resource):
     @account_initialization_required
     @enterprise_license_required
     def patch(self, template_id: str):
-        parser = (
-            reqparse.RequestParser()
-            .add_argument(
-                "name",
-                nullable=False,
-                required=True,
-                help="Name must be between 1 to 40 characters.",
-                type=_validate_name,
-            )
-            .add_argument(
-                "description",
-                type=_validate_description_length,
-                nullable=True,
-                required=False,
-                default="",
-            )
-            .add_argument(
-                "icon_info",
-                type=dict,
-                location="json",
-                nullable=True,
-            )
-        )
-        args = parser.parse_args()
-        pipeline_template_info = PipelineTemplateInfoEntity.model_validate(args)
+        payload = Payload.model_validate(console_ns.payload or {})
+        pipeline_template_info = PipelineTemplateInfoEntity.model_validate(payload.model_dump())
         RagPipelineService.update_customized_pipeline_template(template_id, pipeline_template_info)
         return 200
 
@@ -119,36 +95,14 @@ class CustomizedPipelineTemplateApi(Resource):
 
 @console_ns.route("/rag/pipelines/<string:pipeline_id>/customized/publish")
 class PublishCustomizedPipelineTemplateApi(Resource):
+    @console_ns.expect(console_ns.models[Payload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
     @enterprise_license_required
     @knowledge_pipeline_publish_enabled
     def post(self, pipeline_id: str):
-        parser = (
-            reqparse.RequestParser()
-            .add_argument(
-                "name",
-                nullable=False,
-                required=True,
-                help="Name must be between 1 to 40 characters.",
-                type=_validate_name,
-            )
-            .add_argument(
-                "description",
-                type=_validate_description_length,
-                nullable=True,
-                required=False,
-                default="",
-            )
-            .add_argument(
-                "icon_info",
-                type=dict,
-                location="json",
-                nullable=True,
-            )
-        )
-        args = parser.parse_args()
+        payload = Payload.model_validate(console_ns.payload or {})
         rag_pipeline_service = RagPipelineService()
-        rag_pipeline_service.publish_customized_pipeline_template(pipeline_id, args)
+        rag_pipeline_service.publish_customized_pipeline_template(pipeline_id, payload.model_dump())
         return {"result": "success"}

api/controllers/console/datasets/rag_pipeline/rag_pipeline_datasets.py

@@ -1,8 +1,10 @@
-from flask_restx import Resource, marshal, reqparse
+from flask_restx import Resource, marshal
+from pydantic import BaseModel
 from sqlalchemy.orm import Session
 from werkzeug.exceptions import Forbidden
 
 import services
+from controllers.common.schema import register_schema_model
 from controllers.console import console_ns
 from controllers.console.datasets.error import DatasetNameDuplicateError
 from controllers.console.wraps import (
@@ -19,22 +21,22 @@ from services.entities.knowledge_entities.rag_pipeline_entities import IconInfo,
 from services.rag_pipeline.rag_pipeline_dsl_service import RagPipelineDslService
 
 
+class RagPipelineDatasetImportPayload(BaseModel):
+    yaml_content: str
+
+
+register_schema_model(console_ns, RagPipelineDatasetImportPayload)
+
+
 @console_ns.route("/rag/pipeline/dataset")
 class CreateRagPipelineDatasetApi(Resource):
+    @console_ns.expect(console_ns.models[RagPipelineDatasetImportPayload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
     @cloud_edition_billing_rate_limit_check("knowledge")
     def post(self):
-        parser = reqparse.RequestParser().add_argument(
-            "yaml_content",
-            type=str,
-            nullable=False,
-            required=True,
-            help="yaml_content is required.",
-        )
-
-        args = parser.parse_args()
+        payload = RagPipelineDatasetImportPayload.model_validate(console_ns.payload or {})
         current_user, current_tenant_id = current_account_with_tenant()
         # The role of the current user in the ta table must be admin, owner, or editor, or dataset_operator
         if not current_user.is_dataset_editor:
@@ -49,7 +51,7 @@ class CreateRagPipelineDatasetApi(Resource):
             ),
             permission=DatasetPermissionEnum.ONLY_ME,
             partial_member_list=None,
-            yaml_content=args["yaml_content"],
+            yaml_content=payload.yaml_content,
         )
         try:
             with Session(db.engine) as session:

api/controllers/console/datasets/rag_pipeline/rag_pipeline_draft_variable.py

@@ -1,11 +1,13 @@
 import logging
-from typing import NoReturn
+from typing import Any, NoReturn
 
-from flask import Response
-from flask_restx import Resource, fields, inputs, marshal, marshal_with, reqparse
+from flask import Response, request
+from flask_restx import Resource, fields, marshal, marshal_with
+from pydantic import BaseModel, Field
 from sqlalchemy.orm import Session
 from werkzeug.exceptions import Forbidden
 
+from controllers.common.schema import register_schema_models
 from controllers.console import console_ns
 from controllers.console.app.error import (
     DraftWorkflowNotExist,
@@ -33,19 +35,21 @@ logger = logging.getLogger(__name__)
 
 
 def _create_pagination_parser():
-    parser = (
-        reqparse.RequestParser()
-        .add_argument(
-            "page",
-            type=inputs.int_range(1, 100_000),
-            required=False,
-            default=1,
-            location="args",
-            help="the page of data requested",
-        )
-        .add_argument("limit", type=inputs.int_range(1, 100), required=False, default=20, location="args")
-    )
-    return parser
+    class PaginationQuery(BaseModel):
+        page: int = Field(default=1, ge=1, le=100_000)
+        limit: int = Field(default=20, ge=1, le=100)
+
+    register_schema_models(console_ns, PaginationQuery)
+
+    return PaginationQuery
+
+
+class WorkflowDraftVariablePatchPayload(BaseModel):
+    name: str | None = None
+    value: Any | None = None
+
+
+register_schema_models(console_ns, WorkflowDraftVariablePatchPayload)
 
 
 def _get_items(var_list: WorkflowDraftVariableList) -> list[WorkflowDraftVariable]:
@@ -93,8 +97,8 @@ class RagPipelineVariableCollectionApi(Resource):
         """
         Get draft workflow
         """
-        parser = _create_pagination_parser()
-        args = parser.parse_args()
+        pagination = _create_pagination_parser()
+        query = pagination.model_validate(request.args.to_dict())
 
         # fetch draft workflow by app_model
         rag_pipeline_service = RagPipelineService()
@@ -109,8 +113,8 @@ class RagPipelineVariableCollectionApi(Resource):
             )
         workflow_vars = draft_var_srv.list_variables_without_values(
             app_id=pipeline.id,
-            page=args.page,
-            limit=args.limit,
+            page=query.page,
+            limit=query.limit,
         )
 
         return workflow_vars
@@ -186,6 +190,7 @@ class RagPipelineVariableApi(Resource):
 
     @_api_prerequisite
     @marshal_with(_WORKFLOW_DRAFT_VARIABLE_FIELDS)
+    @console_ns.expect(console_ns.models[WorkflowDraftVariablePatchPayload.__name__])
     def patch(self, pipeline: Pipeline, variable_id: str):
         # Request payload for file types:
         #
@@ -208,16 +213,11 @@ class RagPipelineVariableApi(Resource):
         #         "upload_file_id": "1602650a-4fe4-423c-85a2-af76c083e3c4"
         #     }
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument(self._PATCH_NAME_FIELD, type=str, required=False, nullable=True, location="json")
-            .add_argument(self._PATCH_VALUE_FIELD, type=lambda x: x, required=False, nullable=True, location="json")
-        )
-
         draft_var_srv = WorkflowDraftVariableService(
             session=db.session(),
         )
-        args = parser.parse_args(strict=True)
+        payload = WorkflowDraftVariablePatchPayload.model_validate(console_ns.payload or {})
+        args = payload.model_dump(exclude_none=True)
 
         variable = draft_var_srv.get_variable(variable_id=variable_id)
         if variable is None:

api/controllers/console/datasets/rag_pipeline/rag_pipeline_import.py

@@ -1,6 +1,9 @@
-from flask_restx import Resource, marshal_with, reqparse  # type: ignore
+from flask import request
+from flask_restx import Resource, marshal_with  # type: ignore
+from pydantic import BaseModel, Field
 from sqlalchemy.orm import Session
 
+from controllers.common.schema import register_schema_models
 from controllers.console import console_ns
 from controllers.console.datasets.wraps import get_rag_pipeline
 from controllers.console.wraps import (
@@ -16,6 +19,25 @@ from services.app_dsl_service import ImportStatus
 from services.rag_pipeline.rag_pipeline_dsl_service import RagPipelineDslService
 
 
+class RagPipelineImportPayload(BaseModel):
+    mode: str
+    yaml_content: str | None = None
+    yaml_url: str | None = None
+    name: str | None = None
+    description: str | None = None
+    icon_type: str | None = None
+    icon: str | None = None
+    icon_background: str | None = None
+    pipeline_id: str | None = None
+
+
+class IncludeSecretQuery(BaseModel):
+    include_secret: str = Field(default="false")
+
+
+register_schema_models(console_ns, RagPipelineImportPayload, IncludeSecretQuery)
+
+
 @console_ns.route("/rag/pipelines/imports")
 class RagPipelineImportApi(Resource):
     @setup_required
@@ -23,23 +45,11 @@ class RagPipelineImportApi(Resource):
     @account_initialization_required
     @edit_permission_required
     @marshal_with(pipeline_import_fields)
+    @console_ns.expect(console_ns.models[RagPipelineImportPayload.__name__])
     def post(self):
         # Check user role first
         current_user, _ = current_account_with_tenant()
-
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("mode", type=str, required=True, location="json")
-            .add_argument("yaml_content", type=str, location="json")
-            .add_argument("yaml_url", type=str, location="json")
-            .add_argument("name", type=str, location="json")
-            .add_argument("description", type=str, location="json")
-            .add_argument("icon_type", type=str, location="json")
-            .add_argument("icon", type=str, location="json")
-            .add_argument("icon_background", type=str, location="json")
-            .add_argument("pipeline_id", type=str, location="json")
-        )
-        args = parser.parse_args()
+        payload = RagPipelineImportPayload.model_validate(console_ns.payload or {})
 
         # Create service with session
         with Session(db.engine) as session:
@@ -48,11 +58,11 @@ class RagPipelineImportApi(Resource):
             account = current_user
             result = import_service.import_rag_pipeline(
                 account=account,
-                import_mode=args["mode"],
-                yaml_content=args.get("yaml_content"),
-                yaml_url=args.get("yaml_url"),
-                pipeline_id=args.get("pipeline_id"),
-                dataset_name=args.get("name"),
+                import_mode=payload.mode,
+                yaml_content=payload.yaml_content,
+                yaml_url=payload.yaml_url,
+                pipeline_id=payload.pipeline_id,
+                dataset_name=payload.name,
             )
             session.commit()
 
@@ -114,13 +124,12 @@ class RagPipelineExportApi(Resource):
     @edit_permission_required
     def get(self, pipeline: Pipeline):
         # Add include_secret params
-        parser = reqparse.RequestParser().add_argument("include_secret", type=str, default="false", location="args")
-        args = parser.parse_args()
+        query = IncludeSecretQuery.model_validate(request.args.to_dict())
 
         with Session(db.engine) as session:
             export_service = RagPipelineDslService(session)
             result = export_service.export_rag_pipeline_dsl(
-                pipeline=pipeline, include_secret=args["include_secret"] == "true"
+                pipeline=pipeline, include_secret=query.include_secret == "true"
             )
 
         return {"data": result}, 200

api/controllers/console/datasets/rag_pipeline/rag_pipeline_workflow.py

@@ -1,14 +1,16 @@
 import json
 import logging
-from typing import cast
+from typing import Any, Literal, cast
+from uuid import UUID
 
 from flask import abort, request
-from flask_restx import Resource, inputs, marshal_with, reqparse  # type: ignore  # type: ignore
-from flask_restx.inputs import int_range  # type: ignore
+from flask_restx import Resource, marshal_with  # type: ignore
+from pydantic import BaseModel, Field
 from sqlalchemy.orm import Session
 from werkzeug.exceptions import Forbidden, InternalServerError, NotFound
 
 import services
+from controllers.common.schema import register_schema_models
 from controllers.console import console_ns
 from controllers.console.app.error import (
     ConversationCompletedError,
@@ -36,7 +38,7 @@ from fields.workflow_run_fields import (
     workflow_run_pagination_fields,
 )
 from libs import helper
-from libs.helper import TimestampField, uuid_value
+from libs.helper import TimestampField
 from libs.login import current_account_with_tenant, current_user, login_required
 from models import Account
 from models.dataset import Pipeline
@@ -51,6 +53,91 @@ from services.rag_pipeline.rag_pipeline_transform_service import RagPipelineTran
 logger = logging.getLogger(__name__)
 
 
+class DraftWorkflowSyncPayload(BaseModel):
+    graph: dict[str, Any]
+    hash: str | None = None
+    environment_variables: list[dict[str, Any]] | None = None
+    conversation_variables: list[dict[str, Any]] | None = None
+    rag_pipeline_variables: list[dict[str, Any]] | None = None
+    features: dict[str, Any] | None = None
+
+
+class NodeRunPayload(BaseModel):
+    inputs: dict[str, Any] | None = None
+
+
+class NodeRunRequiredPayload(BaseModel):
+    inputs: dict[str, Any]
+
+
+class DatasourceNodeRunPayload(BaseModel):
+    inputs: dict[str, Any]
+    datasource_type: str
+    credential_id: str | None = None
+
+
+class DraftWorkflowRunPayload(BaseModel):
+    inputs: dict[str, Any]
+    datasource_type: str
+    datasource_info_list: list[dict[str, Any]]
+    start_node_id: str
+
+
+class PublishedWorkflowRunPayload(DraftWorkflowRunPayload):
+    is_preview: bool = False
+    response_mode: Literal["streaming", "blocking"] = "streaming"
+    original_document_id: str | None = None
+
+
+class DefaultBlockConfigQuery(BaseModel):
+    q: str | None = None
+
+
+class WorkflowListQuery(BaseModel):
+    page: int = Field(default=1, ge=1, le=99999)
+    limit: int = Field(default=10, ge=1, le=100)
+    user_id: str | None = None
+    named_only: bool = False
+
+
+class WorkflowUpdatePayload(BaseModel):
+    marked_name: str | None = Field(default=None, max_length=20)
+    marked_comment: str | None = Field(default=None, max_length=100)
+
+
+class NodeIdQuery(BaseModel):
+    node_id: str
+
+
+class WorkflowRunQuery(BaseModel):
+    last_id: UUID | None = None
+    limit: int = Field(default=20, ge=1, le=100)
+
+
+class DatasourceVariablesPayload(BaseModel):
+    datasource_type: str
+    datasource_info: dict[str, Any]
+    start_node_id: str
+    start_node_title: str
+
+
+register_schema_models(
+    console_ns,
+    DraftWorkflowSyncPayload,
+    NodeRunPayload,
+    NodeRunRequiredPayload,
+    DatasourceNodeRunPayload,
+    DraftWorkflowRunPayload,
+    PublishedWorkflowRunPayload,
+    DefaultBlockConfigQuery,
+    WorkflowListQuery,
+    WorkflowUpdatePayload,
+    NodeIdQuery,
+    WorkflowRunQuery,
+    DatasourceVariablesPayload,
+)
+
+
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/draft")
 class DraftRagPipelineApi(Resource):
     @setup_required
@@ -88,15 +175,7 @@ class DraftRagPipelineApi(Resource):
         content_type = request.headers.get("Content-Type", "")
 
         if "application/json" in content_type:
-            parser = (
-                reqparse.RequestParser()
-                .add_argument("graph", type=dict, required=True, nullable=False, location="json")
-                .add_argument("hash", type=str, required=False, location="json")
-                .add_argument("environment_variables", type=list, required=False, location="json")
-                .add_argument("conversation_variables", type=list, required=False, location="json")
-                .add_argument("rag_pipeline_variables", type=list, required=False, location="json")
-            )
-            args = parser.parse_args()
+            payload_dict = console_ns.payload or {}
         elif "text/plain" in content_type:
             try:
                 data = json.loads(request.data.decode("utf-8"))
@@ -106,7 +185,7 @@ class DraftRagPipelineApi(Resource):
                 if not isinstance(data.get("graph"), dict):
                     raise ValueError("graph is not a dict")
 
-                args = {
+                payload_dict = {
                     "graph": data.get("graph"),
                     "features": data.get("features"),
                     "hash": data.get("hash"),
@@ -119,24 +198,26 @@ class DraftRagPipelineApi(Resource):
         else:
             abort(415)
 
+        payload = DraftWorkflowSyncPayload.model_validate(payload_dict)
+
         try:
-            environment_variables_list = args.get("environment_variables") or []
+            environment_variables_list = payload.environment_variables or []
             environment_variables = [
                 variable_factory.build_environment_variable_from_mapping(obj) for obj in environment_variables_list
             ]
-            conversation_variables_list = args.get("conversation_variables") or []
+            conversation_variables_list = payload.conversation_variables or []
             conversation_variables = [
                 variable_factory.build_conversation_variable_from_mapping(obj) for obj in conversation_variables_list
             ]
             rag_pipeline_service = RagPipelineService()
             workflow = rag_pipeline_service.sync_draft_workflow(
                 pipeline=pipeline,
-                graph=args["graph"],
-                unique_hash=args.get("hash"),
+                graph=payload.graph,
+                unique_hash=payload.hash,
                 account=current_user,
                 environment_variables=environment_variables,
                 conversation_variables=conversation_variables,
-                rag_pipeline_variables=args.get("rag_pipeline_variables") or [],
+                rag_pipeline_variables=payload.rag_pipeline_variables or [],
             )
         except WorkflowHashNotEqualError:
             raise DraftWorkflowNotSync()
@@ -148,12 +229,9 @@ class DraftRagPipelineApi(Resource):
         }
 
 
-parser_run = reqparse.RequestParser().add_argument("inputs", type=dict, location="json")
-
-
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/draft/iteration/nodes/<string:node_id>/run")
 class RagPipelineDraftRunIterationNodeApi(Resource):
-    @console_ns.expect(parser_run)
+    @console_ns.expect(console_ns.models[NodeRunPayload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
@@ -166,7 +244,8 @@ class RagPipelineDraftRunIterationNodeApi(Resource):
         # The role of the current user in the ta table must be admin, owner, or editor
         current_user, _ = current_account_with_tenant()
 
-        args = parser_run.parse_args()
+        payload = NodeRunPayload.model_validate(console_ns.payload or {})
+        args = payload.model_dump(exclude_none=True)
 
         try:
             response = PipelineGenerateService.generate_single_iteration(
@@ -187,7 +266,7 @@ class RagPipelineDraftRunIterationNodeApi(Resource):
 
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/draft/loop/nodes/<string:node_id>/run")
 class RagPipelineDraftRunLoopNodeApi(Resource):
-    @console_ns.expect(parser_run)
+    @console_ns.expect(console_ns.models[NodeRunPayload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
@@ -200,7 +279,8 @@ class RagPipelineDraftRunLoopNodeApi(Resource):
         # The role of the current user in the ta table must be admin, owner, or editor
         current_user, _ = current_account_with_tenant()
 
-        args = parser_run.parse_args()
+        payload = NodeRunPayload.model_validate(console_ns.payload or {})
+        args = payload.model_dump(exclude_none=True)
 
         try:
             response = PipelineGenerateService.generate_single_loop(
@@ -219,18 +299,9 @@ class RagPipelineDraftRunLoopNodeApi(Resource):
             raise InternalServerError()
 
 
-parser_draft_run = (
-    reqparse.RequestParser()
-    .add_argument("inputs", type=dict, required=True, nullable=False, location="json")
-    .add_argument("datasource_type", type=str, required=True, location="json")
-    .add_argument("datasource_info_list", type=list, required=True, location="json")
-    .add_argument("start_node_id", type=str, required=True, location="json")
-)
-
-
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/draft/run")
 class DraftRagPipelineRunApi(Resource):
-    @console_ns.expect(parser_draft_run)
+    @console_ns.expect(console_ns.models[DraftWorkflowRunPayload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
@@ -243,7 +314,8 @@ class DraftRagPipelineRunApi(Resource):
         # The role of the current user in the ta table must be admin, owner, or editor
         current_user, _ = current_account_with_tenant()
 
-        args = parser_draft_run.parse_args()
+        payload = DraftWorkflowRunPayload.model_validate(console_ns.payload or {})
+        args = payload.model_dump()
 
         try:
             response = PipelineGenerateService.generate(
@@ -259,21 +331,9 @@ class DraftRagPipelineRunApi(Resource):
             raise InvokeRateLimitHttpError(ex.description)
 
 
-parser_published_run = (
-    reqparse.RequestParser()
-    .add_argument("inputs", type=dict, required=True, nullable=False, location="json")
-    .add_argument("datasource_type", type=str, required=True, location="json")
-    .add_argument("datasource_info_list", type=list, required=True, location="json")
-    .add_argument("start_node_id", type=str, required=True, location="json")
-    .add_argument("is_preview", type=bool, required=True, location="json", default=False)
-    .add_argument("response_mode", type=str, required=True, location="json", default="streaming")
-    .add_argument("original_document_id", type=str, required=False, location="json")
-)
-
-
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/published/run")
 class PublishedRagPipelineRunApi(Resource):
-    @console_ns.expect(parser_published_run)
+    @console_ns.expect(console_ns.models[PublishedWorkflowRunPayload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
@@ -286,16 +346,16 @@ class PublishedRagPipelineRunApi(Resource):
         # The role of the current user in the ta table must be admin, owner, or editor
         current_user, _ = current_account_with_tenant()
 
-        args = parser_published_run.parse_args()
-
-        streaming = args["response_mode"] == "streaming"
+        payload = PublishedWorkflowRunPayload.model_validate(console_ns.payload or {})
+        args = payload.model_dump(exclude_none=True)
+        streaming = payload.response_mode == "streaming"
 
         try:
             response = PipelineGenerateService.generate(
                 pipeline=pipeline,
                 user=current_user,
                 args=args,
-                invoke_from=InvokeFrom.DEBUGGER if args.get("is_preview") else InvokeFrom.PUBLISHED,
+                invoke_from=InvokeFrom.DEBUGGER if payload.is_preview else InvokeFrom.PUBLISHED,
                 streaming=streaming,
             )
 
@@ -387,17 +447,9 @@ class PublishedRagPipelineRunApi(Resource):
 #
 #         return result
 #
-parser_rag_run = (
-    reqparse.RequestParser()
-    .add_argument("inputs", type=dict, required=True, nullable=False, location="json")
-    .add_argument("datasource_type", type=str, required=True, location="json")
-    .add_argument("credential_id", type=str, required=False, location="json")
-)
-
-
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/published/datasource/nodes/<string:node_id>/run")
 class RagPipelinePublishedDatasourceNodeRunApi(Resource):
-    @console_ns.expect(parser_rag_run)
+    @console_ns.expect(console_ns.models[DatasourceNodeRunPayload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
@@ -410,14 +462,7 @@ class RagPipelinePublishedDatasourceNodeRunApi(Resource):
         # The role of the current user in the ta table must be admin, owner, or editor
         current_user, _ = current_account_with_tenant()
 
-        args = parser_rag_run.parse_args()
-
-        inputs = args.get("inputs")
-        if inputs is None:
-            raise ValueError("missing inputs")
-        datasource_type = args.get("datasource_type")
-        if datasource_type is None:
-            raise ValueError("missing datasource_type")
+        payload = DatasourceNodeRunPayload.model_validate(console_ns.payload or {})
 
         rag_pipeline_service = RagPipelineService()
         return helper.compact_generate_response(
@@ -425,11 +470,11 @@ class RagPipelinePublishedDatasourceNodeRunApi(Resource):
                 rag_pipeline_service.run_datasource_workflow_node(
                     pipeline=pipeline,
                     node_id=node_id,
-                    user_inputs=inputs,
+                    user_inputs=payload.inputs,
                     account=current_user,
-                    datasource_type=datasource_type,
+                    datasource_type=payload.datasource_type,
                     is_published=False,
-                    credential_id=args.get("credential_id"),
+                    credential_id=payload.credential_id,
                 )
             )
         )
@@ -437,7 +482,7 @@ class RagPipelinePublishedDatasourceNodeRunApi(Resource):
 
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/draft/datasource/nodes/<string:node_id>/run")
 class RagPipelineDraftDatasourceNodeRunApi(Resource):
-    @console_ns.expect(parser_rag_run)
+    @console_ns.expect(console_ns.models[DatasourceNodeRunPayload.__name__])
     @setup_required
     @login_required
     @edit_permission_required
@@ -450,14 +495,7 @@ class RagPipelineDraftDatasourceNodeRunApi(Resource):
         # The role of the current user in the ta table must be admin, owner, or editor
         current_user, _ = current_account_with_tenant()
 
-        args = parser_rag_run.parse_args()
-
-        inputs = args.get("inputs")
-        if inputs is None:
-            raise ValueError("missing inputs")
-        datasource_type = args.get("datasource_type")
-        if datasource_type is None:
-            raise ValueError("missing datasource_type")
+        payload = DatasourceNodeRunPayload.model_validate(console_ns.payload or {})
 
         rag_pipeline_service = RagPipelineService()
         return helper.compact_generate_response(
@@ -465,24 +503,19 @@ class RagPipelineDraftDatasourceNodeRunApi(Resource):
                 rag_pipeline_service.run_datasource_workflow_node(
                     pipeline=pipeline,
                     node_id=node_id,
-                    user_inputs=inputs,
+                    user_inputs=payload.inputs,
                     account=current_user,
-                    datasource_type=datasource_type,
+                    datasource_type=payload.datasource_type,
                     is_published=False,
-                    credential_id=args.get("credential_id"),
+                    credential_id=payload.credential_id,
                 )
             )
         )
 
 
-parser_run_api = reqparse.RequestParser().add_argument(
-    "inputs", type=dict, required=True, nullable=False, location="json"
-)
-
-
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/draft/nodes/<string:node_id>/run")
 class RagPipelineDraftNodeRunApi(Resource):
-    @console_ns.expect(parser_run_api)
+    @console_ns.expect(console_ns.models[NodeRunRequiredPayload.__name__])
     @setup_required
     @login_required
     @edit_permission_required
@@ -496,11 +529,8 @@ class RagPipelineDraftNodeRunApi(Resource):
         # The role of the current user in the ta table must be admin, owner, or editor
         current_user, _ = current_account_with_tenant()
 
-        args = parser_run_api.parse_args()
-
-        inputs = args.get("inputs")
-        if inputs == None:
-            raise ValueError("missing inputs")
+        payload = NodeRunRequiredPayload.model_validate(console_ns.payload or {})
+        inputs = payload.inputs
 
         rag_pipeline_service = RagPipelineService()
         workflow_node_execution = rag_pipeline_service.run_draft_workflow_node(
@@ -602,12 +632,8 @@ class DefaultRagPipelineBlockConfigsApi(Resource):
         return rag_pipeline_service.get_default_block_configs()
 
 
-parser_default = reqparse.RequestParser().add_argument("q", type=str, location="args")
-
-
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/default-workflow-block-configs/<string:block_type>")
 class DefaultRagPipelineBlockConfigApi(Resource):
-    @console_ns.expect(parser_default)
     @setup_required
     @login_required
     @account_initialization_required
@@ -617,14 +643,12 @@ class DefaultRagPipelineBlockConfigApi(Resource):
         """
         Get default block config
         """
-        args = parser_default.parse_args()
-
-        q = args.get("q")
+        query = DefaultBlockConfigQuery.model_validate(request.args.to_dict())
 
         filters = None
-        if q:
+        if query.q:
             try:
-                filters = json.loads(args.get("q", ""))
+                filters = json.loads(query.q)
             except json.JSONDecodeError:
                 raise ValueError("Invalid filters")
 
@@ -633,18 +657,8 @@ class DefaultRagPipelineBlockConfigApi(Resource):
         return rag_pipeline_service.get_default_block_config(node_type=block_type, filters=filters)
 
 
-parser_wf = (
-    reqparse.RequestParser()
-    .add_argument("page", type=inputs.int_range(1, 99999), required=False, default=1, location="args")
-    .add_argument("limit", type=inputs.int_range(1, 100), required=False, default=10, location="args")
-    .add_argument("user_id", type=str, required=False, location="args")
-    .add_argument("named_only", type=inputs.boolean, required=False, default=False, location="args")
-)
-
-
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows")
 class PublishedAllRagPipelineApi(Resource):
-    @console_ns.expect(parser_wf)
     @setup_required
     @login_required
     @account_initialization_required
@@ -657,16 +671,16 @@ class PublishedAllRagPipelineApi(Resource):
         """
         current_user, _ = current_account_with_tenant()
 
-        args = parser_wf.parse_args()
-        page = args["page"]
-        limit = args["limit"]
-        user_id = args.get("user_id")
-        named_only = args.get("named_only", False)
+        query = WorkflowListQuery.model_validate(request.args.to_dict())
+
+        page = query.page
+        limit = query.limit
+        user_id = query.user_id
+        named_only = query.named_only
 
         if user_id:
             if user_id != current_user.id:
                 raise Forbidden()
-            user_id = cast(str, user_id)
 
         rag_pipeline_service = RagPipelineService()
         with Session(db.engine) as session:
@@ -687,16 +701,8 @@ class PublishedAllRagPipelineApi(Resource):
             }
 
 
-parser_wf_id = (
-    reqparse.RequestParser()
-    .add_argument("marked_name", type=str, required=False, location="json")
-    .add_argument("marked_comment", type=str, required=False, location="json")
-)
-
-
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/<string:workflow_id>")
 class RagPipelineByIdApi(Resource):
-    @console_ns.expect(parser_wf_id)
     @setup_required
     @login_required
     @account_initialization_required
@@ -710,20 +716,8 @@ class RagPipelineByIdApi(Resource):
         # Check permission
         current_user, _ = current_account_with_tenant()
 
-        args = parser_wf_id.parse_args()
-
-        # Validate name and comment length
-        if args.marked_name and len(args.marked_name) > 20:
-            raise ValueError("Marked name cannot exceed 20 characters")
-        if args.marked_comment and len(args.marked_comment) > 100:
-            raise ValueError("Marked comment cannot exceed 100 characters")
-
-        # Prepare update data
-        update_data = {}
-        if args.get("marked_name") is not None:
-            update_data["marked_name"] = args["marked_name"]
-        if args.get("marked_comment") is not None:
-            update_data["marked_comment"] = args["marked_comment"]
+        payload = WorkflowUpdatePayload.model_validate(console_ns.payload or {})
+        update_data = payload.model_dump(exclude_unset=True)
 
         if not update_data:
             return {"message": "No valid fields to update"}, 400
@@ -749,12 +743,8 @@ class RagPipelineByIdApi(Resource):
         return workflow
 
 
-parser_parameters = reqparse.RequestParser().add_argument("node_id", type=str, required=True, location="args")
-
-
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/published/processing/parameters")
 class PublishedRagPipelineSecondStepApi(Resource):
-    @console_ns.expect(parser_parameters)
     @setup_required
     @login_required
     @account_initialization_required
@@ -764,10 +754,8 @@ class PublishedRagPipelineSecondStepApi(Resource):
         """
         Get second step parameters of rag pipeline
         """
-        args = parser_parameters.parse_args()
-        node_id = args.get("node_id")
-        if not node_id:
-            raise ValueError("Node ID is required")
+        query = NodeIdQuery.model_validate(request.args.to_dict())
+        node_id = query.node_id
         rag_pipeline_service = RagPipelineService()
         variables = rag_pipeline_service.get_second_step_parameters(pipeline=pipeline, node_id=node_id, is_draft=False)
         return {
@@ -777,7 +765,6 @@ class PublishedRagPipelineSecondStepApi(Resource):
 
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/published/pre-processing/parameters")
 class PublishedRagPipelineFirstStepApi(Resource):
-    @console_ns.expect(parser_parameters)
     @setup_required
     @login_required
     @account_initialization_required
@@ -787,10 +774,8 @@ class PublishedRagPipelineFirstStepApi(Resource):
         """
         Get first step parameters of rag pipeline
         """
-        args = parser_parameters.parse_args()
-        node_id = args.get("node_id")
-        if not node_id:
-            raise ValueError("Node ID is required")
+        query = NodeIdQuery.model_validate(request.args.to_dict())
+        node_id = query.node_id
         rag_pipeline_service = RagPipelineService()
         variables = rag_pipeline_service.get_first_step_parameters(pipeline=pipeline, node_id=node_id, is_draft=False)
         return {
@@ -800,7 +785,6 @@ class PublishedRagPipelineFirstStepApi(Resource):
 
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/draft/pre-processing/parameters")
 class DraftRagPipelineFirstStepApi(Resource):
-    @console_ns.expect(parser_parameters)
     @setup_required
     @login_required
     @account_initialization_required
@@ -810,10 +794,8 @@ class DraftRagPipelineFirstStepApi(Resource):
         """
         Get first step parameters of rag pipeline
         """
-        args = parser_parameters.parse_args()
-        node_id = args.get("node_id")
-        if not node_id:
-            raise ValueError("Node ID is required")
+        query = NodeIdQuery.model_validate(request.args.to_dict())
+        node_id = query.node_id
         rag_pipeline_service = RagPipelineService()
         variables = rag_pipeline_service.get_first_step_parameters(pipeline=pipeline, node_id=node_id, is_draft=True)
         return {
@@ -823,7 +805,6 @@ class DraftRagPipelineFirstStepApi(Resource):
 
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/draft/processing/parameters")
 class DraftRagPipelineSecondStepApi(Resource):
-    @console_ns.expect(parser_parameters)
     @setup_required
     @login_required
     @account_initialization_required
@@ -833,10 +814,8 @@ class DraftRagPipelineSecondStepApi(Resource):
         """
         Get second step parameters of rag pipeline
         """
-        args = parser_parameters.parse_args()
-        node_id = args.get("node_id")
-        if not node_id:
-            raise ValueError("Node ID is required")
+        query = NodeIdQuery.model_validate(request.args.to_dict())
+        node_id = query.node_id
 
         rag_pipeline_service = RagPipelineService()
         variables = rag_pipeline_service.get_second_step_parameters(pipeline=pipeline, node_id=node_id, is_draft=True)
@@ -845,16 +824,8 @@ class DraftRagPipelineSecondStepApi(Resource):
         }
 
 
-parser_wf_run = (
-    reqparse.RequestParser()
-    .add_argument("last_id", type=uuid_value, location="args")
-    .add_argument("limit", type=int_range(1, 100), required=False, default=20, location="args")
-)
-
-
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflow-runs")
 class RagPipelineWorkflowRunListApi(Resource):
-    @console_ns.expect(parser_wf_run)
     @setup_required
     @login_required
     @account_initialization_required
@@ -864,7 +835,16 @@ class RagPipelineWorkflowRunListApi(Resource):
         """
         Get workflow run list
         """
-        args = parser_wf_run.parse_args()
+        query = WorkflowRunQuery.model_validate(
+            {
+                "last_id": request.args.get("last_id"),
+                "limit": request.args.get("limit", type=int, default=20),
+            }
+        )
+        args = {
+            "last_id": str(query.last_id) if query.last_id else None,
+            "limit": query.limit,
+        }
 
         rag_pipeline_service = RagPipelineService()
         result = rag_pipeline_service.get_rag_pipeline_paginate_workflow_runs(pipeline=pipeline, args=args)
@@ -964,18 +944,9 @@ class RagPipelineTransformApi(Resource):
         return result
 
 
-parser_var = (
-    reqparse.RequestParser()
-    .add_argument("datasource_type", type=str, required=True, location="json")
-    .add_argument("datasource_info", type=dict, required=True, location="json")
-    .add_argument("start_node_id", type=str, required=True, location="json")
-    .add_argument("start_node_title", type=str, required=True, location="json")
-)
-
-
 @console_ns.route("/rag/pipelines/<uuid:pipeline_id>/workflows/draft/datasource/variables-inspect")
 class RagPipelineDatasourceVariableApi(Resource):
-    @console_ns.expect(parser_var)
+    @console_ns.expect(console_ns.models[DatasourceVariablesPayload.__name__])
     @setup_required
     @login_required
     @account_initialization_required
@@ -987,7 +958,7 @@ class RagPipelineDatasourceVariableApi(Resource):
         Set datasource variables
         """
         current_user, _ = current_account_with_tenant()
-        args = parser_var.parse_args()
+        args = DatasourceVariablesPayload.model_validate(console_ns.payload or {}).model_dump()
 
         rag_pipeline_service = RagPipelineService()
         workflow_node_execution = rag_pipeline_service.set_datasource_variables(

api/controllers/console/datasets/website.py

@@ -1,5 +1,10 @@
-from flask_restx import Resource, fields, reqparse
+from typing import Literal
 
+from flask import request
+from flask_restx import Resource
+from pydantic import BaseModel
+
+from controllers.common.schema import register_schema_models
 from controllers.console import console_ns
 from controllers.console.datasets.error import WebsiteCrawlError
 from controllers.console.wraps import account_initialization_required, setup_required
@@ -7,48 +12,35 @@ from libs.login import login_required
 from services.website_service import WebsiteCrawlApiRequest, WebsiteCrawlStatusApiRequest, WebsiteService
 
 
+class WebsiteCrawlPayload(BaseModel):
+    provider: Literal["firecrawl", "watercrawl", "jinareader"]
+    url: str
+    options: dict[str, object]
+
+
+class WebsiteCrawlStatusQuery(BaseModel):
+    provider: Literal["firecrawl", "watercrawl", "jinareader"]
+
+
+register_schema_models(console_ns, WebsiteCrawlPayload, WebsiteCrawlStatusQuery)
+
+
 @console_ns.route("/website/crawl")
 class WebsiteCrawlApi(Resource):
     @console_ns.doc("crawl_website")
     @console_ns.doc(description="Crawl website content")
-    @console_ns.expect(
-        console_ns.model(
-            "WebsiteCrawlRequest",
-            {
-                "provider": fields.String(
-                    required=True,
-                    description="Crawl provider (firecrawl/watercrawl/jinareader)",
-                    enum=["firecrawl", "watercrawl", "jinareader"],
-                ),
-                "url": fields.String(required=True, description="URL to crawl"),
-                "options": fields.Raw(required=True, description="Crawl options"),
-            },
-        )
-    )
+    @console_ns.expect(console_ns.models[WebsiteCrawlPayload.__name__])
     @console_ns.response(200, "Website crawl initiated successfully")
     @console_ns.response(400, "Invalid crawl parameters")
     @setup_required
     @login_required
     @account_initialization_required
     def post(self):
-        parser = (
-            reqparse.RequestParser()
-            .add_argument(
-                "provider",
-                type=str,
-                choices=["firecrawl", "watercrawl", "jinareader"],
-                required=True,
-                nullable=True,
-                location="json",
-            )
-            .add_argument("url", type=str, required=True, nullable=True, location="json")
-            .add_argument("options", type=dict, required=True, nullable=True, location="json")
-        )
-        args = parser.parse_args()
+        payload = WebsiteCrawlPayload.model_validate(console_ns.payload or {})
 
         # Create typed request and validate
         try:
-            api_request = WebsiteCrawlApiRequest.from_args(args)
+            api_request = WebsiteCrawlApiRequest.from_args(payload.model_dump())
         except ValueError as e:
             raise WebsiteCrawlError(str(e))
 
@@ -65,6 +57,7 @@ class WebsiteCrawlStatusApi(Resource):
     @console_ns.doc("get_crawl_status")
     @console_ns.doc(description="Get website crawl status")
     @console_ns.doc(params={"job_id": "Crawl job ID", "provider": "Crawl provider (firecrawl/watercrawl/jinareader)"})
+    @console_ns.expect(console_ns.models[WebsiteCrawlStatusQuery.__name__])
     @console_ns.response(200, "Crawl status retrieved successfully")
     @console_ns.response(404, "Crawl job not found")
     @console_ns.response(400, "Invalid provider")
@@ -72,14 +65,11 @@ class WebsiteCrawlStatusApi(Resource):
     @login_required
     @account_initialization_required
     def get(self, job_id: str):
-        parser = reqparse.RequestParser().add_argument(
-            "provider", type=str, choices=["firecrawl", "watercrawl", "jinareader"], required=True, location="args"
-        )
-        args = parser.parse_args()
+        args = WebsiteCrawlStatusQuery.model_validate(request.args.to_dict())
 
         # Create typed request and validate
         try:
-            api_request = WebsiteCrawlStatusApiRequest.from_args(args, job_id)
+            api_request = WebsiteCrawlStatusApiRequest.from_args(args.model_dump(), job_id)
         except ValueError as e:
             raise WebsiteCrawlError(str(e))
 

api/controllers/console/explore/audio.py

@@ -1,9 +1,11 @@
 import logging
 
 from flask import request
+from pydantic import BaseModel, Field
 from werkzeug.exceptions import InternalServerError
 
 import services
+from controllers.common.schema import register_schema_model
 from controllers.console.app.error import (
     AppUnavailableError,
     AudioTooLargeError,
@@ -31,6 +33,16 @@ from .. import console_ns
 logger = logging.getLogger(__name__)
 
 
+class TextToAudioPayload(BaseModel):
+    message_id: str | None = None
+    voice: str | None = None
+    text: str | None = None
+    streaming: bool | None = Field(default=None, description="Enable streaming response")
+
+
+register_schema_model(console_ns, TextToAudioPayload)
+
+
 @console_ns.route(
     "/installed-apps/<uuid:installed_app_id>/audio-to-text",
     endpoint="installed_app_audio",
@@ -76,23 +88,15 @@ class ChatAudioApi(InstalledAppResource):
     endpoint="installed_app_text",
 )
 class ChatTextApi(InstalledAppResource):
+    @console_ns.expect(console_ns.models[TextToAudioPayload.__name__])
     def post(self, installed_app):
-        from flask_restx import reqparse
-
         app_model = installed_app.app
         try:
-            parser = (
-                reqparse.RequestParser()
-                .add_argument("message_id", type=str, required=False, location="json")
-                .add_argument("voice", type=str, location="json")
-                .add_argument("text", type=str, location="json")
-                .add_argument("streaming", type=bool, location="json")
-            )
-            args = parser.parse_args()
+            payload = TextToAudioPayload.model_validate(console_ns.payload or {})
 
-            message_id = args.get("message_id", None)
-            text = args.get("text", None)
-            voice = args.get("voice", None)
+            message_id = payload.message_id
+            text = payload.text
+            voice = payload.voice
 
             response = AudioService.transcript_tts(app_model=app_model, text=text, voice=voice, message_id=message_id)
             return response

api/controllers/console/explore/completion.py

@@ -1,9 +1,12 @@
 import logging
+from typing import Any, Literal
+from uuid import UUID
 
-from flask_restx import reqparse
+from pydantic import BaseModel, Field
 from werkzeug.exceptions import InternalServerError, NotFound
 
 import services
+from controllers.common.schema import register_schema_models
 from controllers.console.app.error import (
     AppUnavailableError,
     CompletionRequestError,
@@ -25,7 +28,6 @@ from core.model_runtime.errors.invoke import InvokeError
 from extensions.ext_database import db
 from libs import helper
 from libs.datetime_utils import naive_utc_now
-from libs.helper import uuid_value
 from libs.login import current_user
 from models import Account
 from models.model import AppMode
@@ -38,28 +40,42 @@ from .. import console_ns
 logger = logging.getLogger(__name__)
 
 
+class CompletionMessagePayload(BaseModel):
+    inputs: dict[str, Any]
+    query: str = ""
+    files: list[dict[str, Any]] | None = None
+    response_mode: Literal["blocking", "streaming"] | None = None
+    retriever_from: str = Field(default="explore_app")
+
+
+class ChatMessagePayload(BaseModel):
+    inputs: dict[str, Any]
+    query: str
+    files: list[dict[str, Any]] | None = None
+    conversation_id: UUID | None = None
+    parent_message_id: UUID | None = None
+    retriever_from: str = Field(default="explore_app")
+
+
+register_schema_models(console_ns, CompletionMessagePayload, ChatMessagePayload)
+
+
 # define completion api for user
 @console_ns.route(
     "/installed-apps/<uuid:installed_app_id>/completion-messages",
     endpoint="installed_app_completion",
 )
 class CompletionApi(InstalledAppResource):
+    @console_ns.expect(console_ns.models[CompletionMessagePayload.__name__])
     def post(self, installed_app):
         app_model = installed_app.app
         if app_model.mode != AppMode.COMPLETION:
             raise NotCompletionAppError()
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("inputs", type=dict, required=True, location="json")
-            .add_argument("query", type=str, location="json", default="")
-            .add_argument("files", type=list, required=False, location="json")
-            .add_argument("response_mode", type=str, choices=["blocking", "streaming"], location="json")
-            .add_argument("retriever_from", type=str, required=False, default="explore_app", location="json")
-        )
-        args = parser.parse_args()
+        payload = CompletionMessagePayload.model_validate(console_ns.payload or {})
+        args = payload.model_dump(exclude_none=True)
 
-        streaming = args["response_mode"] == "streaming"
+        streaming = payload.response_mode == "streaming"
         args["auto_generate_name"] = False
 
         installed_app.last_used_at = naive_utc_now()
@@ -123,22 +139,15 @@ class CompletionStopApi(InstalledAppResource):
     endpoint="installed_app_chat_completion",
 )
 class ChatApi(InstalledAppResource):
+    @console_ns.expect(console_ns.models[ChatMessagePayload.__name__])
     def post(self, installed_app):
         app_model = installed_app.app
         app_mode = AppMode.value_of(app_model.mode)
         if app_mode not in {AppMode.CHAT, AppMode.AGENT_CHAT, AppMode.ADVANCED_CHAT}:
             raise NotChatAppError()
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("inputs", type=dict, required=True, location="json")
-            .add_argument("query", type=str, required=True, location="json")
-            .add_argument("files", type=list, required=False, location="json")
-            .add_argument("conversation_id", type=uuid_value, location="json")
-            .add_argument("parent_message_id", type=uuid_value, required=False, location="json")
-            .add_argument("retriever_from", type=str, required=False, default="explore_app", location="json")
-        )
-        args = parser.parse_args()
+        payload = ChatMessagePayload.model_validate(console_ns.payload or {})
+        args = payload.model_dump(exclude_none=True)
 
         args["auto_generate_name"] = False
 

api/controllers/console/explore/conversation.py

@@ -1,14 +1,18 @@
-from flask_restx import marshal_with, reqparse
-from flask_restx.inputs import int_range
+from typing import Any
+from uuid import UUID
+
+from flask import request
+from flask_restx import marshal_with
+from pydantic import BaseModel, Field
 from sqlalchemy.orm import Session
 from werkzeug.exceptions import NotFound
 
+from controllers.common.schema import register_schema_models
 from controllers.console.explore.error import NotChatAppError
 from controllers.console.explore.wraps import InstalledAppResource
 from core.app.entities.app_invoke_entities import InvokeFrom
 from extensions.ext_database import db
 from fields.conversation_fields import conversation_infinite_scroll_pagination_fields, simple_conversation_fields
-from libs.helper import uuid_value
 from libs.login import current_user
 from models import Account
 from models.model import AppMode
@@ -19,29 +23,44 @@ from services.web_conversation_service import WebConversationService
 from .. import console_ns
 
 
+class ConversationListQuery(BaseModel):
+    last_id: UUID | None = None
+    limit: int = Field(default=20, ge=1, le=100)
+    pinned: bool | None = None
+
+
+class ConversationRenamePayload(BaseModel):
+    name: str
+    auto_generate: bool = False
+
+
+register_schema_models(console_ns, ConversationListQuery, ConversationRenamePayload)
+
+
 @console_ns.route(
     "/installed-apps/<uuid:installed_app_id>/conversations",
     endpoint="installed_app_conversations",
 )
 class ConversationListApi(InstalledAppResource):
     @marshal_with(conversation_infinite_scroll_pagination_fields)
+    @console_ns.expect(console_ns.models[ConversationListQuery.__name__])
     def get(self, installed_app):
         app_model = installed_app.app
         app_mode = AppMode.value_of(app_model.mode)
         if app_mode not in {AppMode.CHAT, AppMode.AGENT_CHAT, AppMode.ADVANCED_CHAT}:
             raise NotChatAppError()
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("last_id", type=uuid_value, location="args")
-            .add_argument("limit", type=int_range(1, 100), required=False, default=20, location="args")
-            .add_argument("pinned", type=str, choices=["true", "false", None], location="args")
-        )
-        args = parser.parse_args()
-
-        pinned = None
-        if "pinned" in args and args["pinned"] is not None:
-            pinned = args["pinned"] == "true"
+        raw_args: dict[str, Any] = {
+            "last_id": request.args.get("last_id"),
+            "limit": request.args.get("limit", default=20, type=int),
+            "pinned": request.args.get("pinned"),
+        }
+        if raw_args["last_id"] is None:
+            raw_args["last_id"] = None
+        pinned_value = raw_args["pinned"]
+        if isinstance(pinned_value, str):
+            raw_args["pinned"] = pinned_value == "true"
+        args = ConversationListQuery.model_validate(raw_args)
 
         try:
             if not isinstance(current_user, Account):
@@ -51,10 +70,10 @@ class ConversationListApi(InstalledAppResource):
                     session=session,
                     app_model=app_model,
                     user=current_user,
-                    last_id=args["last_id"],
-                    limit=args["limit"],
+                    last_id=str(args.last_id) if args.last_id else None,
+                    limit=args.limit,
                     invoke_from=InvokeFrom.EXPLORE,
-                    pinned=pinned,
+                    pinned=args.pinned,
                 )
         except LastConversationNotExistsError:
             raise NotFound("Last Conversation Not Exists.")
@@ -88,6 +107,7 @@ class ConversationApi(InstalledAppResource):
 )
 class ConversationRenameApi(InstalledAppResource):
     @marshal_with(simple_conversation_fields)
+    @console_ns.expect(console_ns.models[ConversationRenamePayload.__name__])
     def post(self, installed_app, c_id):
         app_model = installed_app.app
         app_mode = AppMode.value_of(app_model.mode)
@@ -96,18 +116,13 @@ class ConversationRenameApi(InstalledAppResource):
 
         conversation_id = str(c_id)
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("name", type=str, required=False, location="json")
-            .add_argument("auto_generate", type=bool, required=False, default=False, location="json")
-        )
-        args = parser.parse_args()
+        payload = ConversationRenamePayload.model_validate(console_ns.payload or {})
 
         try:
             if not isinstance(current_user, Account):
                 raise ValueError("current_user must be an Account instance")
             return ConversationService.rename(
-                app_model, conversation_id, current_user, args["name"], args["auto_generate"]
+                app_model, conversation_id, current_user, payload.name, payload.auto_generate
             )
         except ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")

api/controllers/console/explore/message.py

@@ -1,9 +1,13 @@
 import logging
+from typing import Literal
+from uuid import UUID
 
-from flask_restx import marshal_with, reqparse
-from flask_restx.inputs import int_range
+from flask import request
+from flask_restx import marshal_with
+from pydantic import BaseModel, Field
 from werkzeug.exceptions import InternalServerError, NotFound
 
+from controllers.common.schema import register_schema_models
 from controllers.console.app.error import (
     AppMoreLikeThisDisabledError,
     CompletionRequestError,
@@ -22,7 +26,6 @@ from core.errors.error import ModelCurrentlyNotSupportError, ProviderTokenNotIni
 from core.model_runtime.errors.invoke import InvokeError
 from fields.message_fields import message_infinite_scroll_pagination_fields
 from libs import helper
-from libs.helper import uuid_value
 from libs.login import current_account_with_tenant
 from models.model import AppMode
 from services.app_generate_service import AppGenerateService
@@ -40,12 +43,31 @@ from .. import console_ns
 logger = logging.getLogger(__name__)
 
 
+class MessageListQuery(BaseModel):
+    conversation_id: UUID
+    first_id: UUID | None = None
+    limit: int = Field(default=20, ge=1, le=100)
+
+
+class MessageFeedbackPayload(BaseModel):
+    rating: Literal["like", "dislike"] | None = None
+    content: str | None = None
+
+
+class MoreLikeThisQuery(BaseModel):
+    response_mode: Literal["blocking", "streaming"]
+
+
+register_schema_models(console_ns, MessageListQuery, MessageFeedbackPayload, MoreLikeThisQuery)
+
+
 @console_ns.route(
     "/installed-apps/<uuid:installed_app_id>/messages",
     endpoint="installed_app_messages",
 )
 class MessageListApi(InstalledAppResource):
     @marshal_with(message_infinite_scroll_pagination_fields)
+    @console_ns.expect(console_ns.models[MessageListQuery.__name__])
     def get(self, installed_app):
         current_user, _ = current_account_with_tenant()
         app_model = installed_app.app
@@ -53,18 +75,15 @@ class MessageListApi(InstalledAppResource):
         app_mode = AppMode.value_of(app_model.mode)
         if app_mode not in {AppMode.CHAT, AppMode.AGENT_CHAT, AppMode.ADVANCED_CHAT}:
             raise NotChatAppError()
-
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("conversation_id", required=True, type=uuid_value, location="args")
-            .add_argument("first_id", type=uuid_value, location="args")
-            .add_argument("limit", type=int_range(1, 100), required=False, default=20, location="args")
-        )
-        args = parser.parse_args()
+        args = MessageListQuery.model_validate(request.args.to_dict())
 
         try:
             return MessageService.pagination_by_first_id(
-                app_model, current_user, args["conversation_id"], args["first_id"], args["limit"]
+                app_model,
+                current_user,
+                str(args.conversation_id),
+                str(args.first_id) if args.first_id else None,
+                args.limit,
             )
         except ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")
@@ -77,26 +96,22 @@ class MessageListApi(InstalledAppResource):
     endpoint="installed_app_message_feedback",
 )
 class MessageFeedbackApi(InstalledAppResource):
+    @console_ns.expect(console_ns.models[MessageFeedbackPayload.__name__])
     def post(self, installed_app, message_id):
         current_user, _ = current_account_with_tenant()
         app_model = installed_app.app
 
         message_id = str(message_id)
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("rating", type=str, choices=["like", "dislike", None], location="json")
-            .add_argument("content", type=str, location="json")
-        )
-        args = parser.parse_args()
+        payload = MessageFeedbackPayload.model_validate(console_ns.payload or {})
 
         try:
             MessageService.create_feedback(
                 app_model=app_model,
                 message_id=message_id,
                 user=current_user,
-                rating=args.get("rating"),
-                content=args.get("content"),
+                rating=payload.rating,
+                content=payload.content,
             )
         except MessageNotExistsError:
             raise NotFound("Message Not Exists.")
@@ -109,6 +124,7 @@ class MessageFeedbackApi(InstalledAppResource):
     endpoint="installed_app_more_like_this",
 )
 class MessageMoreLikeThisApi(InstalledAppResource):
+    @console_ns.expect(console_ns.models[MoreLikeThisQuery.__name__])
     def get(self, installed_app, message_id):
         current_user, _ = current_account_with_tenant()
         app_model = installed_app.app
@@ -117,12 +133,9 @@ class MessageMoreLikeThisApi(InstalledAppResource):
 
         message_id = str(message_id)
 
-        parser = reqparse.RequestParser().add_argument(
-            "response_mode", type=str, required=True, choices=["blocking", "streaming"], location="args"
-        )
-        args = parser.parse_args()
+        args = MoreLikeThisQuery.model_validate(request.args.to_dict())
 
-        streaming = args["response_mode"] == "streaming"
+        streaming = args.response_mode == "streaming"
 
         try:
             response = AppGenerateService.generate_more_like_this(

api/controllers/console/explore/saved_message.py

@@ -1,16 +1,33 @@
-from flask_restx import fields, marshal_with, reqparse
-from flask_restx.inputs import int_range
+from uuid import UUID
+
+from flask import request
+from flask_restx import fields, marshal_with
+from pydantic import BaseModel, Field
 from werkzeug.exceptions import NotFound
 
+from controllers.common.schema import register_schema_models
 from controllers.console import console_ns
 from controllers.console.explore.error import NotCompletionAppError
 from controllers.console.explore.wraps import InstalledAppResource
 from fields.conversation_fields import message_file_fields
-from libs.helper import TimestampField, uuid_value
+from libs.helper import TimestampField
 from libs.login import current_account_with_tenant
 from services.errors.message import MessageNotExistsError
 from services.saved_message_service import SavedMessageService
 
+
+class SavedMessageListQuery(BaseModel):
+    last_id: UUID | None = None
+    limit: int = Field(default=20, ge=1, le=100)
+
+
+class SavedMessageCreatePayload(BaseModel):
+    message_id: UUID
+
+
+register_schema_models(console_ns, SavedMessageListQuery, SavedMessageCreatePayload)
+
+
 feedback_fields = {"rating": fields.String}
 
 message_fields = {
@@ -33,32 +50,33 @@ class SavedMessageListApi(InstalledAppResource):
     }
 
     @marshal_with(saved_message_infinite_scroll_pagination_fields)
+    @console_ns.expect(console_ns.models[SavedMessageListQuery.__name__])
     def get(self, installed_app):
         current_user, _ = current_account_with_tenant()
         app_model = installed_app.app
         if app_model.mode != "completion":
             raise NotCompletionAppError()
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("last_id", type=uuid_value, location="args")
-            .add_argument("limit", type=int_range(1, 100), required=False, default=20, location="args")
+        args = SavedMessageListQuery.model_validate(request.args.to_dict())
+
+        return SavedMessageService.pagination_by_last_id(
+            app_model,
+            current_user,
+            str(args.last_id) if args.last_id else None,
+            args.limit,
         )
-        args = parser.parse_args()
-
-        return SavedMessageService.pagination_by_last_id(app_model, current_user, args["last_id"], args["limit"])
 
+    @console_ns.expect(console_ns.models[SavedMessageCreatePayload.__name__])
     def post(self, installed_app):
         current_user, _ = current_account_with_tenant()
         app_model = installed_app.app
         if app_model.mode != "completion":
             raise NotCompletionAppError()
 
-        parser = reqparse.RequestParser().add_argument("message_id", type=uuid_value, required=True, location="json")
-        args = parser.parse_args()
+        payload = SavedMessageCreatePayload.model_validate(console_ns.payload or {})
 
         try:
-            SavedMessageService.save(app_model, current_user, args["message_id"])
+            SavedMessageService.save(app_model, current_user, str(payload.message_id))
         except MessageNotExistsError:
             raise NotFound("Message Not Exists.")
 

api/controllers/console/explore/workflow.py

@@ -1,8 +1,10 @@
 import logging
+from typing import Any
 
-from flask_restx import reqparse
+from pydantic import BaseModel
 from werkzeug.exceptions import InternalServerError
 
+from controllers.common.schema import register_schema_model
 from controllers.console.app.error import (
     CompletionRequestError,
     ProviderModelCurrentlyNotSupportError,
@@ -32,8 +34,17 @@ from .. import console_ns
 logger = logging.getLogger(__name__)
 
 
+class WorkflowRunPayload(BaseModel):
+    inputs: dict[str, Any]
+    files: list[dict[str, Any]] | None = None
+
+
+register_schema_model(console_ns, WorkflowRunPayload)
+
+
 @console_ns.route("/installed-apps/<uuid:installed_app_id>/workflows/run")
 class InstalledAppWorkflowRunApi(InstalledAppResource):
+    @console_ns.expect(console_ns.models[WorkflowRunPayload.__name__])
     def post(self, installed_app: InstalledApp):
         """
         Run workflow
@@ -46,12 +57,8 @@ class InstalledAppWorkflowRunApi(InstalledAppResource):
         if app_mode != AppMode.WORKFLOW:
             raise NotWorkflowAppError()
 
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("inputs", type=dict, required=True, nullable=False, location="json")
-            .add_argument("files", type=list, required=False, location="json")
-        )
-        args = parser.parse_args()
+        payload = WorkflowRunPayload.model_validate(console_ns.payload or {})
+        args = payload.model_dump(exclude_none=True)
         try:
             response = AppGenerateService.generate(
                 app_model=app_model, user=current_user, args=args, invoke_from=InvokeFrom.EXPLORE, streaming=True

api/controllers/console/files.py

@@ -45,6 +45,9 @@ class FileApi(Resource):
             "video_file_size_limit": dify_config.UPLOAD_VIDEO_FILE_SIZE_LIMIT,
             "audio_file_size_limit": dify_config.UPLOAD_AUDIO_FILE_SIZE_LIMIT,
             "workflow_file_upload_limit": dify_config.WORKFLOW_FILE_UPLOAD_LIMIT,
+            "image_file_batch_limit": dify_config.IMAGE_FILE_BATCH_LIMIT,
+            "single_chunk_attachment_limit": dify_config.SINGLE_CHUNK_ATTACHMENT_LIMIT,
+            "attachment_image_file_size_limit": dify_config.ATTACHMENT_IMAGE_FILE_SIZE_LIMIT,
         }, 200
 
     @setup_required

api/controllers/console/workspace/trigger_providers.py

@@ -22,7 +22,12 @@ from services.trigger.trigger_subscription_builder_service import TriggerSubscri
 from services.trigger.trigger_subscription_operator_service import TriggerSubscriptionOperatorService
 
 from .. import console_ns
-from ..wraps import account_initialization_required, is_admin_or_owner_required, setup_required
+from ..wraps import (
+    account_initialization_required,
+    edit_permission_required,
+    is_admin_or_owner_required,
+    setup_required,
+)
 
 logger = logging.getLogger(__name__)
 
@@ -72,7 +77,7 @@ class TriggerProviderInfoApi(Resource):
 class TriggerSubscriptionListApi(Resource):
     @setup_required
     @login_required
-    @is_admin_or_owner_required
+    @edit_permission_required
     @account_initialization_required
     def get(self, provider):
         """List all trigger subscriptions for the current tenant's provider"""
@@ -104,7 +109,7 @@ class TriggerSubscriptionBuilderCreateApi(Resource):
     @console_ns.expect(parser)
     @setup_required
     @login_required
-    @is_admin_or_owner_required
+    @edit_permission_required
     @account_initialization_required
     def post(self, provider):
         """Add a new subscription instance for a trigger provider"""
@@ -133,6 +138,7 @@ class TriggerSubscriptionBuilderCreateApi(Resource):
 class TriggerSubscriptionBuilderGetApi(Resource):
     @setup_required
     @login_required
+    @edit_permission_required
     @account_initialization_required
     def get(self, provider, subscription_builder_id):
         """Get a subscription instance for a trigger provider"""
@@ -155,7 +161,7 @@ class TriggerSubscriptionBuilderVerifyApi(Resource):
     @console_ns.expect(parser_api)
     @setup_required
     @login_required
-    @is_admin_or_owner_required
+    @edit_permission_required
     @account_initialization_required
     def post(self, provider, subscription_builder_id):
         """Verify a subscription instance for a trigger provider"""
@@ -200,6 +206,7 @@ class TriggerSubscriptionBuilderUpdateApi(Resource):
     @console_ns.expect(parser_update_api)
     @setup_required
     @login_required
+    @edit_permission_required
     @account_initialization_required
     def post(self, provider, subscription_builder_id):
         """Update a subscription instance for a trigger provider"""
@@ -233,6 +240,7 @@ class TriggerSubscriptionBuilderUpdateApi(Resource):
 class TriggerSubscriptionBuilderLogsApi(Resource):
     @setup_required
     @login_required
+    @edit_permission_required
     @account_initialization_required
     def get(self, provider, subscription_builder_id):
         """Get the request logs for a subscription instance for a trigger provider"""
@@ -255,7 +263,7 @@ class TriggerSubscriptionBuilderBuildApi(Resource):
     @console_ns.expect(parser_update_api)
     @setup_required
     @login_required
-    @is_admin_or_owner_required
+    @edit_permission_required
     @account_initialization_required
     def post(self, provider, subscription_builder_id):
         """Build a subscription instance for a trigger provider"""

api/controllers/inner_api/mail.py

@@ -1,29 +1,38 @@
-from flask_restx import Resource, reqparse
+from typing import Any
 
+from flask_restx import Resource
+from pydantic import BaseModel, Field
+
+from controllers.common.schema import register_schema_model
 from controllers.console.wraps import setup_required
 from controllers.inner_api import inner_api_ns
 from controllers.inner_api.wraps import billing_inner_api_only, enterprise_inner_api_only
 from tasks.mail_inner_task import send_inner_email_task
 
-_mail_parser = (
-    reqparse.RequestParser()
-    .add_argument("to", type=str, action="append", required=True)
-    .add_argument("subject", type=str, required=True)
-    .add_argument("body", type=str, required=True)
-    .add_argument("substitutions", type=dict, required=False)
-)
+
+class InnerMailPayload(BaseModel):
+    to: list[str] = Field(description="Recipient email addresses", min_length=1)
+    subject: str
+    body: str
+    substitutions: dict[str, Any] | None = None
+
+
+register_schema_model(inner_api_ns, InnerMailPayload)
 
 
 class BaseMail(Resource):
     """Shared logic for sending an inner email."""
 
+    @inner_api_ns.doc("send_inner_mail")
+    @inner_api_ns.doc(description="Send internal email")
+    @inner_api_ns.expect(inner_api_ns.models[InnerMailPayload.__name__])
     def post(self):
-        args = _mail_parser.parse_args()
-        send_inner_email_task.delay(  # type: ignore
-            to=args["to"],
-            subject=args["subject"],
-            body=args["body"],
-            substitutions=args["substitutions"],
+        args = InnerMailPayload.model_validate(inner_api_ns.payload or {})
+        send_inner_email_task.delay(
+            to=args.to,
+            subject=args.subject,
+            body=args.body,
+            substitutions=args.substitutions,  # type: ignore
         )
         return {"message": "success"}, 200
 
@@ -34,7 +43,7 @@ class EnterpriseMail(BaseMail):
 
     @inner_api_ns.doc("send_enterprise_mail")
     @inner_api_ns.doc(description="Send internal email for enterprise features")
-    @inner_api_ns.expect(_mail_parser)
+    @inner_api_ns.expect(inner_api_ns.models[InnerMailPayload.__name__])
     @inner_api_ns.doc(
         responses={200: "Email sent successfully", 401: "Unauthorized - invalid API key", 404: "Service not available"}
     )
@@ -56,7 +65,7 @@ class BillingMail(BaseMail):
 
     @inner_api_ns.doc("send_billing_mail")
     @inner_api_ns.doc(description="Send internal email for billing notifications")
-    @inner_api_ns.expect(_mail_parser)
+    @inner_api_ns.expect(inner_api_ns.models[InnerMailPayload.__name__])
     @inner_api_ns.doc(
         responses={200: "Email sent successfully", 401: "Unauthorized - invalid API key", 404: "Service not available"}
     )

api/controllers/inner_api/plugin/wraps.py

@@ -1,10 +1,9 @@
 from collections.abc import Callable
 from functools import wraps
-from typing import ParamSpec, TypeVar, cast
+from typing import ParamSpec, TypeVar
 
 from flask import current_app, request
 from flask_login import user_logged_in
-from flask_restx import reqparse
 from pydantic import BaseModel
 from sqlalchemy.orm import Session
 
@@ -17,6 +16,11 @@ P = ParamSpec("P")
 R = TypeVar("R")
 
 
+class TenantUserPayload(BaseModel):
+    tenant_id: str
+    user_id: str
+
+
 def get_user(tenant_id: str, user_id: str | None) -> EndUser:
     """
     Get current user
@@ -67,58 +71,45 @@ def get_user(tenant_id: str, user_id: str | None) -> EndUser:
     return user_model
 
 
-def get_user_tenant(view: Callable[P, R] | None = None):
-    def decorator(view_func: Callable[P, R]):
-        @wraps(view_func)
-        def decorated_view(*args: P.args, **kwargs: P.kwargs):
-            # fetch json body
-            parser = (
-                reqparse.RequestParser()
-                .add_argument("tenant_id", type=str, required=True, location="json")
-                .add_argument("user_id", type=str, required=True, location="json")
-            )
+def get_user_tenant(view_func: Callable[P, R]):
+    @wraps(view_func)
+    def decorated_view(*args: P.args, **kwargs: P.kwargs):
+        payload = TenantUserPayload.model_validate(request.get_json(silent=True) or {})
 
-            p = parser.parse_args()
+        user_id = payload.user_id
+        tenant_id = payload.tenant_id
 
-            user_id = cast(str, p.get("user_id"))
-            tenant_id = cast(str, p.get("tenant_id"))
+        if not tenant_id:
+            raise ValueError("tenant_id is required")
 
-            if not tenant_id:
-                raise ValueError("tenant_id is required")
+        if not user_id:
+            user_id = DefaultEndUserSessionID.DEFAULT_SESSION_ID
 
-            if not user_id:
-                user_id = DefaultEndUserSessionID.DEFAULT_SESSION_ID
-
-            try:
-                tenant_model = (
-                    db.session.query(Tenant)
-                    .where(
-                        Tenant.id == tenant_id,
-                    )
-                    .first()
+        try:
+            tenant_model = (
+                db.session.query(Tenant)
+                .where(
+                    Tenant.id == tenant_id,
                 )
-            except Exception:
-                raise ValueError("tenant not found")
+                .first()
+            )
+        except Exception:
+            raise ValueError("tenant not found")
 
-            if not tenant_model:
-                raise ValueError("tenant not found")
+        if not tenant_model:
+            raise ValueError("tenant not found")
 
-            kwargs["tenant_model"] = tenant_model
+        kwargs["tenant_model"] = tenant_model
 
-            user = get_user(tenant_id, user_id)
-            kwargs["user_model"] = user
+        user = get_user(tenant_id, user_id)
+        kwargs["user_model"] = user
 
-            current_app.login_manager._update_request_context_with_user(user)  # type: ignore
-            user_logged_in.send(current_app._get_current_object(), user=current_user)  # type: ignore
+        current_app.login_manager._update_request_context_with_user(user)  # type: ignore
+        user_logged_in.send(current_app._get_current_object(), user=current_user)  # type: ignore
 
-            return view_func(*args, **kwargs)
+        return view_func(*args, **kwargs)
 
-        return decorated_view
-
-    if view is None:
-        return decorator
-    else:
-        return decorator(view)
+    return decorated_view
 
 
 def plugin_data(view: Callable[P, R] | None = None, *, payload_type: type[BaseModel]):

api/controllers/inner_api/workspace/workspace.py

@@ -1,7 +1,9 @@
 import json
 
-from flask_restx import Resource, reqparse
+from flask_restx import Resource
+from pydantic import BaseModel
 
+from controllers.common.schema import register_schema_models
 from controllers.console.wraps import setup_required
 from controllers.inner_api import inner_api_ns
 from controllers.inner_api.wraps import enterprise_inner_api_only
@@ -11,12 +13,25 @@ from models import Account
 from services.account_service import TenantService
 
 
+class WorkspaceCreatePayload(BaseModel):
+    name: str
+    owner_email: str
+
+
+class WorkspaceOwnerlessPayload(BaseModel):
+    name: str
+
+
+register_schema_models(inner_api_ns, WorkspaceCreatePayload, WorkspaceOwnerlessPayload)
+
+
 @inner_api_ns.route("/enterprise/workspace")
 class EnterpriseWorkspace(Resource):
     @setup_required
     @enterprise_inner_api_only
     @inner_api_ns.doc("create_enterprise_workspace")
     @inner_api_ns.doc(description="Create a new enterprise workspace with owner assignment")
+    @inner_api_ns.expect(inner_api_ns.models[WorkspaceCreatePayload.__name__])
     @inner_api_ns.doc(
         responses={
             200: "Workspace created successfully",
@@ -25,18 +40,13 @@ class EnterpriseWorkspace(Resource):
         }
     )
     def post(self):
-        parser = (
-            reqparse.RequestParser()
-            .add_argument("name", type=str, required=True, location="json")
-            .add_argument("owner_email", type=str, required=True, location="json")
-        )
-        args = parser.parse_args()
+        args = WorkspaceCreatePayload.model_validate(inner_api_ns.payload or {})
 
-        account = db.session.query(Account).filter_by(email=args["owner_email"]).first()
+        account = db.session.query(Account).filter_by(email=args.owner_email).first()
         if account is None:
             return {"message": "owner account not found."}, 404
 
-        tenant = TenantService.create_tenant(args["name"], is_from_dashboard=True)
+        tenant = TenantService.create_tenant(args.name, is_from_dashboard=True)
         TenantService.create_tenant_member(tenant, account, role="owner")
 
         tenant_was_created.send(tenant)
@@ -62,6 +72,7 @@ class EnterpriseWorkspaceNoOwnerEmail(Resource):
     @enterprise_inner_api_only
     @inner_api_ns.doc("create_enterprise_workspace_ownerless")
     @inner_api_ns.doc(description="Create a new enterprise workspace without initial owner assignment")
+    @inner_api_ns.expect(inner_api_ns.models[WorkspaceOwnerlessPayload.__name__])
     @inner_api_ns.doc(
         responses={
             200: "Workspace created successfully",
@@ -70,10 +81,9 @@ class EnterpriseWorkspaceNoOwnerEmail(Resource):
         }
     )
     def post(self):
-        parser = reqparse.RequestParser().add_argument("name", type=str, required=True, location="json")
-        args = parser.parse_args()
+        args = WorkspaceOwnerlessPayload.model_validate(inner_api_ns.payload or {})
 
-        tenant = TenantService.create_tenant(args["name"], is_from_dashboard=True)
+        tenant = TenantService.create_tenant(args.name, is_from_dashboard=True)
 
         tenant_was_created.send(tenant)
 

api/controllers/mcp/mcp.py

@@ -1,10 +1,11 @@
-from typing import Union
+from typing import Any, Union
 
 from flask import Response
-from flask_restx import Resource, reqparse
-from pydantic import ValidationError
+from flask_restx import Resource
+from pydantic import BaseModel, Field, ValidationError
 from sqlalchemy.orm import Session
 
+from controllers.common.schema import register_schema_model
 from controllers.console.app.mcp_server import AppMCPServerStatus
 from controllers.mcp import mcp_ns
 from core.app.app_config.entities import VariableEntity
@@ -24,27 +25,19 @@ class MCPRequestError(Exception):
         super().__init__(message)
 
 
-def int_or_str(value):
-    """Validate that a value is either an integer or string."""
-    if isinstance(value, (int, str)):
-        return value
-    else:
-        return None
+class MCPRequestPayload(BaseModel):
+    jsonrpc: str = Field(description="JSON-RPC version (should be '2.0')")
+    method: str = Field(description="The method to invoke")
+    params: dict[str, Any] | None = Field(default=None, description="Parameters for the method")
+    id: int | str | None = Field(default=None, description="Request ID for tracking responses")
 
 
-# Define parser for both documentation and validation
-mcp_request_parser = (
-    reqparse.RequestParser()
-    .add_argument("jsonrpc", type=str, required=True, location="json", help="JSON-RPC version (should be '2.0')")
-    .add_argument("method", type=str, required=True, location="json", help="The method to invoke")
-    .add_argument("params", type=dict, required=False, location="json", help="Parameters for the method")
-    .add_argument("id", type=int_or_str, required=False, location="json", help="Request ID for tracking responses")
-)
+register_schema_model(mcp_ns, MCPRequestPayload)
 
 
 @mcp_ns.route("/server/<string:server_code>/mcp")
 class MCPAppApi(Resource):
-    @mcp_ns.expect(mcp_request_parser)
+    @mcp_ns.expect(mcp_ns.models[MCPRequestPayload.__name__])
     @mcp_ns.doc("handle_mcp_request")
     @mcp_ns.doc(description="Handle Model Context Protocol (MCP) requests for a specific server")
     @mcp_ns.doc(params={"server_code": "Unique identifier for the MCP server"})
@@ -70,9 +63,9 @@ class MCPAppApi(Resource):
         Raises:
             ValidationError: Invalid request format or parameters
         """
-        args = mcp_request_parser.parse_args()
-        request_id: Union[int, str] | None = args.get("id")
-        mcp_request = self._parse_mcp_request(args)
+        args = MCPRequestPayload.model_validate(mcp_ns.payload or {})
+        request_id: Union[int, str] | None = args.id
+        mcp_request = self._parse_mcp_request(args.model_dump(exclude_none=True))
 
         with Session(db.engine, expire_on_commit=False) as session:
             # Get MCP server and app

api/controllers/service_api/app/annotation.py

@@ -1,9 +1,11 @@
 from typing import Literal
 
 from flask import request
-from flask_restx import Api, Namespace, Resource, fields, reqparse
+from flask_restx import Api, Namespace, Resource, fields
 from flask_restx.api import HTTPStatus
+from pydantic import BaseModel, Field
 
+from controllers.common.schema import register_schema_models
 from controllers.console.wraps import edit_permission_required
 from controllers.service_api import service_api_ns
 from controllers.service_api.wraps import validate_app_token
@@ -12,26 +14,24 @@ from fields.annotation_fields import annotation_fields, build_annotation_model
 from models.model import App
 from services.annotation_service import AppAnnotationService
 
-# Define parsers for annotation API
-annotation_create_parser = (
-    reqparse.RequestParser()
-    .add_argument("question", required=True, type=str, location="json", help="Annotation question")
-    .add_argument("answer", required=True, type=str, location="json", help="Annotation answer")
-)
 
-annotation_reply_action_parser = (
-    reqparse.RequestParser()
-    .add_argument(
-        "score_threshold", required=True, type=float, location="json", help="Score threshold for annotation matching"
-    )
-    .add_argument("embedding_provider_name", required=True, type=str, location="json", help="Embedding provider name")
-    .add_argument("embedding_model_name", required=True, type=str, location="json", help="Embedding model name")
-)
+class AnnotationCreatePayload(BaseModel):
+    question: str = Field(description="Annotation question")
+    answer: str = Field(description="Annotation answer")
+
+
+class AnnotationReplyActionPayload(BaseModel):
+    score_threshold: float = Field(description="Score threshold for annotation matching")
+    embedding_provider_name: str = Field(description="Embedding provider name")
+    embedding_model_name: str = Field(description="Embedding model name")
+
+
+register_schema_models(service_api_ns, AnnotationCreatePayload, AnnotationReplyActionPayload)
 
 
 @service_api_ns.route("/apps/annotation-reply/<string:action>")
 class AnnotationReplyActionApi(Resource):
-    @service_api_ns.expect(annotation_reply_action_parser)
+    @service_api_ns.expect(service_api_ns.models[AnnotationReplyActionPayload.__name__])
     @service_api_ns.doc("annotation_reply_action")
     @service_api_ns.doc(description="Enable or disable annotation reply feature")
     @service_api_ns.doc(params={"action": "Action to perform: 'enable' or 'disable'"})
@@ -44,7 +44,7 @@ class AnnotationReplyActionApi(Resource):
     @validate_app_token
     def post(self, app_model: App, action: Literal["enable", "disable"]):
         """Enable or disable annotation reply feature."""
-        args = annotation_reply_action_parser.parse_args()
+        args = AnnotationReplyActionPayload.model_validate(service_api_ns.payload or {}).model_dump()
         if action == "enable":
             result = AppAnnotationService.enable_app_annotation(args, app_model.id)
         elif action == "disable":
@@ -126,7 +126,7 @@ class AnnotationListApi(Resource):
             "page": page,
         }
 
-    @service_api_ns.expect(annotation_create_parser)
+    @service_api_ns.expect(service_api_ns.models[AnnotationCreatePayload.__name__])
     @service_api_ns.doc("create_annotation")
     @service_api_ns.doc(description="Create a new annotation")
     @service_api_ns.doc(
@@ -139,14 +139,14 @@ class AnnotationListApi(Resource):
     @service_api_ns.marshal_with(build_annotation_model(service_api_ns), code=HTTPStatus.CREATED)
     def post(self, app_model: App):
         """Create a new annotation."""
-        args = annotation_create_parser.parse_args()
+        args = AnnotationCreatePayload.model_validate(service_api_ns.payload or {}).model_dump()
         annotation = AppAnnotationService.insert_app_annotation_directly(args, app_model.id)
         return annotation, 201
 
 
 @service_api_ns.route("/apps/annotations/<uuid:annotation_id>")
 class AnnotationUpdateDeleteApi(Resource):
-    @service_api_ns.expect(annotation_create_parser)
+    @service_api_ns.expect(service_api_ns.models[AnnotationCreatePayload.__name__])
     @service_api_ns.doc("update_annotation")
     @service_api_ns.doc(description="Update an existing annotation")
     @service_api_ns.doc(params={"annotation_id": "Annotation ID"})
@@ -163,7 +163,7 @@ class AnnotationUpdateDeleteApi(Resource):
     @service_api_ns.marshal_with(build_annotation_model(service_api_ns))
     def put(self, app_model: App, annotation_id: str):
         """Update an existing annotation."""
-        args = annotation_create_parser.parse_args()
+        args = AnnotationCreatePayload.model_validate(service_api_ns.payload or {}).model_dump()
         annotation = AppAnnotationService.update_app_annotation_directly(args, app_model.id, annotation_id)
         return annotation
 

api/controllers/service_api/app/audio.py

@@ -1,10 +1,12 @@
 import logging
 
 from flask import request
-from flask_restx import Resource, reqparse
+from flask_restx import Resource
+from pydantic import BaseModel, Field
 from werkzeug.exceptions import InternalServerError
 
 import services
+from controllers.common.schema import register_schema_model
 from controllers.service_api import service_api_ns
 from controllers.service_api.app.error import (
     AppUnavailableError,
@@ -84,19 +86,19 @@ class AudioApi(Resource):
             raise InternalServerError()
 
 
-# Define parser for text-to-audio API
-text_to_audio_parser = (
-    reqparse.RequestParser()
-    .add_argument("message_id", type=str, required=False, location="json", help="Message ID")
-    .add_argument("voice", type=str, location="json", help="Voice to use for TTS")
-    .add_argument("text", type=str, location="json", help="Text to convert to audio")
-    .add_argument("streaming", type=bool, location="json", help="Enable streaming response")
-)
+class TextToAudioPayload(BaseModel):
+    message_id: str | None = Field(default=None, description="Message ID")
+    voice: str | None = Field(default=None, description="Voice to use for TTS")
+    text: str | None = Field(default=None, description="Text to convert to audio")
+    streaming: bool | None = Field(default=None, description="Enable streaming response")
+
+
+register_schema_model(service_api_ns, TextToAudioPayload)
 
 
 @service_api_ns.route("/text-to-audio")
 class TextApi(Resource):
-    @service_api_ns.expect(text_to_audio_parser)
+    @service_api_ns.expect(service_api_ns.models[TextToAudioPayload.__name__])
     @service_api_ns.doc("text_to_audio")
     @service_api_ns.doc(description="Convert text to audio using text-to-speech")
     @service_api_ns.doc(
@@ -114,11 +116,11 @@ class TextApi(Resource):
         Converts the provided text to audio using the specified voice.
         """
         try:
-            args = text_to_audio_parser.parse_args()
+            payload = TextToAudioPayload.model_validate(service_api_ns.payload or {})
 
-            message_id = args.get("message_id", None)
-            text = args.get("text", None)
-            voice = args.get("voice", None)
+            message_id = payload.message_id
+            text = payload.text
+            voice = payload.voice
             response = AudioService.transcript_tts(
                 app_model=app_model, text=text, voice=voice, end_user=end_user.external_user_id, message_id=message_id
             )

api/controllers/service_api/app/completion.py

@@ -1,10 +1,14 @@
 import logging
+from typing import Any, Literal
+from uuid import UUID
 
 from flask import request
-from flask_restx import Resource, reqparse
+from flask_restx import Resource
+from pydantic import BaseModel, Field
 from werkzeug.exceptions import BadRequest, InternalServerError, NotFound
 
 import services
+from controllers.common.schema import register_schema_models
 from controllers.service_api import service_api_ns
 from controllers.service_api.app.error import (
     AppUnavailableError,
@@ -26,7 +30,6 @@ from core.errors.error import (
 from core.helper.trace_id_helper import get_external_trace_id
 from core.model_runtime.errors.invoke import InvokeError
 from libs import helper
-from libs.helper import uuid_value
 from models.model import App, AppMode, EndUser
 from services.app_generate_service import AppGenerateService
 from services.app_task_service import AppTaskService
@@ -36,40 +39,31 @@ from services.errors.llm import InvokeRateLimitError
 logger = logging.getLogger(__name__)
 
 
-# Define parser for completion API
-completion_parser = (
-    reqparse.RequestParser()
-    .add_argument("inputs", type=dict, required=True, location="json", help="Input parameters for completion")
-    .add_argument("query", type=str, location="json", default="", help="The query string")
-    .add_argument("files", type=list, required=False, location="json", help="List of file attachments")
-    .add_argument("response_mode", type=str, choices=["blocking", "streaming"], location="json", help="Response mode")
-    .add_argument("retriever_from", type=str, required=False, default="dev", location="json", help="Retriever source")
-)
+class CompletionRequestPayload(BaseModel):
+    inputs: dict[str, Any]
+    query: str = Field(default="")
+    files: list[dict[str, Any]] | None = None
+    response_mode: Literal["blocking", "streaming"] | None = None
+    retriever_from: str = Field(default="dev")
 
-# Define parser for chat API
-chat_parser = (
-    reqparse.RequestParser()
-    .add_argument("inputs", type=dict, required=True, location="json", help="Input parameters for chat")
-    .add_argument("query", type=str, required=True, location="json", help="The chat query")
-    .add_argument("files", type=list, required=False, location="json", help="List of file attachments")
-    .add_argument("response_mode", type=str, choices=["blocking", "streaming"], location="json", help="Response mode")
-    .add_argument("conversation_id", type=uuid_value, location="json", help="Existing conversation ID")
-    .add_argument("retriever_from", type=str, required=False, default="dev", location="json", help="Retriever source")
-    .add_argument(
-        "auto_generate_name",
-        type=bool,
-        required=False,
-        default=True,
-        location="json",
-        help="Auto generate conversation name",
-    )
-    .add_argument("workflow_id", type=str, required=False, location="json", help="Workflow ID for advanced chat")
-)
+
+class ChatRequestPayload(BaseModel):
+    inputs: dict[str, Any]
+    query: str
+    files: list[dict[str, Any]] | None = None
+    response_mode: Literal["blocking", "streaming"] | None = None
+    conversation_id: UUID | None = None
+    retriever_from: str = Field(default="dev")
+    auto_generate_name: bool = Field(default=True, description="Auto generate conversation name")
+    workflow_id: str | None = Field(default=None, description="Workflow ID for advanced chat")
+
+
+register_schema_models(service_api_ns, CompletionRequestPayload, ChatRequestPayload)
 
 
 @service_api_ns.route("/completion-messages")
 class CompletionApi(Resource):
-    @service_api_ns.expect(completion_parser)
+    @service_api_ns.expect(service_api_ns.models[CompletionRequestPayload.__name__])
     @service_api_ns.doc("create_completion")
     @service_api_ns.doc(description="Create a completion for the given prompt")
     @service_api_ns.doc(
@@ -91,12 +85,13 @@ class CompletionApi(Resource):
         if app_model.mode != AppMode.COMPLETION:
             raise AppUnavailableError()
 
-        args = completion_parser.parse_args()
+        payload = CompletionRequestPayload.model_validate(service_api_ns.payload or {})
         external_trace_id = get_external_trace_id(request)
+        args = payload.model_dump(exclude_none=True)
         if external_trace_id:
             args["external_trace_id"] = external_trace_id
 
-        streaming = args["response_mode"] == "streaming"
+        streaming = payload.response_mode == "streaming"
 
         args["auto_generate_name"] = False
 
@@ -162,7 +157,7 @@ class CompletionStopApi(Resource):
 
 @service_api_ns.route("/chat-messages")
 class ChatApi(Resource):
-    @service_api_ns.expect(chat_parser)
+    @service_api_ns.expect(service_api_ns.models[ChatRequestPayload.__name__])
     @service_api_ns.doc("create_chat_message")
     @service_api_ns.doc(description="Send a message in a chat conversation")
     @service_api_ns.doc(
@@ -186,13 +181,14 @@ class ChatApi(Resource):
         if app_mode not in {AppMode.CHAT, AppMode.AGENT_CHAT, AppMode.ADVANCED_CHAT}:
             raise NotChatAppError()
 
-        args = chat_parser.parse_args()
+        payload = ChatRequestPayload.model_validate(service_api_ns.payload or {})
 
         external_trace_id = get_external_trace_id(request)
+        args = payload.model_dump(exclude_none=True)
         if external_trace_id:
             args["external_trace_id"] = external_trace_id
 
-        streaming = args["response_mode"] == "streaming"
+        streaming = payload.response_mode == "streaming"
 
         try:
             response = AppGenerateService.generate(

api/controllers/service_api/app/conversation.py

@@ -1,10 +1,15 @@
-from flask_restx import Resource, reqparse
+from typing import Any, Literal
+from uuid import UUID
+
+from flask import request
+from flask_restx import Resource
 from flask_restx._http import HTTPStatus
-from flask_restx.inputs import int_range
+from pydantic import BaseModel, Field
 from sqlalchemy.orm import Session
 from werkzeug.exceptions import BadRequest, NotFound
 
 import services
+from controllers.common.schema import register_schema_models
 from controllers.service_api import service_api_ns
 from controllers.service_api.app.error import NotChatAppError
 from controllers.service_api.wraps import FetchUserArg, WhereisUserArg, validate_app_token
@@ -19,74 +24,44 @@ from fields.conversation_variable_fields import (
     build_conversation_variable_infinite_scroll_pagination_model,
     build_conversation_variable_model,
 )
-from libs.helper import uuid_value
 from models.model import App, AppMode, EndUser
 from services.conversation_service import ConversationService
 
-# Define parsers for conversation APIs
-conversation_list_parser = (
-    reqparse.RequestParser()
-    .add_argument("last_id", type=uuid_value, location="args", help="Last conversation ID for pagination")
-    .add_argument(
-        "limit",
-        type=int_range(1, 100),
-        required=False,
-        default=20,
-        location="args",
-        help="Number of conversations to return",
-    )
-    .add_argument(
-        "sort_by",
-        type=str,
-        choices=["created_at", "-created_at", "updated_at", "-updated_at"],
-        required=False,
-        default="-updated_at",
-        location="args",
-        help="Sort order for conversations",
-    )
-)
 
-conversation_rename_parser = (
-    reqparse.RequestParser()
-    .add_argument("name", type=str, required=False, location="json", help="New conversation name")
-    .add_argument(
-        "auto_generate",
-        type=bool,
-        required=False,
-        default=False,
-        location="json",
-        help="Auto-generate conversation name",
+class ConversationListQuery(BaseModel):
+    last_id: UUID | None = Field(default=None, description="Last conversation ID for pagination")
+    limit: int = Field(default=20, ge=1, le=100, description="Number of conversations to return")
+    sort_by: Literal["created_at", "-created_at", "updated_at", "-updated_at"] = Field(
+        default="-updated_at", description="Sort order for conversations"
     )
-)
 
-conversation_variables_parser = (
-    reqparse.RequestParser()
-    .add_argument("last_id", type=uuid_value, location="args", help="Last variable ID for pagination")
-    .add_argument(
-        "limit",
-        type=int_range(1, 100),
-        required=False,
-        default=20,
-        location="args",
-        help="Number of variables to return",
-    )
-)
 
-conversation_variable_update_parser = reqparse.RequestParser().add_argument(
-    # using lambda is for passing the already-typed value without modification
-    # if no lambda, it will be converted to string
-    # the string cannot be converted using json.loads
-    "value",
-    required=True,
-    location="json",
-    type=lambda x: x,
-    help="New value for the conversation variable",
+class ConversationRenamePayload(BaseModel):
+    name: str = Field(description="New conversation name")
+    auto_generate: bool = Field(default=False, description="Auto-generate conversation name")
+
+
+class ConversationVariablesQuery(BaseModel):
+    last_id: UUID | None = Field(default=None, description="Last variable ID for pagination")
+    limit: int = Field(default=20, ge=1, le=100, description="Number of variables to return")
+
+
+class ConversationVariableUpdatePayload(BaseModel):
+    value: Any
+
+
+register_schema_models(
+    service_api_ns,
+    ConversationListQuery,
+    ConversationRenamePayload,
+    ConversationVariablesQuery,
+    ConversationVariableUpdatePayload,
 )
 
 
 @service_api_ns.route("/conversations")
 class ConversationApi(Resource):
-    @service_api_ns.expect(conversation_list_parser)
+    @service_api_ns.expect(service_api_ns.models[ConversationListQuery.__name__])
     @service_api_ns.doc("list_conversations")
     @service_api_ns.doc(description="List all conversations for the current user")
     @service_api_ns.doc(
@@ -107,7 +82,8 @@ class ConversationApi(Resource):
         if app_mode not in {AppMode.CHAT, AppMode.AGENT_CHAT, AppMode.ADVANCED_CHAT}:
             raise NotChatAppError()
 
-        args = conversation_list_parser.parse_args()
+        query_args = ConversationListQuery.model_validate(request.args.to_dict())
+        last_id = str(query_args.last_id) if query_args.last_id else None
 
         try:
             with Session(db.engine) as session:
@@ -115,10 +91,10 @@ class ConversationApi(Resource):
                     session=session,
                     app_model=app_model,
                     user=end_user,
-                    last_id=args["last_id"],
-                    limit=args["limit"],
+                    last_id=last_id,
+                    limit=query_args.limit,
                     invoke_from=InvokeFrom.SERVICE_API,
-                    sort_by=args["sort_by"],
+                    sort_by=query_args.sort_by,
                 )
         except services.errors.conversation.LastConversationNotExistsError:
             raise NotFound("Last Conversation Not Exists.")
@@ -155,7 +131,7 @@ class ConversationDetailApi(Resource):
 
 @service_api_ns.route("/conversations/<uuid:c_id>/name")
 class ConversationRenameApi(Resource):
-    @service_api_ns.expect(conversation_rename_parser)
+    @service_api_ns.expect(service_api_ns.models[ConversationRenamePayload.__name__])
     @service_api_ns.doc("rename_conversation")
     @service_api_ns.doc(description="Rename a conversation or auto-generate a name")
     @service_api_ns.doc(params={"c_id": "Conversation ID"})
@@ -176,17 +152,17 @@ class ConversationRenameApi(Resource):
 
         conversation_id = str(c_id)
 
-        args = conversation_rename_parser.parse_args()
+        payload = ConversationRenamePayload.model_validate(service_api_ns.payload or {})
 
         try:
-            return ConversationService.rename(app_model, conversation_id, end_user, args["name"], args["auto_generate"])
+            return ConversationService.rename(app_model, conversation_id, end_user, payload.name, payload.auto_generate)
         except services.errors.conversation.ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")
 
 
 @service_api_ns.route("/conversations/<uuid:c_id>/variables")
 class ConversationVariablesApi(Resource):
-    @service_api_ns.expect(conversation_variables_parser)
+    @service_api_ns.expect(service_api_ns.models[ConversationVariablesQuery.__name__])
     @service_api_ns.doc("list_conversation_variables")
     @service_api_ns.doc(description="List all variables for a conversation")
     @service_api_ns.doc(params={"c_id": "Conversation ID"})
@@ -211,11 +187,12 @@ class ConversationVariablesApi(Resource):
 
         conversation_id = str(c_id)
 
-        args = conversation_variables_parser.parse_args()
+        query_args = ConversationVariablesQuery.model_validate(request.args.to_dict())
+        last_id = str(query_args.last_id) if query_args.last_id else None
 
         try:
             return ConversationService.get_conversational_variable(
-                app_model, conversation_id, end_user, args["limit"], args["last_id"]
+                app_model, conversation_id, end_user, query_args.limit, last_id
             )
         except services.errors.conversation.ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")
@@ -223,7 +200,7 @@ class ConversationVariablesApi(Resource):
 
 @service_api_ns.route("/conversations/<uuid:c_id>/variables/<uuid:variable_id>")
 class ConversationVariableDetailApi(Resource):
-    @service_api_ns.expect(conversation_variable_update_parser)
+    @service_api_ns.expect(service_api_ns.models[ConversationVariableUpdatePayload.__name__])
     @service_api_ns.doc("update_conversation_variable")
     @service_api_ns.doc(description="Update a conversation variable's value")
     @service_api_ns.doc(params={"c_id": "Conversation ID", "variable_id": "Variable ID"})
@@ -250,11 +227,11 @@ class ConversationVariableDetailApi(Resource):
         conversation_id = str(c_id)
         variable_id = str(variable_id)
 
-        args = conversation_variable_update_parser.parse_args()
+        payload = ConversationVariableUpdatePayload.model_validate(service_api_ns.payload or {})
 
         try:
             return ConversationService.update_conversation_variable(
-                app_model, conversation_id, variable_id, end_user, args["value"]
+                app_model, conversation_id, variable_id, end_user, payload.value
             )
         except services.errors.conversation.ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")

api/controllers/service_api/app/file_preview.py

@@ -1,9 +1,11 @@
 import logging
 from urllib.parse import quote
 
-from flask import Response
-from flask_restx import Resource, reqparse
+from flask import Response, request
+from flask_restx import Resource
+from pydantic import BaseModel, Field
 
+from controllers.common.schema import register_schema_model
 from controllers.service_api import service_api_ns
 from controllers.service_api.app.error import (
     FileAccessDeniedError,
@@ -17,10 +19,11 @@ from models.model import App, EndUser, Message, MessageFile, UploadFile
 logger = logging.getLogger(__name__)
 
 
-# Define parser for file preview API
-file_preview_parser = reqparse.RequestParser().add_argument(
-    "as_attachment", type=bool, required=False, default=False, location="args", help="Download as attachment"
-)
+class FilePreviewQuery(BaseModel):
+    as_attachment: bool = Field(default=False, description="Download as attachment")
+
+
+register_schema_model(service_api_ns, FilePreviewQuery)
 
 
 @service_api_ns.route("/files/<uuid:file_id>/preview")
@@ -32,7 +35,7 @@ class FilePreviewApi(Resource):
     Files can only be accessed if they belong to messages within the requesting app's context.
     """
 
-    @service_api_ns.expect(file_preview_parser)
+    @service_api_ns.expect(service_api_ns.models[FilePreviewQuery.__name__])
     @service_api_ns.doc("preview_file")
     @service_api_ns.doc(description="Preview or download a file uploaded via Service API")
     @service_api_ns.doc(params={"file_id": "UUID of the file to preview"})
@@ -55,7 +58,7 @@ class FilePreviewApi(Resource):
         file_id = str(file_id)
 
         # Parse query parameters
-        args = file_preview_parser.parse_args()
+        args = FilePreviewQuery.model_validate(request.args.to_dict())
 
         # Validate file ownership and get file objects
         _, upload_file = self._validate_file_ownership(file_id, app_model.id)
@@ -67,7 +70,7 @@ class FilePreviewApi(Resource):
             raise FileNotFoundError(f"Failed to load file content: {str(e)}")
 
         # Build response with appropriate headers
-        response = self._build_file_response(generator, upload_file, args["as_attachment"])
+        response = self._build_file_response(generator, upload_file, args.as_attachment)
 
         return response
 

api/controllers/service_api/app/message.py

@@ -1,11 +1,15 @@
 import json
 import logging
+from typing import Literal
+from uuid import UUID
 
-from flask_restx import Api, Namespace, Resource, fields, reqparse
-from flask_restx.inputs import int_range
+from flask import request
+from flask_restx import Namespace, Resource, fields
+from pydantic import BaseModel, Field
 from werkzeug.exceptions import BadRequest, InternalServerError, NotFound
 
 import services
+from controllers.common.schema import register_schema_models
 from controllers.service_api import service_api_ns
 from controllers.service_api.app.error import NotChatAppError
 from controllers.service_api.wraps import FetchUserArg, WhereisUserArg, validate_app_token
@@ -13,7 +17,7 @@ from core.app.entities.app_invoke_entities import InvokeFrom
 from fields.conversation_fields import build_message_file_model
 from fields.message_fields import build_agent_thought_model, build_feedback_model
 from fields.raws import FilesContainedField
-from libs.helper import TimestampField, uuid_value
+from libs.helper import TimestampField
 from models.model import App, AppMode, EndUser
 from services.errors.message import (
     FirstMessageNotExistsError,
@@ -25,42 +29,26 @@ from services.message_service import MessageService
 logger = logging.getLogger(__name__)
 
 
-# Define parsers for message APIs
-message_list_parser = (
-    reqparse.RequestParser()
-    .add_argument("conversation_id", required=True, type=uuid_value, location="args", help="Conversation ID")
-    .add_argument("first_id", type=uuid_value, location="args", help="First message ID for pagination")
-    .add_argument(
-        "limit",
-        type=int_range(1, 100),
-        required=False,
-        default=20,
-        location="args",
-        help="Number of messages to return",
-    )
-)
-
-message_feedback_parser = (
-    reqparse.RequestParser()
-    .add_argument("rating", type=str, choices=["like", "dislike", None], location="json", help="Feedback rating")
-    .add_argument("content", type=str, location="json", help="Feedback content")
-)
-
-feedback_list_parser = (
-    reqparse.RequestParser()
-    .add_argument("page", type=int, default=1, location="args", help="Page number")
-    .add_argument(
-        "limit",
-        type=int_range(1, 101),
-        required=False,
-        default=20,
-        location="args",
-        help="Number of feedbacks per page",
-    )
-)
+class MessageListQuery(BaseModel):
+    conversation_id: UUID
+    first_id: UUID | None = None
+    limit: int = Field(default=20, ge=1, le=100, description="Number of messages to return")
 
 
-def build_message_model(api_or_ns: Api | Namespace):
+class MessageFeedbackPayload(BaseModel):
+    rating: Literal["like", "dislike"] | None = Field(default=None, description="Feedback rating")
+    content: str | None = Field(default=None, description="Feedback content")
+
+
+class FeedbackListQuery(BaseModel):
+    page: int = Field(default=1, ge=1, description="Page number")
+    limit: int = Field(default=20, ge=1, le=101, description="Number of feedbacks per page")
+
+
+register_schema_models(service_api_ns, MessageListQuery, MessageFeedbackPayload, FeedbackListQuery)
+
+
+def build_message_model(api_or_ns: Namespace):
     """Build the message model for the API or Namespace."""
     # First build the nested models
     feedback_model = build_feedback_model(api_or_ns)
@@ -90,7 +78,7 @@ def build_message_model(api_or_ns: Api | Namespace):
     return api_or_ns.model("Message", message_fields)
 
 
-def build_message_infinite_scroll_pagination_model(api_or_ns: Api | Namespace):
+def build_message_infinite_scroll_pagination_model(api_or_ns: Namespace):
     """Build the message infinite scroll pagination model for the API or Namespace."""
     # Build the nested message model first
     message_model = build_message_model(api_or_ns)
@@ -105,7 +93,7 @@ def build_message_infinite_scroll_pagination_model(api_or_ns: Api | Namespace):
 
 @service_api_ns.route("/messages")
 class MessageListApi(Resource):
-    @service_api_ns.expect(message_list_parser)
+    @service_api_ns.expect(service_api_ns.models[MessageListQuery.__name__])
     @service_api_ns.doc("list_messages")
     @service_api_ns.doc(description="List messages in a conversation")
     @service_api_ns.doc(
@@ -126,11 +114,13 @@ class MessageListApi(Resource):
         if app_mode not in {AppMode.CHAT, AppMode.AGENT_CHAT, AppMode.ADVANCED_CHAT}:
             raise NotChatAppError()
 
-        args = message_list_parser.parse_args()
+        query_args = MessageListQuery.model_validate(request.args.to_dict())
+        conversation_id = str(query_args.conversation_id)
+        first_id = str(query_args.first_id) if query_args.first_id else None
 
         try:
             return MessageService.pagination_by_first_id(
-                app_model, end_user, args["conversation_id"], args["first_id"], args["limit"]
+                app_model, end_user, conversation_id, first_id, query_args.limit
             )
         except services.errors.conversation.ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")
@@ -140,7 +130,7 @@ class MessageListApi(Resource):
 
 @service_api_ns.route("/messages/<uuid:message_id>/feedbacks")
 class MessageFeedbackApi(Resource):
-    @service_api_ns.expect(message_feedback_parser)
+    @service_api_ns.expect(service_api_ns.models[MessageFeedbackPayload.__name__])
     @service_api_ns.doc("create_message_feedback")
     @service_api_ns.doc(description="Submit feedback for a message")
     @service_api_ns.doc(params={"message_id": "Message ID"})
@@ -159,15 +149,15 @@ class MessageFeedbackApi(Resource):
         """
         message_id = str(message_id)
 
-        args = message_feedback_parser.parse_args()
+        payload = MessageFeedbackPayload.model_validate(service_api_ns.payload or {})
 
         try:
             MessageService.create_feedback(
                 app_model=app_model,
                 message_id=message_id,
                 user=end_user,
-                rating=args.get("rating"),
-                content=args.get("content"),
+                rating=payload.rating,
+                content=payload.content,
             )
         except MessageNotExistsError:
             raise NotFound("Message Not Exists.")
@@ -177,7 +167,7 @@ class MessageFeedbackApi(Resource):
 
 @service_api_ns.route("/app/feedbacks")
 class AppGetFeedbacksApi(Resource):
-    @service_api_ns.expect(feedback_list_parser)
+    @service_api_ns.expect(service_api_ns.models[FeedbackListQuery.__name__])
     @service_api_ns.doc("get_app_feedbacks")
     @service_api_ns.doc(description="Get all feedbacks for the application")
     @service_api_ns.doc(
@@ -192,8 +182,8 @@ class AppGetFeedbacksApi(Resource):
 
         Returns paginated list of all feedback submitted for messages in this app.
         """
-        args = feedback_list_parser.parse_args()
-        feedbacks = MessageService.get_all_messages_feedbacks(app_model, page=args["page"], limit=args["limit"])
+        query_args = FeedbackListQuery.model_validate(request.args.to_dict())
+        feedbacks = MessageService.get_all_messages_feedbacks(app_model, page=query_args.page, limit=query_args.limit)
         return {"data": feedbacks}
 
 

api/controllers/service_api/app/workflow.py

@@ -1,12 +1,14 @@
 import logging
+from typing import Any, Literal
 
 from dateutil.parser import isoparse
 from flask import request
-from flask_restx import Api, Namespace, Resource, fields, reqparse
-from flask_restx.inputs import int_range
+from flask_restx import Api, Namespace, Resource, fields
+from pydantic import BaseModel, Field
 from sqlalchemy.orm import Session, sessionmaker
 from werkzeug.exceptions import BadRequest, InternalServerError, NotFound
 
+from controllers.common.schema import register_schema_models
 from controllers.service_api import service_api_ns
 from controllers.service_api.app.error import (
     CompletionRequestError,
@@ -41,37 +43,25 @@ from services.workflow_app_service import WorkflowAppService
 
 logger = logging.getLogger(__name__)
 
-# Define parsers for workflow APIs
-workflow_run_parser = (
-    reqparse.RequestParser()
-    .add_argument("inputs", type=dict, required=True, nullable=False, location="json")
-    .add_argument("files", type=list, required=False, location="json")
-    .add_argument("response_mode", type=str, choices=["blocking", "streaming"], location="json")
-)
 
-workflow_log_parser = (
-    reqparse.RequestParser()
-    .add_argument("keyword", type=str, location="args")
-    .add_argument("status", type=str, choices=["succeeded", "failed", "stopped"], location="args")
-    .add_argument("created_at__before", type=str, location="args")
-    .add_argument("created_at__after", type=str, location="args")
-    .add_argument(
-        "created_by_end_user_session_id",
-        type=str,
-        location="args",
-        required=False,
-        default=None,
-    )
-    .add_argument(
-        "created_by_account",
-        type=str,
-        location="args",
-        required=False,
-        default=None,
-    )
-    .add_argument("page", type=int_range(1, 99999), default=1, location="args")
-    .add_argument("limit", type=int_range(1, 100), default=20, location="args")
-)
+class WorkflowRunPayload(BaseModel):
+    inputs: dict[str, Any]
+    files: list[dict[str, Any]] | None = None
+    response_mode: Literal["blocking", "streaming"] | None = None
+
+
+class WorkflowLogQuery(BaseModel):
+    keyword: str | None = None
+    status: Literal["succeeded", "failed", "stopped"] | None = None
+    created_at__before: str | None = None
+    created_at__after: str | None = None
+    created_by_end_user_session_id: str | None = None
+    created_by_account: str | None = None
+    page: int = Field(default=1, ge=1, le=99999)
+    limit: int = Field(default=20, ge=1, le=100)
+
+
+register_schema_models(service_api_ns, WorkflowRunPayload, WorkflowLogQuery)
 
 workflow_run_fields = {
     "id": fields.String,
@@ -130,7 +120,7 @@ class WorkflowRunDetailApi(Resource):
 
 @service_api_ns.route("/workflows/run")
 class WorkflowRunApi(Resource):
-    @service_api_ns.expect(workflow_run_parser)
+    @service_api_ns.expect(service_api_ns.models[WorkflowRunPayload.__name__])
     @service_api_ns.doc("run_workflow")
     @service_api_ns.doc(description="Execute a workflow")
     @service_api_ns.doc(
@@ -154,11 +144,12 @@ class WorkflowRunApi(Resource):
         if app_mode != AppMode.WORKFLOW:
             raise NotWorkflowAppError()
 
-        args = workflow_run_parser.parse_args()
+        payload = WorkflowRunPayload.model_validate(service_api_ns.payload or {})
+        args = payload.model_dump(exclude_none=True)
         external_trace_id = get_external_trace_id(request)
         if external_trace_id:
             args["external_trace_id"] = external_trace_id
-        streaming = args.get("response_mode") == "streaming"
+        streaming = payload.response_mode == "streaming"
 
         try:
             response = AppGenerateService.generate(
@@ -185,7 +176,7 @@ class WorkflowRunApi(Resource):
 
 @service_api_ns.route("/workflows/<string:workflow_id>/run")
 class WorkflowRunByIdApi(Resource):
-    @service_api_ns.expect(workflow_run_parser)
+    @service_api_ns.expect(service_api_ns.models[WorkflowRunPayload.__name__])
     @service_api_ns.doc("run_workflow_by_id")
     @service_api_ns.doc(description="Execute a specific workflow by ID")
     @service_api_ns.doc(params={"workflow_id": "Workflow ID to execute"})
@@ -209,7 +200,8 @@ class WorkflowRunByIdApi(Resource):
         if app_mode != AppMode.WORKFLOW:
             raise NotWorkflowAppError()
 
-        args = workflow_run_parser.parse_args()
+        payload = WorkflowRunPayload.model_validate(service_api_ns.payload or {})
+        args = payload.model_dump(exclude_none=True)
 
         # Add workflow_id to args for AppGenerateService
         args["workflow_id"] = workflow_id
@@ -217,7 +209,7 @@ class WorkflowRunByIdApi(Resource):
         external_trace_id = get_external_trace_id(request)
         if external_trace_id:
             args["external_trace_id"] = external_trace_id
-        streaming = args.get("response_mode") == "streaming"
+        streaming = payload.response_mode == "streaming"
 
         try:
             response = AppGenerateService.generate(
@@ -279,7 +271,7 @@ class WorkflowTaskStopApi(Resource):
 
 @service_api_ns.route("/workflows/logs")
 class WorkflowAppLogApi(Resource):
-    @service_api_ns.expect(workflow_log_parser)
+    @service_api_ns.expect(service_api_ns.models[WorkflowLogQuery.__name__])
     @service_api_ns.doc("get_workflow_logs")
     @service_api_ns.doc(description="Get workflow execution logs")
     @service_api_ns.doc(
@@ -295,14 +287,11 @@ class WorkflowAppLogApi(Resource):
 
         Returns paginated workflow execution logs with filtering options.
         """
-        args = workflow_log_parser.parse_args()
+        args = WorkflowLogQuery.model_validate(request.args.to_dict())
 
-        args.status = WorkflowExecutionStatus(args.status) if args.status else None
-        if args.created_at__before:
-            args.created_at__before = isoparse(args.created_at__before)
-
-        if args.created_at__after:
-            args.created_at__after = isoparse(args.created_at__after)
+        status = WorkflowExecutionStatus(args.status) if args.status else None
+        created_at_before = isoparse(args.created_at__before) if args.created_at__before else None
+        created_at_after = isoparse(args.created_at__after) if args.created_at__after else None
 
         # get paginate workflow app logs
         workflow_app_service = WorkflowAppService()
@@ -311,9 +300,9 @@ class WorkflowAppLogApi(Resource):
                 session=session,
                 app_model=app_model,
                 keyword=args.keyword,
-                status=args.status,
-                created_at_before=args.created_at__before,
-                created_at_after=args.created_at__after,
+                status=status,
+                created_at_before=created_at_before,
+                created_at_after=created_at_after,
                 page=args.page,
                 limit=args.limit,
                 created_by_end_user_session_id=args.created_by_end_user_session_id,

api/controllers/service_api/dataset/dataset.py

@@ -1,10 +1,12 @@
 from typing import Any, Literal, cast
 
 from flask import request
-from flask_restx import marshal, reqparse
+from flask_restx import marshal
+from pydantic import BaseModel, Field, field_validator
 from werkzeug.exceptions import Forbidden, NotFound
 
 import services
+from controllers.common.schema import register_schema_models
 from controllers.console.wraps import edit_permission_required
 from controllers.service_api import service_api_ns
 from controllers.service_api.dataset.error import DatasetInUseError, DatasetNameDuplicateError, InvalidActionError
@@ -18,173 +20,83 @@ from core.provider_manager import ProviderManager
 from fields.dataset_fields import dataset_detail_fields
 from fields.tag_fields import build_dataset_tag_fields
 from libs.login import current_user
-from libs.validators import validate_description_length
 from models.account import Account
-from models.dataset import Dataset, DatasetPermissionEnum
+from models.dataset import DatasetPermissionEnum
 from models.provider_ids import ModelProviderID
 from services.dataset_service import DatasetPermissionService, DatasetService, DocumentService
 from services.entities.knowledge_entities.knowledge_entities import RetrievalModel
 from services.tag_service import TagService
 
 
-def _validate_name(name):
-    if not name or len(name) < 1 or len(name) > 40:
-        raise ValueError("Name must be between 1 to 40 characters.")
-    return name
+class DatasetCreatePayload(BaseModel):
+    name: str = Field(..., min_length=1, max_length=40)
+    description: str = Field(default="", description="Dataset description (max 400 chars)", max_length=400)
+    indexing_technique: Literal["high_quality", "economy"] | None = None
+    permission: DatasetPermissionEnum | None = DatasetPermissionEnum.ONLY_ME
+    external_knowledge_api_id: str | None = None
+    provider: str = "vendor"
+    external_knowledge_id: str | None = None
+    retrieval_model: RetrievalModel | None = None
+    embedding_model: str | None = None
+    embedding_model_provider: str | None = None
 
 
-# Define parsers for dataset operations
-dataset_create_parser = (
-    reqparse.RequestParser()
-    .add_argument(
-        "name",
-        nullable=False,
-        required=True,
-        help="type is required. Name must be between 1 to 40 characters.",
-        type=_validate_name,
-    )
-    .add_argument(
-        "description",
-        type=validate_description_length,
-        nullable=True,
-        required=False,
-        default="",
-    )
-    .add_argument(
-        "indexing_technique",
-        type=str,
-        location="json",
-        choices=Dataset.INDEXING_TECHNIQUE_LIST,
-        help="Invalid indexing technique.",
-    )
-    .add_argument(
-        "permission",
-        type=str,
-        location="json",
-        choices=(DatasetPermissionEnum.ONLY_ME, DatasetPermissionEnum.ALL_TEAM, DatasetPermissionEnum.PARTIAL_TEAM),
-        help="Invalid permission.",
-        required=False,
-        nullable=False,
-    )
-    .add_argument(
-        "external_knowledge_api_id",
-        type=str,
-        nullable=True,
-        required=False,
-        default="_validate_name",
-    )
-    .add_argument(
-        "provider",
-        type=str,
-        nullable=True,
-        required=False,
-        default="vendor",
-    )
-    .add_argument(
-        "external_knowledge_id",
-        type=str,
-        nullable=True,
-        required=False,
-    )
-    .add_argument("retrieval_model", type=dict, required=False, nullable=True, location="json")
-    .add_argument("embedding_model", type=str, required=False, nullable=True, location="json")
-    .add_argument("embedding_model_provider", type=str, required=False, nullable=True, location="json")
-)
+class DatasetUpdatePayload(BaseModel):
+    name: str | None = Field(default=None, min_length=1, max_length=40)
+    description: str | None = Field(default=None, description="Dataset description (max 400 chars)", max_length=400)
+    indexing_technique: Literal["high_quality", "economy"] | None = None
+    permission: DatasetPermissionEnum | None = None
+    embedding_model: str | None = None
+    embedding_model_provider: str | None = None
+    retrieval_model: RetrievalModel | None = None
+    partial_member_list: list[str] | None = None
+    external_retrieval_model: dict[str, Any] | None = None
+    external_knowledge_id: str | None = None
+    external_knowledge_api_id: str | None = None
 
-dataset_update_parser = (
-    reqparse.RequestParser()
-    .add_argument(
-        "name",
-        nullable=False,
-        help="type is required. Name must be between 1 to 40 characters.",
-        type=_validate_name,
-    )
-    .add_argument("description", location="json", store_missing=False, type=validate_description_length)
-    .add_argument(
-        "indexing_technique",
-        type=str,
-        location="json",
-        choices=Dataset.INDEXING_TECHNIQUE_LIST,
-        nullable=True,
-        help="Invalid indexing technique.",
-    )
-    .add_argument(
-        "permission",
-        type=str,
-        location="json",
-        choices=(DatasetPermissionEnum.ONLY_ME, DatasetPermissionEnum.ALL_TEAM, DatasetPermissionEnum.PARTIAL_TEAM),
-        help="Invalid permission.",
-    )
-    .add_argument("embedding_model", type=str, location="json", help="Invalid embedding model.")
-    .add_argument("embedding_model_provider", type=str, location="json", help="Invalid embedding model provider.")
-    .add_argument("retrieval_model", type=dict, location="json", help="Invalid retrieval model.")
-    .add_argument("partial_member_list", type=list, location="json", help="Invalid parent user list.")
-    .add_argument(
-        "external_retrieval_model",
-        type=dict,
-        required=False,
-        nullable=True,
-        location="json",
-        help="Invalid external retrieval model.",
-    )
-    .add_argument(
-        "external_knowledge_id",
-        type=str,
-        required=False,
-        nullable=True,
-        location="json",
-        help="Invalid external knowledge id.",
-    )
-    .add_argument(
-        "external_knowledge_api_id",
-        type=str,
-        required=False,
-        nullable=True,
-        location="json",
-        help="Invalid external knowledge api id.",
-    )
-)
 
-tag_create_parser = reqparse.RequestParser().add_argument(
-    "name",
-    nullable=False,
-    required=True,
-    help="Name must be between 1 to 50 characters.",
-    type=lambda x: x
-    if x and 1 <= len(x) <= 50
-    else (_ for _ in ()).throw(ValueError("Name must be between 1 to 50 characters.")),
-)
+class TagNamePayload(BaseModel):
+    name: str = Field(..., min_length=1, max_length=50)
 
-tag_update_parser = (
-    reqparse.RequestParser()
-    .add_argument(
-        "name",
-        nullable=False,
-        required=True,
-        help="Name must be between 1 to 50 characters.",
-        type=lambda x: x
-        if x and 1 <= len(x) <= 50
-        else (_ for _ in ()).throw(ValueError("Name must be between 1 to 50 characters.")),
-    )
-    .add_argument("tag_id", nullable=False, required=True, help="Id of a tag.", type=str)
-)
 
-tag_delete_parser = reqparse.RequestParser().add_argument(
-    "tag_id", nullable=False, required=True, help="Id of a tag.", type=str
-)
+class TagCreatePayload(TagNamePayload):
+    pass
 
-tag_binding_parser = (
-    reqparse.RequestParser()
-    .add_argument("tag_ids", type=list, nullable=False, required=True, location="json", help="Tag IDs is required.")
-    .add_argument(
-        "target_id", type=str, nullable=False, required=True, location="json", help="Target Dataset ID is required."
-    )
-)
 
-tag_unbinding_parser = (
-    reqparse.RequestParser()
-    .add_argument("tag_id", type=str, nullable=False, required=True, help="Tag ID is required.")
-    .add_argument("target_id", type=str, nullable=False, required=True, help="Target ID is required.")
+class TagUpdatePayload(TagNamePayload):
+    tag_id: str
+
+
+class TagDeletePayload(BaseModel):
+    tag_id: str
+
+
+class TagBindingPayload(BaseModel):
+    tag_ids: list[str]
+    target_id: str
+
+    @field_validator("tag_ids")
+    @classmethod
+    def validate_tag_ids(cls, value: list[str]) -> list[str]:
+        if not value:
+            raise ValueError("Tag IDs is required.")
+        return value
+
+
+class TagUnbindingPayload(BaseModel):
+    tag_id: str
+    target_id: str
+
+
+register_schema_models(
+    service_api_ns,
+    DatasetCreatePayload,
+    DatasetUpdatePayload,
+    TagCreatePayload,
+    TagUpdatePayload,
+    TagDeletePayload,
+    TagBindingPayload,
+    TagUnbindingPayload,
 )
 
 
@@ -239,7 +151,7 @@ class DatasetListApi(DatasetApiResource):
         response = {"data": data, "has_more": len(datasets) == limit, "limit": limit, "total": total, "page": page}
         return response, 200
 
-    @service_api_ns.expect(dataset_create_parser)
+    @service_api_ns.expect(service_api_ns.models[DatasetCreatePayload.__name__])
     @service_api_ns.doc("create_dataset")
     @service_api_ns.doc(description="Create a new dataset")
     @service_api_ns.doc(
@@ -252,42 +164,41 @@ class DatasetListApi(DatasetApiResource):
     @cloud_edition_billing_rate_limit_check("knowledge", "dataset")
     def post(self, tenant_id):
         """Resource for creating datasets."""
-        args = dataset_create_parser.parse_args()
+        payload = DatasetCreatePayload.model_validate(service_api_ns.payload or {})
 
-        embedding_model_provider = args.get("embedding_model_provider")
-        embedding_model = args.get("embedding_model")
+        embedding_model_provider = payload.embedding_model_provider
+        embedding_model = payload.embedding_model
         if embedding_model_provider and embedding_model:
             DatasetService.check_embedding_model_setting(tenant_id, embedding_model_provider, embedding_model)
 
-        retrieval_model = args.get("retrieval_model")
+        retrieval_model = payload.retrieval_model
         if (
             retrieval_model
-            and retrieval_model.get("reranking_model")
-            and retrieval_model.get("reranking_model").get("reranking_provider_name")
+            and retrieval_model.reranking_model
+            and retrieval_model.reranking_model.reranking_provider_name
+            and retrieval_model.reranking_model.reranking_model_name
         ):
             DatasetService.check_reranking_model_setting(
                 tenant_id,
-                retrieval_model.get("reranking_model").get("reranking_provider_name"),
-                retrieval_model.get("reranking_model").get("reranking_model_name"),
+                retrieval_model.reranking_model.reranking_provider_name,
+                retrieval_model.reranking_model.reranking_model_name,
             )
 
         try:
             assert isinstance(current_user, Account)
             dataset = DatasetService.create_empty_dataset(
                 tenant_id=tenant_id,
-                name=args["name"],
-                description=args["description"],
-                indexing_technique=args["indexing_technique"],
+                name=payload.name,
+                description=payload.description,
+                indexing_technique=payload.indexing_technique,
                 account=current_user,
-                permission=args["permission"],
-                provider=args["provider"],
-                external_knowledge_api_id=args["external_knowledge_api_id"],
-                external_knowledge_id=args["external_knowledge_id"],
-                embedding_model_provider=args["embedding_model_provider"],
-                embedding_model_name=args["embedding_model"],
-                retrieval_model=RetrievalModel.model_validate(args["retrieval_model"])
-                if args["retrieval_model"] is not None
-                else None,
+                permission=str(payload.permission) if payload.permission else None,
+                provider=payload.provider,
+                external_knowledge_api_id=payload.external_knowledge_api_id,
+                external_knowledge_id=payload.external_knowledge_id,
+                embedding_model_provider=payload.embedding_model_provider,
+                embedding_model_name=payload.embedding_model,
+                retrieval_model=payload.retrieval_model,
             )
         except services.errors.dataset.DatasetNameDuplicateError:
             raise DatasetNameDuplicateError()
@@ -353,7 +264,7 @@ class DatasetApi(DatasetApiResource):
 
         return data, 200
 
-    @service_api_ns.expect(dataset_update_parser)
+    @service_api_ns.expect(service_api_ns.models[DatasetUpdatePayload.__name__])
     @service_api_ns.doc("update_dataset")
     @service_api_ns.doc(description="Update an existing dataset")
     @service_api_ns.doc(params={"dataset_id": "Dataset ID"})
@@ -372,36 +283,45 @@ class DatasetApi(DatasetApiResource):
         if dataset is None:
             raise NotFound("Dataset not found.")
 
-        args = dataset_update_parser.parse_args()
-        data = request.get_json()
+        payload_dict = service_api_ns.payload or {}
+        payload = DatasetUpdatePayload.model_validate(payload_dict)
+        update_data = payload.model_dump(exclude_unset=True)
+        if payload.permission is not None:
+            update_data["permission"] = str(payload.permission)
+        if payload.retrieval_model is not None:
+            update_data["retrieval_model"] = payload.retrieval_model.model_dump()
 
         # check embedding model setting
-        embedding_model_provider = data.get("embedding_model_provider")
-        embedding_model = data.get("embedding_model")
-        if data.get("indexing_technique") == "high_quality" or embedding_model_provider:
+        embedding_model_provider = payload.embedding_model_provider
+        embedding_model = payload.embedding_model
+        if payload.indexing_technique == "high_quality" or embedding_model_provider:
             if embedding_model_provider and embedding_model:
                 DatasetService.check_embedding_model_setting(
                     dataset.tenant_id, embedding_model_provider, embedding_model
                 )
 
-        retrieval_model = data.get("retrieval_model")
+        retrieval_model = payload.retrieval_model
         if (
             retrieval_model
-            and retrieval_model.get("reranking_model")
-            and retrieval_model.get("reranking_model").get("reranking_provider_name")
+            and retrieval_model.reranking_model
+            and retrieval_model.reranking_model.reranking_provider_name
+            and retrieval_model.reranking_model.reranking_model_name
         ):
             DatasetService.check_reranking_model_setting(
                 dataset.tenant_id,
-                retrieval_model.get("reranking_model").get("reranking_provider_name"),
-                retrieval_model.get("reranking_model").get("reranking_model_name"),
+                retrieval_model.reranking_model.reranking_provider_name,
+                retrieval_model.reranking_model.reranking_model_name,
             )
 
         # The role of the current user in the ta table must be admin, owner, editor, or dataset_operator
         DatasetPermissionService.check_permission(
-            current_user, dataset, data.get("permission"), data.get("partial_member_list")
+            current_user,
+            dataset,
+            str(payload.permission) if payload.permission else None,
+            payload.partial_member_list,
         )
 
-        dataset = DatasetService.update_dataset(dataset_id_str, args, current_user)
+        dataset = DatasetService.update_dataset(dataset_id_str, update_data, current_user)
 
         if dataset is None:
             raise NotFound("Dataset not found.")
@@ -410,15 +330,10 @@ class DatasetApi(DatasetApiResource):
         assert isinstance(current_user, Account)
         tenant_id = current_user.current_tenant_id
 
-        if data.get("partial_member_list") and data.get("permission") == "partial_members":
-            DatasetPermissionService.update_partial_member_list(
-                tenant_id, dataset_id_str, data.get("partial_member_list")
-            )
+        if payload.partial_member_list and payload.permission == DatasetPermissionEnum.PARTIAL_TEAM:
+            DatasetPermissionService.update_partial_member_list(tenant_id, dataset_id_str, payload.partial_member_list)
         # clear partial member list when permission is only_me or all_team_members
-        elif (
-            data.get("permission") == DatasetPermissionEnum.ONLY_ME
-            or data.get("permission") == DatasetPermissionEnum.ALL_TEAM
-        ):
+        elif payload.permission in {DatasetPermissionEnum.ONLY_ME, DatasetPermissionEnum.ALL_TEAM}:
             DatasetPermissionService.clear_partial_member_list(dataset_id_str)
 
         partial_member_list = DatasetPermissionService.get_dataset_partial_member_list(dataset_id_str)
@@ -556,7 +471,7 @@ class DatasetTagsApi(DatasetApiResource):
 
         return tags, 200
 
-    @service_api_ns.expect(tag_create_parser)
+    @service_api_ns.expect(service_api_ns.models[TagCreatePayload.__name__])
     @service_api_ns.doc("create_dataset_tag")
     @service_api_ns.doc(description="Add a knowledge type tag")
     @service_api_ns.doc(
@@ -574,14 +489,13 @@ class DatasetTagsApi(DatasetApiResource):
         if not (current_user.has_edit_permission or current_user.is_dataset_editor):
             raise Forbidden()
 
-        args = tag_create_parser.parse_args()
-        args["type"] = "knowledge"
-        tag = TagService.save_tags(args)
+        payload = TagCreatePayload.model_validate(service_api_ns.payload or {})
+        tag = TagService.save_tags({"name": payload.name, "type": "knowledge"})
 
         response = {"id": tag.id, "name": tag.name, "type": tag.type, "binding_count": 0}
         return response, 200
 
-    @service_api_ns.expect(tag_update_parser)
+    @service_api_ns.expect(service_api_ns.models[TagUpdatePayload.__name__])
     @service_api_ns.doc("update_dataset_tag")
     @service_api_ns.doc(description="Update a knowledge type tag")
     @service_api_ns.doc(
@@ -598,10 +512,10 @@ class DatasetTagsApi(DatasetApiResource):
         if not (current_user.has_edit_permission or current_user.is_dataset_editor):
             raise Forbidden()
 
-        args = tag_update_parser.parse_args()
-        args["type"] = "knowledge"
-        tag_id = args["tag_id"]
-        tag = TagService.update_tags(args, tag_id)
+        payload = TagUpdatePayload.model_validate(service_api_ns.payload or {})
+        params = {"name": payload.name, "type": "knowledge"}
+        tag_id = payload.tag_id
+        tag = TagService.update_tags(params, tag_id)
 
         binding_count = TagService.get_tag_binding_count(tag_id)
 
@@ -609,7 +523,7 @@ class DatasetTagsApi(DatasetApiResource):
 
         return response, 200
 
-    @service_api_ns.expect(tag_delete_parser)
+    @service_api_ns.expect(service_api_ns.models[TagDeletePayload.__name__])
     @service_api_ns.doc("delete_dataset_tag")
     @service_api_ns.doc(description="Delete a knowledge type tag")
     @service_api_ns.doc(
@@ -623,15 +537,15 @@ class DatasetTagsApi(DatasetApiResource):
     @edit_permission_required
     def delete(self, _, dataset_id):
         """Delete a knowledge type tag."""
-        args = tag_delete_parser.parse_args()
-        TagService.delete_tag(args["tag_id"])
+        payload = TagDeletePayload.model_validate(service_api_ns.payload or {})
+        TagService.delete_tag(payload.tag_id)
 
         return 204
 
 
 @service_api_ns.route("/datasets/tags/binding")
 class DatasetTagBindingApi(DatasetApiResource):
-    @service_api_ns.expect(tag_binding_parser)
+    @service_api_ns.expect(service_api_ns.models[TagBindingPayload.__name__])
     @service_api_ns.doc("bind_dataset_tags")
     @service_api_ns.doc(description="Bind tags to a dataset")
     @service_api_ns.doc(
@@ -648,16 +562,15 @@ class DatasetTagBindingApi(DatasetApiResource):
         if not (current_user.has_edit_permission or current_user.is_dataset_editor):
             raise Forbidden()
 
-        args = tag_binding_parser.parse_args()
-        args["type"] = "knowledge"
-        TagService.save_tag_binding(args)
+        payload = TagBindingPayload.model_validate(service_api_ns.payload or {})
+        TagService.save_tag_binding({"tag_ids": payload.tag_ids, "target_id": payload.target_id, "type": "knowledge"})
 
         return 204
 
 
 @service_api_ns.route("/datasets/tags/unbinding")
 class DatasetTagUnbindingApi(DatasetApiResource):
-    @service_api_ns.expect(tag_unbinding_parser)
+    @service_api_ns.expect(service_api_ns.models[TagUnbindingPayload.__name__])
     @service_api_ns.doc("unbind_dataset_tag")
     @service_api_ns.doc(description="Unbind a tag from a dataset")
     @service_api_ns.doc(
@@ -674,9 +587,8 @@ class DatasetTagUnbindingApi(DatasetApiResource):
         if not (current_user.has_edit_permission or current_user.is_dataset_editor):
             raise Forbidden()
 
-        args = tag_unbinding_parser.parse_args()
-        args["type"] = "knowledge"
-        TagService.delete_tag_binding(args)
+        payload = TagUnbindingPayload.model_validate(service_api_ns.payload or {})
+        TagService.delete_tag_binding({"tag_id": payload.tag_id, "target_id": payload.target_id, "type": "knowledge"})
 
         return 204
 

api/controllers/service_api/dataset/document.py

@@ -3,8 +3,8 @@ from typing import Self
 from uuid import UUID
 
 from flask import request
-from flask_restx import marshal, reqparse
-from pydantic import BaseModel, model_validator
+from flask_restx import marshal
+from pydantic import BaseModel, Field, model_validator
 from sqlalchemy import desc, select
 from werkzeug.exceptions import Forbidden, NotFound
 
@@ -37,22 +37,19 @@ from services.dataset_service import DatasetService, DocumentService
 from services.entities.knowledge_entities.knowledge_entities import KnowledgeConfig, ProcessRule, RetrievalModel
 from services.file_service import FileService
 
-# Define parsers for document operations
-document_text_create_parser = (
-    reqparse.RequestParser()
-    .add_argument("name", type=str, required=True, nullable=False, location="json")
-    .add_argument("text", type=str, required=True, nullable=False, location="json")
-    .add_argument("process_rule", type=dict, required=False, nullable=True, location="json")
-    .add_argument("original_document_id", type=str, required=False, location="json")
-    .add_argument("doc_form", type=str, default="text_model", required=False, nullable=False, location="json")
-    .add_argument("doc_language", type=str, default="English", required=False, nullable=False, location="json")
-    .add_argument(
-        "indexing_technique", type=str, choices=Dataset.INDEXING_TECHNIQUE_LIST, nullable=False, location="json"
-    )
-    .add_argument("retrieval_model", type=dict, required=False, nullable=True, location="json")
-    .add_argument("embedding_model", type=str, required=False, nullable=True, location="json")
-    .add_argument("embedding_model_provider", type=str, required=False, nullable=True, location="json")
-)
+
+class DocumentTextCreatePayload(BaseModel):
+    name: str
+    text: str
+    process_rule: ProcessRule | None = None
+    original_document_id: str | None = None
+    doc_form: str = Field(default="text_model")
+    doc_language: str = Field(default="English")
+    indexing_technique: str | None = None
+    retrieval_model: RetrievalModel | None = None
+    embedding_model: str | None = None
+    embedding_model_provider: str | None = None
+
 
 DEFAULT_REF_TEMPLATE_SWAGGER_2_0 = "#/definitions/{model}"
 
@@ -72,7 +69,7 @@ class DocumentTextUpdate(BaseModel):
         return self
 
 
-for m in [ProcessRule, RetrievalModel, DocumentTextUpdate]:
+for m in [ProcessRule, RetrievalModel, DocumentTextCreatePayload, DocumentTextUpdate]:
     service_api_ns.schema_model(m.__name__, m.model_json_schema(ref_template=DEFAULT_REF_TEMPLATE_SWAGGER_2_0))  # type: ignore
 
 
@@ -83,7 +80,7 @@ for m in [ProcessRule, RetrievalModel, DocumentTextUpdate]:
 class DocumentAddByTextApi(DatasetApiResource):
     """Resource for documents."""
 
-    @service_api_ns.expect(document_text_create_parser)
+    @service_api_ns.expect(service_api_ns.models[DocumentTextCreatePayload.__name__])
     @service_api_ns.doc("create_document_by_text")
     @service_api_ns.doc(description="Create a new document by providing text content")
     @service_api_ns.doc(params={"dataset_id": "Dataset ID"})
@@ -99,7 +96,8 @@ class DocumentAddByTextApi(DatasetApiResource):
     @cloud_edition_billing_rate_limit_check("knowledge", "dataset")
     def post(self, tenant_id, dataset_id):
         """Create document by text."""
-        args = document_text_create_parser.parse_args()
+        payload = DocumentTextCreatePayload.model_validate(service_api_ns.payload or {})
+        args = payload.model_dump(exclude_none=True)
 
         dataset_id = str(dataset_id)
         tenant_id = str(tenant_id)
@@ -111,33 +109,29 @@ class DocumentAddByTextApi(DatasetApiResource):
         if not dataset.indexing_technique and not args["indexing_technique"]:
             raise ValueError("indexing_technique is required.")
 
-        text = args.get("text")
-        name = args.get("name")
-        if text is None or name is None:
-            raise ValueError("Both 'text' and 'name' must be non-null values.")
-
-        embedding_model_provider = args.get("embedding_model_provider")
-        embedding_model = args.get("embedding_model")
+        embedding_model_provider = payload.embedding_model_provider
+        embedding_model = payload.embedding_model
         if embedding_model_provider and embedding_model:
             DatasetService.check_embedding_model_setting(tenant_id, embedding_model_provider, embedding_model)
 
-        retrieval_model = args.get("retrieval_model")
+        retrieval_model = payload.retrieval_model
         if (
             retrieval_model
-            and retrieval_model.get("reranking_model")
-            and retrieval_model.get("reranking_model").get("reranking_provider_name")
+            and retrieval_model.reranking_model
+            and retrieval_model.reranking_model.reranking_provider_name
+            and retrieval_model.reranking_model.reranking_model_name
         ):
             DatasetService.check_reranking_model_setting(
                 tenant_id,
-                retrieval_model.get("reranking_model").get("reranking_provider_name"),
-                retrieval_model.get("reranking_model").get("reranking_model_name"),
+                retrieval_model.reranking_model.reranking_provider_name,
+                retrieval_model.reranking_model.reranking_model_name,
             )
 
         if not current_user:
             raise ValueError("current_user is required")
 
         upload_file = FileService(db.engine).upload_text(
-            text=str(text), text_name=str(name), user_id=current_user.id, tenant_id=tenant_id
+            text=payload.text, text_name=payload.name, user_id=current_user.id, tenant_id=tenant_id
         )
         data_source = {
             "type": "upload_file",
@@ -174,7 +168,7 @@ class DocumentAddByTextApi(DatasetApiResource):
 class DocumentUpdateByTextApi(DatasetApiResource):
     """Resource for update documents."""
 
-    @service_api_ns.expect(service_api_ns.models[DocumentTextUpdate.__name__], validate=True)
+    @service_api_ns.expect(service_api_ns.models[DocumentTextUpdate.__name__])
     @service_api_ns.doc("update_document_by_text")
     @service_api_ns.doc(description="Update an existing document by providing text content")
     @service_api_ns.doc(params={"dataset_id": "Dataset ID", "document_id": "Document ID"})
@@ -189,22 +183,23 @@ class DocumentUpdateByTextApi(DatasetApiResource):
     @cloud_edition_billing_rate_limit_check("knowledge", "dataset")
     def post(self, tenant_id: str, dataset_id: UUID, document_id: UUID):
         """Update document by text."""
-        args = DocumentTextUpdate.model_validate(service_api_ns.payload).model_dump(exclude_unset=True)
+        payload = DocumentTextUpdate.model_validate(service_api_ns.payload or {})
         dataset = db.session.query(Dataset).where(Dataset.tenant_id == tenant_id, Dataset.id == str(dataset_id)).first()
-
+        args = payload.model_dump(exclude_none=True)
         if not dataset:
             raise ValueError("Dataset does not exist.")
 
-        retrieval_model = args.get("retrieval_model")
+        retrieval_model = payload.retrieval_model
         if (
             retrieval_model
-            and retrieval_model.get("reranking_model")
-            and retrieval_model.get("reranking_model").get("reranking_provider_name")
+            and retrieval_model.reranking_model
+            and retrieval_model.reranking_model.reranking_provider_name
+            and retrieval_model.reranking_model.reranking_model_name
         ):
             DatasetService.check_reranking_model_setting(
                 tenant_id,
-                retrieval_model.get("reranking_model").get("reranking_provider_name"),
-                retrieval_model.get("reranking_model").get("reranking_model_name"),
+                retrieval_model.reranking_model.reranking_provider_name,
+                retrieval_model.reranking_model.reranking_model_name,
             )
 
         # indexing_technique is already set in dataset since this is an update

api/controllers/service_api/dataset/metadata.py

@@ -1,9 +1,11 @@
 from typing import Literal
 
 from flask_login import current_user
-from flask_restx import marshal, reqparse
+from flask_restx import marshal
+from pydantic import BaseModel
 from werkzeug.exceptions import NotFound
 
+from controllers.common.schema import register_schema_model, register_schema_models
 from controllers.service_api import service_api_ns
 from controllers.service_api.wraps import DatasetApiResource, cloud_edition_billing_rate_limit_check
 from fields.dataset_fields import dataset_metadata_fields
@@ -14,25 +16,18 @@ from services.entities.knowledge_entities.knowledge_entities import (
 )
 from services.metadata_service import MetadataService
 
-# Define parsers for metadata APIs
-metadata_create_parser = (
-    reqparse.RequestParser()
-    .add_argument("type", type=str, required=True, nullable=False, location="json", help="Metadata type")
-    .add_argument("name", type=str, required=True, nullable=False, location="json", help="Metadata name")
-)
 
-metadata_update_parser = reqparse.RequestParser().add_argument(
-    "name", type=str, required=True, nullable=False, location="json", help="New metadata name"
-)
+class MetadataUpdatePayload(BaseModel):
+    name: str
 
-document_metadata_parser = reqparse.RequestParser().add_argument(
-    "operation_data", type=list, required=True, nullable=False, location="json", help="Metadata operation data"
-)
+
+register_schema_model(service_api_ns, MetadataUpdatePayload)
+register_schema_models(service_api_ns, MetadataArgs, MetadataOperationData)
 
 
 @service_api_ns.route("/datasets/<uuid:dataset_id>/metadata")
 class DatasetMetadataCreateServiceApi(DatasetApiResource):
-    @service_api_ns.expect(metadata_create_parser)
+    @service_api_ns.expect(service_api_ns.models[MetadataArgs.__name__])
     @service_api_ns.doc("create_dataset_metadata")
     @service_api_ns.doc(description="Create metadata for a dataset")
     @service_api_ns.doc(params={"dataset_id": "Dataset ID"})
@@ -46,8 +41,7 @@ class DatasetMetadataCreateServiceApi(DatasetApiResource):
     @cloud_edition_billing_rate_limit_check("knowledge", "dataset")
     def post(self, tenant_id, dataset_id):
         """Create metadata for a dataset."""
-        args = metadata_create_parser.parse_args()
-        metadata_args = MetadataArgs.model_validate(args)
+        metadata_args = MetadataArgs.model_validate(service_api_ns.payload or {})
 
         dataset_id_str = str(dataset_id)
         dataset = DatasetService.get_dataset(dataset_id_str)
@@ -79,7 +73,7 @@ class DatasetMetadataCreateServiceApi(DatasetApiResource):
 
 @service_api_ns.route("/datasets/<uuid:dataset_id>/metadata/<uuid:metadata_id>")
 class DatasetMetadataServiceApi(DatasetApiResource):
-    @service_api_ns.expect(metadata_update_parser)
+    @service_api_ns.expect(service_api_ns.models[MetadataUpdatePayload.__name__])
     @service_api_ns.doc("update_dataset_metadata")
     @service_api_ns.doc(description="Update metadata name")
     @service_api_ns.doc(params={"dataset_id": "Dataset ID", "metadata_id": "Metadata ID"})
@@ -93,7 +87,7 @@ class DatasetMetadataServiceApi(DatasetApiResource):
     @cloud_edition_billing_rate_limit_check("knowledge", "dataset")
     def patch(self, tenant_id, dataset_id, metadata_id):
         """Update metadata name."""
-        args = metadata_update_parser.parse_args()
+        payload = MetadataUpdatePayload.model_validate(service_api_ns.payload or {})
 
         dataset_id_str = str(dataset_id)
         metadata_id_str = str(metadata_id)
@@ -102,7 +96,7 @@ class DatasetMetadataServiceApi(DatasetApiResource):
             raise NotFound("Dataset not found.")
         DatasetService.check_dataset_permission(dataset, current_user)
 
-        metadata = MetadataService.update_metadata_name(dataset_id_str, metadata_id_str, args["name"])
+        metadata = MetadataService.update_metadata_name(dataset_id_str, metadata_id_str, payload.name)
         return marshal(metadata, dataset_metadata_fields), 200
 
     @service_api_ns.doc("delete_dataset_metadata")
@@ -175,7 +169,7 @@ class DatasetMetadataBuiltInFieldActionServiceApi(DatasetApiResource):
 
 @service_api_ns.route("/datasets/<uuid:dataset_id>/documents/metadata")
 class DocumentMetadataEditServiceApi(DatasetApiResource):
-    @service_api_ns.expect(document_metadata_parser)
+    @service_api_ns.expect(service_api_ns.models[MetadataOperationData.__name__])
     @service_api_ns.doc("update_documents_metadata")
     @service_api_ns.doc(description="Update metadata for multiple documents")
     @service_api_ns.doc(params={"dataset_id": "Dataset ID"})
@@ -195,8 +189,7 @@ class DocumentMetadataEditServiceApi(DatasetApiResource):
             raise NotFound("Dataset not found.")
         DatasetService.check_dataset_permission(dataset, current_user)
 
-        args = document_metadata_parser.parse_args()
-        metadata_args = MetadataOperationData.model_validate(args)
+        metadata_args = MetadataOperationData.model_validate(service_api_ns.payload or {})
 
         MetadataService.update_documents_metadata(dataset, metadata_args)
 

api/controllers/service_api/dataset/rag_pipeline/rag_pipeline_workflow.py

@@ -4,12 +4,12 @@ from collections.abc import Generator
 from typing import Any
 
 from flask import request
-from flask_restx import reqparse
-from flask_restx.reqparse import ParseResult, RequestParser
+from pydantic import BaseModel
 from werkzeug.exceptions import Forbidden
 
 import services
 from controllers.common.errors import FilenameNotExistsError, NoFileUploadedError, TooManyFilesError
+from controllers.common.schema import register_schema_model
 from controllers.service_api import service_api_ns
 from controllers.service_api.dataset.error import PipelineRunError
 from controllers.service_api.wraps import DatasetApiResource
@@ -22,11 +22,25 @@ from models.dataset import Pipeline
 from models.engine import db
 from services.errors.file import FileTooLargeError, UnsupportedFileTypeError
 from services.file_service import FileService
-from services.rag_pipeline.entity.pipeline_service_api_entities import DatasourceNodeRunApiEntity
+from services.rag_pipeline.entity.pipeline_service_api_entities import (
+    DatasourceNodeRunApiEntity,
+    PipelineRunApiEntity,
+)
 from services.rag_pipeline.pipeline_generate_service import PipelineGenerateService
 from services.rag_pipeline.rag_pipeline import RagPipelineService
 
 
+class DatasourceNodeRunPayload(BaseModel):
+    inputs: dict[str, Any]
+    datasource_type: str
+    credential_id: str | None = None
+    is_published: bool
+
+
+register_schema_model(service_api_ns, DatasourceNodeRunPayload)
+register_schema_model(service_api_ns, PipelineRunApiEntity)
+
+
 @service_api_ns.route(f"/datasets/{uuid:dataset_id}/pipeline/datasource-plugins")
 class DatasourcePluginsApi(DatasetApiResource):
     """Resource for datasource plugins."""
@@ -88,22 +102,20 @@ class DatasourceNodeRunApi(DatasetApiResource):
             401: "Unauthorized - invalid API token",
         }
     )
+    @service_api_ns.expect(service_api_ns.models[DatasourceNodeRunPayload.__name__])
     def post(self, tenant_id: str, dataset_id: str, node_id: str):
         """Resource for getting datasource plugins."""
-        # Get query parameter to determine published or draft
-        parser: RequestParser = (
-            reqparse.RequestParser()
-            .add_argument("inputs", type=dict, required=True, nullable=False, location="json")
-            .add_argument("datasource_type", type=str, required=True, location="json")
-            .add_argument("credential_id", type=str, required=False, location="json")
-            .add_argument("is_published", type=bool, required=True, location="json")
-        )
-        args: ParseResult = parser.parse_args()
-
-        datasource_node_run_api_entity = DatasourceNodeRunApiEntity.model_validate(args)
+        payload = DatasourceNodeRunPayload.model_validate(service_api_ns.payload or {})
         assert isinstance(current_user, Account)
         rag_pipeline_service: RagPipelineService = RagPipelineService()
         pipeline: Pipeline = rag_pipeline_service.get_pipeline(tenant_id=tenant_id, dataset_id=dataset_id)
+        datasource_node_run_api_entity = DatasourceNodeRunApiEntity.model_validate(
+            {
+                **payload.model_dump(exclude_none=True),
+                "pipeline_id": str(pipeline.id),
+                "node_id": node_id,
+            }
+        )
         return helper.compact_generate_response(
             PipelineGenerator.convert_to_event_stream(
                 rag_pipeline_service.run_datasource_workflow_node(
@@ -147,25 +159,10 @@ class PipelineRunApi(DatasetApiResource):
             401: "Unauthorized - invalid API token",
         }
     )
+    @service_api_ns.expect(service_api_ns.models[PipelineRunApiEntity.__name__])
     def post(self, tenant_id: str, dataset_id: str):
         """Resource for running a rag pipeline."""
-        parser: RequestParser = (
-            reqparse.RequestParser()
-            .add_argument("inputs", type=dict, required=True, nullable=False, location="json")
-            .add_argument("datasource_type", type=str, required=True, location="json")
-            .add_argument("datasource_info_list", type=list, required=True, location="json")
-            .add_argument("start_node_id", type=str, required=True, location="json")
-            .add_argument("is_published", type=bool, required=True, default=True, location="json")
-            .add_argument(
-                "response_mode",
-                type=str,
-                required=True,
-                choices=["streaming", "blocking"],
-                default="blocking",
-                location="json",
-            )
-        )
-        args: ParseResult = parser.parse_args()
+        payload = PipelineRunApiEntity.model_validate(service_api_ns.payload or {})
 
         if not isinstance(current_user, Account):
             raise Forbidden()
@@ -176,9 +173,9 @@ class PipelineRunApi(DatasetApiResource):
             response: dict[Any, Any] | Generator[str, Any, None] = PipelineGenerateService.generate(
                 pipeline=pipeline,
                 user=current_user,
-                args=args,
-                invoke_from=InvokeFrom.PUBLISHED if args.get("is_published") else InvokeFrom.DEBUGGER,
-                streaming=args.get("response_mode") == "streaming",
+                args=payload.model_dump(),
+                invoke_from=InvokeFrom.PUBLISHED if payload.is_published else InvokeFrom.DEBUGGER,
+                streaming=payload.response_mode == "streaming",
             )
 
             return helper.compact_generate_response(response)

api/controllers/service_api/dataset/segment.py

@@ -1,8 +1,12 @@
+from typing import Any
+
 from flask import request
-from flask_restx import marshal, reqparse
+from flask_restx import marshal
+from pydantic import BaseModel, Field
 from werkzeug.exceptions import NotFound
 
 from configs import dify_config
+from controllers.common.schema import register_schema_models
 from controllers.service_api import service_api_ns
 from controllers.service_api.app.error import ProviderNotInitializeError
 from controllers.service_api.wraps import (
@@ -24,34 +28,42 @@ from services.errors.chunk import ChildChunkDeleteIndexError, ChildChunkIndexing
 from services.errors.chunk import ChildChunkDeleteIndexError as ChildChunkDeleteIndexServiceError
 from services.errors.chunk import ChildChunkIndexingError as ChildChunkIndexingServiceError
 
-# Define parsers for segment operations
-segment_create_parser = reqparse.RequestParser().add_argument(
-    "segments", type=list, required=False, nullable=True, location="json"
-)
 
-segment_list_parser = (
-    reqparse.RequestParser()
-    .add_argument("status", type=str, action="append", default=[], location="args")
-    .add_argument("keyword", type=str, default=None, location="args")
-)
+class SegmentCreatePayload(BaseModel):
+    segments: list[dict[str, Any]] | None = None
 
-segment_update_parser = reqparse.RequestParser().add_argument(
-    "segment", type=dict, required=False, nullable=True, location="json"
-)
 
-child_chunk_create_parser = reqparse.RequestParser().add_argument(
-    "content", type=str, required=True, nullable=False, location="json"
-)
+class SegmentListQuery(BaseModel):
+    status: list[str] = Field(default_factory=list)
+    keyword: str | None = None
 
-child_chunk_list_parser = (
-    reqparse.RequestParser()
-    .add_argument("limit", type=int, default=20, location="args")
-    .add_argument("keyword", type=str, default=None, location="args")
-    .add_argument("page", type=int, default=1, location="args")
-)
 
-child_chunk_update_parser = reqparse.RequestParser().add_argument(
-    "content", type=str, required=True, nullable=False, location="json"
+class SegmentUpdatePayload(BaseModel):
+    segment: SegmentUpdateArgs
+
+
+class ChildChunkCreatePayload(BaseModel):
+    content: str
+
+
+class ChildChunkListQuery(BaseModel):
+    limit: int = Field(default=20, ge=1)
+    keyword: str | None = None
+    page: int = Field(default=1, ge=1)
+
+
+class ChildChunkUpdatePayload(BaseModel):
+    content: str
+
+
+register_schema_models(
+    service_api_ns,
+    SegmentCreatePayload,
+    SegmentListQuery,
+    SegmentUpdatePayload,
+    ChildChunkCreatePayload,
+    ChildChunkListQuery,
+    ChildChunkUpdatePayload,
 )
 
 
@@ -59,7 +71,7 @@ child_chunk_update_parser = reqparse.RequestParser().add_argument(
 class SegmentApi(DatasetApiResource):
     """Resource for segments."""
 
-    @service_api_ns.expect(segment_create_parser)
+    @service_api_ns.expect(service_api_ns.models[SegmentCreatePayload.__name__])
     @service_api_ns.doc("create_segments")
     @service_api_ns.doc(description="Create segments in a document")
     @service_api_ns.doc(params={"dataset_id": "Dataset ID", "document_id": "Document ID"})
@@ -106,20 +118,20 @@ class SegmentApi(DatasetApiResource):
             except ProviderTokenNotInitError as ex:
                 raise ProviderNotInitializeError(ex.description)
         # validate args
-        args = segment_create_parser.parse_args()
-        if args["segments"] is not None:
+        payload = SegmentCreatePayload.model_validate(service_api_ns.payload or {})
+        if payload.segments is not None:
             segments_limit = dify_config.DATASET_MAX_SEGMENTS_PER_REQUEST
-            if segments_limit > 0 and len(args["segments"]) > segments_limit:
+            if segments_limit > 0 and len(payload.segments) > segments_limit:
                 raise ValueError(f"Exceeded maximum segments limit of {segments_limit}.")
 
-            for args_item in args["segments"]:
+            for args_item in payload.segments:
                 SegmentService.segment_create_args_validate(args_item, document)
-            segments = SegmentService.multi_create_segment(args["segments"], document, dataset)
+            segments = SegmentService.multi_create_segment(payload.segments, document, dataset)
             return {"data": marshal(segments, segment_fields), "doc_form": document.doc_form}, 200
         else:
             return {"error": "Segments is required"}, 400
 
-    @service_api_ns.expect(segment_list_parser)
+    @service_api_ns.expect(service_api_ns.models[SegmentListQuery.__name__])
     @service_api_ns.doc("list_segments")
     @service_api_ns.doc(description="List segments in a document")
     @service_api_ns.doc(params={"dataset_id": "Dataset ID", "document_id": "Document ID"})
@@ -160,13 +172,18 @@ class SegmentApi(DatasetApiResource):
             except ProviderTokenNotInitError as ex:
                 raise ProviderNotInitializeError(ex.description)
 
-        args = segment_list_parser.parse_args()
+        args = SegmentListQuery.model_validate(
+            {
+                "status": request.args.getlist("status"),
+                "keyword": request.args.get("keyword"),
+            }
+        )
 
         segments, total = SegmentService.get_segments(
             document_id=document_id,
             tenant_id=current_tenant_id,
-            status_list=args["status"],
-            keyword=args["keyword"],
+            status_list=args.status,
+            keyword=args.keyword,
             page=page,
             limit=limit,
         )
@@ -217,7 +234,7 @@ class DatasetSegmentApi(DatasetApiResource):
         SegmentService.delete_segment(segment, document, dataset)
         return 204
 
-    @service_api_ns.expect(segment_update_parser)
+    @service_api_ns.expect(service_api_ns.models[SegmentUpdatePayload.__name__])
     @service_api_ns.doc("update_segment")
     @service_api_ns.doc(description="Update a specific segment")
     @service_api_ns.doc(
@@ -265,12 +282,9 @@ class DatasetSegmentApi(DatasetApiResource):
         if not segment:
             raise NotFound("Segment not found.")
 
-        # validate args
-        args = segment_update_parser.parse_args()
+        payload = SegmentUpdatePayload.model_validate(service_api_ns.payload or {})
 
-        updated_segment = SegmentService.update_segment(
-            SegmentUpdateArgs.model_validate(args["segment"]), segment, document, dataset
-        )
+        updated_segment = SegmentService.update_segment(payload.segment, segment, document, dataset)
         return {"data": marshal(updated_segment, segment_fields), "doc_form": document.doc_form}, 200
 
     @service_api_ns.doc("get_segment")
@@ -308,7 +322,7 @@ class DatasetSegmentApi(DatasetApiResource):
 class ChildChunkApi(DatasetApiResource):
     """Resource for child chunks."""
 
-    @service_api_ns.expect(child_chunk_create_parser)
+    @service_api_ns.expect(service_api_ns.models[ChildChunkCreatePayload.__name__])
     @service_api_ns.doc("create_child_chunk")
     @service_api_ns.doc(description="Create a new child chunk for a segment")
     @service_api_ns.doc(
@@ -360,16 +374,16 @@ class ChildChunkApi(DatasetApiResource):
                 raise ProviderNotInitializeError(ex.description)
 
         # validate args
-        args = child_chunk_create_parser.parse_args()
+        payload = ChildChunkCreatePayload.model_validate(service_api_ns.payload or {})
 
         try:
-            child_chunk = SegmentService.create_child_chunk(args["content"], segment, document, dataset)
+            child_chunk = SegmentService.create_child_chunk(payload.content, segment, document, dataset)
         except ChildChunkIndexingServiceError as e:
             raise ChildChunkIndexingError(str(e))
 
         return {"data": marshal(child_chunk, child_chunk_fields)}, 200
 
-    @service_api_ns.expect(child_chunk_list_parser)
+    @service_api_ns.expect(service_api_ns.models[ChildChunkListQuery.__name__])
     @service_api_ns.doc("list_child_chunks")
     @service_api_ns.doc(description="List child chunks for a segment")
     @service_api_ns.doc(
@@ -400,11 +414,17 @@ class ChildChunkApi(DatasetApiResource):
         if not segment:
             raise NotFound("Segment not found.")
 
-        args = child_chunk_list_parser.parse_args()
+        args = ChildChunkListQuery.model_validate(
+            {
+                "limit": request.args.get("limit", default=20, type=int),
+                "keyword": request.args.get("keyword"),
+                "page": request.args.get("page", default=1, type=int),
+            }
+        )
 
-        page = args["page"]
-        limit = min(args["limit"], 100)
-        keyword = args["keyword"]
+        page = args.page
+        limit = min(args.limit, 100)
+        keyword = args.keyword
 
         child_chunks = SegmentService.get_child_chunks(segment_id, document_id, dataset_id, page, limit, keyword)
 
@@ -480,7 +500,7 @@ class DatasetChildChunkApi(DatasetApiResource):
 
         return 204
 
-    @service_api_ns.expect(child_chunk_update_parser)
+    @service_api_ns.expect(service_api_ns.models[ChildChunkUpdatePayload.__name__])
     @service_api_ns.doc("update_child_chunk")
     @service_api_ns.doc(description="Update a specific child chunk")
     @service_api_ns.doc(
@@ -533,10 +553,10 @@ class DatasetChildChunkApi(DatasetApiResource):
             raise NotFound("Child chunk not found.")
 
         # validate args
-        args = child_chunk_update_parser.parse_args()
+        payload = ChildChunkUpdatePayload.model_validate(service_api_ns.payload or {})
 
         try:
-            child_chunk = SegmentService.update_child_chunk(args["content"], child_chunk, segment, document, dataset)
+            child_chunk = SegmentService.update_child_chunk(payload.content, child_chunk, segment, document, dataset)
         except ChildChunkIndexingServiceError as e:
             raise ChildChunkIndexingError(str(e))
 

api/controllers/service_api/wraps.py

@@ -1,3 +1,4 @@
+import logging
 import time
 from collections.abc import Callable
 from datetime import timedelta
@@ -28,6 +29,8 @@ P = ParamSpec("P")
 R = TypeVar("R")
 T = TypeVar("T")
 
+logger = logging.getLogger(__name__)
+
 
 class WhereisUserArg(StrEnum):
     """
@@ -238,8 +241,8 @@ def validate_dataset_token(view: Callable[Concatenate[T, P], R] | None = None):
                         # Basic check: UUIDs are 36 chars with hyphens
                         if len(str_id) == 36 and str_id.count("-") == 4:
                             dataset_id = str_id
-                    except:
-                        pass
+                    except Exception:
+                        logger.exception("Failed to parse dataset_id from class method args")
                 elif len(args) > 0:
                     # Not a class method, check if args[0] looks like a UUID
                     potential_id = args[0]
@@ -247,8 +250,8 @@ def validate_dataset_token(view: Callable[Concatenate[T, P], R] | None = None):
                         str_id = str(potential_id)
                         if len(str_id) == 36 and str_id.count("-") == 4:
                             dataset_id = str_id
-                    except:
-                        pass
+                    except Exception:
+                        logger.exception("Failed to parse dataset_id from positional args")
 
             # Validate dataset if dataset_id is provided
             if dataset_id:

api/core/agent/cot_agent_runner.py

@@ -1,4 +1,5 @@
 import json
+import logging
 from abc import ABC, abstractmethod
 from collections.abc import Generator, Mapping, Sequence
 from typing import Any
@@ -23,6 +24,8 @@ from core.tools.entities.tool_entities import ToolInvokeMeta
 from core.tools.tool_engine import ToolEngine
 from models.model import Message
 
+logger = logging.getLogger(__name__)
+
 
 class CotAgentRunner(BaseAgentRunner, ABC):
     _is_first_iteration = True
@@ -400,8 +403,8 @@ class CotAgentRunner(BaseAgentRunner, ABC):
                             action_input=json.loads(message.tool_calls[0].function.arguments),
                         )
                         current_scratchpad.action_str = json.dumps(current_scratchpad.action.to_dict())
-                    except:
-                        pass
+                    except Exception:
+                        logger.exception("Failed to parse tool call from assistant message")
             elif isinstance(message, ToolPromptMessage):
                 if current_scratchpad:
                     assert isinstance(message.content, str)

api/core/app/apps/advanced_chat/app_runner.py

@@ -35,6 +35,7 @@ from core.workflow.variable_loader import VariableLoader
 from core.workflow.workflow_entry import WorkflowEntry
 from extensions.ext_database import db
 from extensions.ext_redis import redis_client
+from extensions.otel import WorkflowAppRunnerHandler, trace_span
 from models import Workflow
 from models.enums import UserFrom
 from models.model import App, Conversation, Message, MessageAnnotation
@@ -80,6 +81,7 @@ class AdvancedChatAppRunner(WorkflowBasedAppRunner):
         self._workflow_execution_repository = workflow_execution_repository
         self._workflow_node_execution_repository = workflow_node_execution_repository
 
+    @trace_span(WorkflowAppRunnerHandler)
     def run(self):
         app_config = self.application_generate_entity.app_config
         app_config = cast(AdvancedChatAppConfig, app_config)

api/core/app/apps/base_app_runner.py

@@ -83,6 +83,7 @@ class AppRunner:
         context: str | None = None,
         memory: TokenBufferMemory | None = None,
         image_detail_config: ImagePromptMessageContent.DETAIL | None = None,
+        context_files: list["File"] | None = None,
     ) -> tuple[list[PromptMessage], list[str] | None]:
         """
         Organize prompt messages
@@ -111,6 +112,7 @@ class AppRunner:
                 memory=memory,
                 model_config=model_config,
                 image_detail_config=image_detail_config,
+                context_files=context_files,
             )
         else:
             memory_config = MemoryConfig(window=MemoryConfig.WindowConfig(enabled=False))

api/core/app/apps/chat/app_runner.py

@@ -11,6 +11,7 @@ from core.app.entities.app_invoke_entities import (
 )
 from core.app.entities.queue_entities import QueueAnnotationReplyEvent
 from core.callback_handler.index_tool_callback_handler import DatasetIndexToolCallbackHandler
+from core.file import File
 from core.memory.token_buffer_memory import TokenBufferMemory
 from core.model_manager import ModelInstance
 from core.model_runtime.entities.message_entities import ImagePromptMessageContent
@@ -146,6 +147,7 @@ class ChatAppRunner(AppRunner):
 
         # get context from datasets
         context = None
+        context_files: list[File] = []
         if app_config.dataset and app_config.dataset.dataset_ids:
             hit_callback = DatasetIndexToolCallbackHandler(
                 queue_manager,
@@ -156,7 +158,7 @@ class ChatAppRunner(AppRunner):
             )
 
             dataset_retrieval = DatasetRetrieval(application_generate_entity)
-            context = dataset_retrieval.retrieve(
+            context, retrieved_files = dataset_retrieval.retrieve(
                 app_id=app_record.id,
                 user_id=application_generate_entity.user_id,
                 tenant_id=app_record.tenant_id,
@@ -171,7 +173,11 @@ class ChatAppRunner(AppRunner):
                 memory=memory,
                 message_id=message.id,
                 inputs=inputs,
+                vision_enabled=application_generate_entity.app_config.app_model_config_dict.get("file_upload", {}).get(
+                    "enabled", False
+                ),
             )
+            context_files = retrieved_files or []
 
         # reorganize all inputs and template to prompt messages
         # Include: prompt template, inputs, query(optional), files(optional)
@@ -186,6 +192,7 @@ class ChatAppRunner(AppRunner):
             context=context,
             memory=memory,
             image_detail_config=image_detail_config,
+            context_files=context_files,
         )
 
         # check hosting moderation

api/core/app/apps/completion/app_runner.py

@@ -10,6 +10,7 @@ from core.app.entities.app_invoke_entities import (
     CompletionAppGenerateEntity,
 )
 from core.callback_handler.index_tool_callback_handler import DatasetIndexToolCallbackHandler
+from core.file import File
 from core.model_manager import ModelInstance
 from core.model_runtime.entities.message_entities import ImagePromptMessageContent
 from core.moderation.base import ModerationError
@@ -102,6 +103,7 @@ class CompletionAppRunner(AppRunner):
 
         # get context from datasets
         context = None
+        context_files: list[File] = []
         if app_config.dataset and app_config.dataset.dataset_ids:
             hit_callback = DatasetIndexToolCallbackHandler(
                 queue_manager,
@@ -116,7 +118,7 @@ class CompletionAppRunner(AppRunner):
                 query = inputs.get(dataset_config.retrieve_config.query_variable, "")
 
             dataset_retrieval = DatasetRetrieval(application_generate_entity)
-            context = dataset_retrieval.retrieve(
+            context, retrieved_files = dataset_retrieval.retrieve(
                 app_id=app_record.id,
                 user_id=application_generate_entity.user_id,
                 tenant_id=app_record.tenant_id,
@@ -130,7 +132,11 @@ class CompletionAppRunner(AppRunner):
                 hit_callback=hit_callback,
                 message_id=message.id,
                 inputs=inputs,
+                vision_enabled=application_generate_entity.app_config.app_model_config_dict.get("file_upload", {}).get(
+                    "enabled", False
+                ),
             )
+            context_files = retrieved_files or []
 
         # reorganize all inputs and template to prompt messages
         # Include: prompt template, inputs, query(optional), files(optional)
@@ -144,6 +150,7 @@ class CompletionAppRunner(AppRunner):
             query=query,
             context=context,
             image_detail_config=image_detail_config,
+            context_files=context_files,
         )
 
         # check hosting moderation

api/core/app/apps/workflow/app_runner.py

@@ -18,6 +18,7 @@ from core.workflow.system_variable import SystemVariable
 from core.workflow.variable_loader import VariableLoader
 from core.workflow.workflow_entry import WorkflowEntry
 from extensions.ext_redis import redis_client
+from extensions.otel import WorkflowAppRunnerHandler, trace_span
 from libs.datetime_utils import naive_utc_now
 from models.enums import UserFrom
 from models.workflow import Workflow
@@ -56,6 +57,7 @@ class WorkflowAppRunner(WorkflowBasedAppRunner):
         self._workflow_execution_repository = workflow_execution_repository
         self._workflow_node_execution_repository = workflow_node_execution_repository
 
+    @trace_span(WorkflowAppRunnerHandler)
     def run(self):
         """
         Run application

api/core/callback_handler/index_tool_callback_handler.py

@@ -7,7 +7,7 @@ from core.app.apps.base_app_queue_manager import AppQueueManager, PublishFrom
 from core.app.entities.app_invoke_entities import InvokeFrom
 from core.app.entities.queue_entities import QueueRetrieverResourcesEvent
 from core.rag.entities.citation_metadata import RetrievalSourceMetadata
-from core.rag.index_processor.constant.index_type import IndexType
+from core.rag.index_processor.constant.index_type import IndexStructureType
 from core.rag.models.document import Document
 from extensions.ext_database import db
 from models.dataset import ChildChunk, DatasetQuery, DocumentSegment
@@ -59,7 +59,7 @@ class DatasetIndexToolCallbackHandler:
                         document_id,
                     )
                     continue
-                if dataset_document.doc_form == IndexType.PARENT_CHILD_INDEX:
+                if dataset_document.doc_form == IndexStructureType.PARENT_CHILD_INDEX:
                     child_chunk_stmt = select(ChildChunk).where(
                         ChildChunk.index_node_id == document.metadata["doc_id"],
                         ChildChunk.dataset_id == dataset_document.dataset_id,

api/core/entities/provider_configuration.py

@@ -253,7 +253,7 @@ class ProviderConfiguration(BaseModel):
                 try:
                     credentials[key] = encrypter.decrypt_token(tenant_id=self.tenant_id, token=credentials[key])
                 except Exception:
-                    pass
+                    logger.exception("Failed to decrypt credential secret variable %s", key)
 
         return self.obfuscated_credentials(
             credentials=credentials,
@@ -765,7 +765,7 @@ class ProviderConfiguration(BaseModel):
                 try:
                     credentials[key] = encrypter.decrypt_token(tenant_id=self.tenant_id, token=credentials[key])
                 except Exception:
-                    pass
+                    logger.exception("Failed to decrypt model credential secret variable %s", key)
 
         current_credential_id = credential_record.id
         current_credential_name = credential_record.credential_name

api/core/helper/marketplace.py

@@ -1,3 +1,4 @@
+import logging
 from collections.abc import Sequence
 
 import httpx
@@ -8,6 +9,7 @@ from core.helper.download import download_with_size_limit
 from core.plugin.entities.marketplace import MarketplacePluginDeclaration
 
 marketplace_api_url = URL(str(dify_config.MARKETPLACE_API_URL))
+logger = logging.getLogger(__name__)
 
 
 def get_plugin_pkg_url(plugin_unique_identifier: str) -> str:
@@ -55,7 +57,9 @@ def batch_fetch_plugin_manifests_ignore_deserialization_error(
         try:
             result.append(MarketplacePluginDeclaration.model_validate(plugin))
         except Exception:
-            pass
+            logger.exception(
+                "Failed to deserialize marketplace plugin manifest for %s", plugin.get("plugin_id", "unknown")
+            )
 
     return result
 

api/core/helper/tool_provider_cache.py

@@ -0,0 +1,56 @@
+import json
+import logging
+from typing import Any
+
+from core.tools.entities.api_entities import ToolProviderTypeApiLiteral
+from extensions.ext_redis import redis_client, redis_fallback
+
+logger = logging.getLogger(__name__)
+
+
+class ToolProviderListCache:
+    """Cache for tool provider lists"""
+
+    CACHE_TTL = 300  # 5 minutes
+
+    @staticmethod
+    def _generate_cache_key(tenant_id: str, typ: ToolProviderTypeApiLiteral = None) -> str:
+        """Generate cache key for tool providers list"""
+        type_filter = typ or "all"
+        return f"tool_providers:tenant_id:{tenant_id}:type:{type_filter}"
+
+    @staticmethod
+    @redis_fallback(default_return=None)
+    def get_cached_providers(tenant_id: str, typ: ToolProviderTypeApiLiteral = None) -> list[dict[str, Any]] | None:
+        """Get cached tool providers"""
+        cache_key = ToolProviderListCache._generate_cache_key(tenant_id, typ)
+        cached_data = redis_client.get(cache_key)
+        if cached_data:
+            try:
+                return json.loads(cached_data.decode("utf-8"))
+            except (json.JSONDecodeError, UnicodeDecodeError):
+                logger.warning("Failed to decode cached tool providers data")
+                return None
+        return None
+
+    @staticmethod
+    @redis_fallback()
+    def set_cached_providers(tenant_id: str, typ: ToolProviderTypeApiLiteral, providers: list[dict[str, Any]]):
+        """Cache tool providers"""
+        cache_key = ToolProviderListCache._generate_cache_key(tenant_id, typ)
+        redis_client.setex(cache_key, ToolProviderListCache.CACHE_TTL, json.dumps(providers))
+
+    @staticmethod
+    @redis_fallback()
+    def invalidate_cache(tenant_id: str, typ: ToolProviderTypeApiLiteral = None):
+        """Invalidate cache for tool providers"""
+        if typ:
+            # Invalidate specific type cache
+            cache_key = ToolProviderListCache._generate_cache_key(tenant_id, typ)
+            redis_client.delete(cache_key)
+        else:
+            # Invalidate all caches for this tenant
+            pattern = f"tool_providers:tenant_id:{tenant_id}:*"
+            keys = list(redis_client.scan_iter(pattern))
+            if keys:
+                redis_client.delete(*keys)

api/core/indexing_runner.py

@@ -7,7 +7,7 @@ import time
 import uuid
 from typing import Any
 
-from flask import current_app
+from flask import Flask, current_app
 from sqlalchemy import select
 from sqlalchemy.orm.exc import ObjectDeletedError
 
@@ -21,7 +21,7 @@ from core.rag.datasource.keyword.keyword_factory import Keyword
 from core.rag.docstore.dataset_docstore import DatasetDocumentStore
 from core.rag.extractor.entity.datasource_type import DatasourceType
 from core.rag.extractor.entity.extract_setting import ExtractSetting, NotionInfo, WebsiteInfo
-from core.rag.index_processor.constant.index_type import IndexType
+from core.rag.index_processor.constant.index_type import IndexStructureType
 from core.rag.index_processor.index_processor_base import BaseIndexProcessor
 from core.rag.index_processor.index_processor_factory import IndexProcessorFactory
 from core.rag.models.document import ChildDocument, Document
@@ -36,6 +36,7 @@ from extensions.ext_redis import redis_client
 from extensions.ext_storage import storage
 from libs import helper
 from libs.datetime_utils import naive_utc_now
+from models import Account
 from models.dataset import ChildChunk, Dataset, DatasetProcessRule, DocumentSegment
 from models.dataset import Document as DatasetDocument
 from models.model import UploadFile
@@ -89,8 +90,17 @@ class IndexingRunner:
                 text_docs = self._extract(index_processor, requeried_document, processing_rule.to_dict())
 
                 # transform
+                current_user = db.session.query(Account).filter_by(id=requeried_document.created_by).first()
+                if not current_user:
+                    raise ValueError("no current user found")
+                current_user.set_tenant_id(dataset.tenant_id)
                 documents = self._transform(
-                    index_processor, dataset, text_docs, requeried_document.doc_language, processing_rule.to_dict()
+                    index_processor,
+                    dataset,
+                    text_docs,
+                    requeried_document.doc_language,
+                    processing_rule.to_dict(),
+                    current_user=current_user,
                 )
                 # save segment
                 self._load_segments(dataset, requeried_document, documents)
@@ -136,7 +146,7 @@ class IndexingRunner:
 
             for document_segment in document_segments:
                 db.session.delete(document_segment)
-                if requeried_document.doc_form == IndexType.PARENT_CHILD_INDEX:
+                if requeried_document.doc_form == IndexStructureType.PARENT_CHILD_INDEX:
                     # delete child chunks
                     db.session.query(ChildChunk).where(ChildChunk.segment_id == document_segment.id).delete()
             db.session.commit()
@@ -152,8 +162,17 @@ class IndexingRunner:
             text_docs = self._extract(index_processor, requeried_document, processing_rule.to_dict())
 
             # transform
+            current_user = db.session.query(Account).filter_by(id=requeried_document.created_by).first()
+            if not current_user:
+                raise ValueError("no current user found")
+            current_user.set_tenant_id(dataset.tenant_id)
             documents = self._transform(
-                index_processor, dataset, text_docs, requeried_document.doc_language, processing_rule.to_dict()
+                index_processor,
+                dataset,
+                text_docs,
+                requeried_document.doc_language,
+                processing_rule.to_dict(),
+                current_user=current_user,
             )
             # save segment
             self._load_segments(dataset, requeried_document, documents)
@@ -209,7 +228,7 @@ class IndexingRunner:
                                 "dataset_id": document_segment.dataset_id,
                             },
                         )
-                        if requeried_document.doc_form == IndexType.PARENT_CHILD_INDEX:
+                        if requeried_document.doc_form == IndexStructureType.PARENT_CHILD_INDEX:
                             child_chunks = document_segment.get_child_chunks()
                             if child_chunks:
                                 child_documents = []
@@ -302,6 +321,7 @@ class IndexingRunner:
             text_docs = index_processor.extract(extract_setting, process_rule_mode=tmp_processing_rule["mode"])
             documents = index_processor.transform(
                 text_docs,
+                current_user=None,
                 embedding_model_instance=embedding_model_instance,
                 process_rule=processing_rule.to_dict(),
                 tenant_id=tenant_id,
@@ -551,7 +571,10 @@ class IndexingRunner:
         indexing_start_at = time.perf_counter()
         tokens = 0
         create_keyword_thread = None
-        if dataset_document.doc_form != IndexType.PARENT_CHILD_INDEX and dataset.indexing_technique == "economy":
+        if (
+            dataset_document.doc_form != IndexStructureType.PARENT_CHILD_INDEX
+            and dataset.indexing_technique == "economy"
+        ):
             # create keyword index
             create_keyword_thread = threading.Thread(
                 target=self._process_keyword_index,
@@ -590,7 +613,7 @@ class IndexingRunner:
                 for future in futures:
                     tokens += future.result()
         if (
-            dataset_document.doc_form != IndexType.PARENT_CHILD_INDEX
+            dataset_document.doc_form != IndexStructureType.PARENT_CHILD_INDEX
             and dataset.indexing_technique == "economy"
             and create_keyword_thread is not None
         ):
@@ -635,7 +658,13 @@ class IndexingRunner:
                 db.session.commit()
 
     def _process_chunk(
-        self, flask_app, index_processor, chunk_documents, dataset, dataset_document, embedding_model_instance
+        self,
+        flask_app: Flask,
+        index_processor: BaseIndexProcessor,
+        chunk_documents: list[Document],
+        dataset: Dataset,
+        dataset_document: DatasetDocument,
+        embedding_model_instance: ModelInstance | None,
     ):
         with flask_app.app_context():
             # check document is paused
@@ -646,8 +675,15 @@ class IndexingRunner:
                 page_content_list = [document.page_content for document in chunk_documents]
                 tokens += sum(embedding_model_instance.get_text_embedding_num_tokens(page_content_list))
 
+            multimodal_documents = []
+            for document in chunk_documents:
+                if document.attachments and dataset.is_multimodal:
+                    multimodal_documents.extend(document.attachments)
+
             # load index
-            index_processor.load(dataset, chunk_documents, with_keywords=False)
+            index_processor.load(
+                dataset, chunk_documents, multimodal_documents=multimodal_documents, with_keywords=False
+            )
 
             document_ids = [document.metadata["doc_id"] for document in chunk_documents]
             db.session.query(DocumentSegment).where(
@@ -710,6 +746,7 @@ class IndexingRunner:
         text_docs: list[Document],
         doc_language: str,
         process_rule: dict,
+        current_user: Account | None = None,
     ) -> list[Document]:
         # get embedding model instance
         embedding_model_instance = None
@@ -729,6 +766,7 @@ class IndexingRunner:
 
         documents = index_processor.transform(
             text_docs,
+            current_user,
             embedding_model_instance=embedding_model_instance,
             process_rule=process_rule,
             tenant_id=dataset.tenant_id,
@@ -737,14 +775,16 @@ class IndexingRunner:
 
         return documents
 
-    def _load_segments(self, dataset, dataset_document, documents):
+    def _load_segments(self, dataset: Dataset, dataset_document: DatasetDocument, documents: list[Document]):
         # save node to document segment
         doc_store = DatasetDocumentStore(
             dataset=dataset, user_id=dataset_document.created_by, document_id=dataset_document.id
         )
 
         # add document segments
-        doc_store.add_documents(docs=documents, save_child=dataset_document.doc_form == IndexType.PARENT_CHILD_INDEX)
+        doc_store.add_documents(
+            docs=documents, save_child=dataset_document.doc_form == IndexStructureType.PARENT_CHILD_INDEX
+        )
 
         # update document status to indexing
         cur_time = naive_utc_now()

api/core/llm_generator/llm_generator.py

@@ -15,6 +15,8 @@ from core.llm_generator.prompts import (
     LLM_MODIFY_CODE_SYSTEM,
     LLM_MODIFY_PROMPT_SYSTEM,
     PYTHON_CODE_GENERATOR_PROMPT_TEMPLATE,
+    SUGGESTED_QUESTIONS_MAX_TOKENS,
+    SUGGESTED_QUESTIONS_TEMPERATURE,
     SYSTEM_STRUCTURED_OUTPUT_GENERATE,
     WORKFLOW_RULE_CONFIG_PROMPT_GENERATE_TEMPLATE,
 )
@@ -124,7 +126,10 @@ class LLMGenerator:
         try:
             response: LLMResult = model_instance.invoke_llm(
                 prompt_messages=list(prompt_messages),
-                model_parameters={"max_tokens": 256, "temperature": 0},
+                model_parameters={
+                    "max_tokens": SUGGESTED_QUESTIONS_MAX_TOKENS,
+                    "temperature": SUGGESTED_QUESTIONS_TEMPERATURE,
+                },
                 stream=False,
             )
 

api/core/llm_generator/prompts.py

@@ -1,4 +1,6 @@
 # Written by YORKI MINAKO🤡, Edited by Xiaoyi, Edited by yasu-oh
+import os
+
 CONVERSATION_TITLE_PROMPT = """You are asked to generate a concise chat title by decomposing the user’s input into two parts: “Intention” and “Subject”.
 
 1. Detect Input Language
@@ -94,7 +96,8 @@ JAVASCRIPT_CODE_GENERATOR_PROMPT_TEMPLATE = (
 )
 
 
-SUGGESTED_QUESTIONS_AFTER_ANSWER_INSTRUCTION_PROMPT = (
+# Default prompt for suggested questions (can be overridden by environment variable)
+_DEFAULT_SUGGESTED_QUESTIONS_AFTER_ANSWER_PROMPT = (
     "Please help me predict the three most likely questions that human would ask, "
     "and keep each question under 20 characters.\n"
     "MAKE SURE your output is the SAME language as the Assistant's latest response. "
@@ -102,6 +105,15 @@ SUGGESTED_QUESTIONS_AFTER_ANSWER_INSTRUCTION_PROMPT = (
     '["question1","question2","question3"]\n'
 )
 
+# Environment variable override for suggested questions prompt
+SUGGESTED_QUESTIONS_AFTER_ANSWER_INSTRUCTION_PROMPT = os.getenv(
+    "SUGGESTED_QUESTIONS_PROMPT", _DEFAULT_SUGGESTED_QUESTIONS_AFTER_ANSWER_PROMPT
+)
+
+# Configurable LLM parameters for suggested questions (can be overridden by environment variables)
+SUGGESTED_QUESTIONS_MAX_TOKENS = int(os.getenv("SUGGESTED_QUESTIONS_MAX_TOKENS", "256"))
+SUGGESTED_QUESTIONS_TEMPERATURE = float(os.getenv("SUGGESTED_QUESTIONS_TEMPERATURE", "0"))
+
 GENERATOR_QA_PROMPT = (
     "<Task> The user will send a long text. Generate a Question and Answer pairs only using the knowledge"
     " in the long text. Please think step by step."

api/core/model_manager.py

@@ -10,9 +10,9 @@ from core.errors.error import ProviderTokenNotInitError
 from core.model_runtime.callbacks.base_callback import Callback
 from core.model_runtime.entities.llm_entities import LLMResult
 from core.model_runtime.entities.message_entities import PromptMessage, PromptMessageTool
-from core.model_runtime.entities.model_entities import ModelType
+from core.model_runtime.entities.model_entities import ModelFeature, ModelType
 from core.model_runtime.entities.rerank_entities import RerankResult
-from core.model_runtime.entities.text_embedding_entities import TextEmbeddingResult
+from core.model_runtime.entities.text_embedding_entities import EmbeddingResult
 from core.model_runtime.errors.invoke import InvokeAuthorizationError, InvokeConnectionError, InvokeRateLimitError
 from core.model_runtime.model_providers.__base.large_language_model import LargeLanguageModel
 from core.model_runtime.model_providers.__base.moderation_model import ModerationModel
@@ -200,7 +200,7 @@ class ModelInstance:
 
     def invoke_text_embedding(
         self, texts: list[str], user: str | None = None, input_type: EmbeddingInputType = EmbeddingInputType.DOCUMENT
-    ) -> TextEmbeddingResult:
+    ) -> EmbeddingResult:
         """
         Invoke large language model
 
@@ -212,7 +212,7 @@ class ModelInstance:
         if not isinstance(self.model_type_instance, TextEmbeddingModel):
             raise Exception("Model type instance is not TextEmbeddingModel")
         return cast(
-            TextEmbeddingResult,
+            EmbeddingResult,
             self._round_robin_invoke(
                 function=self.model_type_instance.invoke,
                 model=self.model,
@@ -223,6 +223,34 @@ class ModelInstance:
             ),
         )
 
+    def invoke_multimodal_embedding(
+        self,
+        multimodel_documents: list[dict],
+        user: str | None = None,
+        input_type: EmbeddingInputType = EmbeddingInputType.DOCUMENT,
+    ) -> EmbeddingResult:
+        """
+        Invoke large language model
+
+        :param multimodel_documents: multimodel documents to embed
+        :param user: unique user id
+        :param input_type: input type
+        :return: embeddings result
+        """
+        if not isinstance(self.model_type_instance, TextEmbeddingModel):
+            raise Exception("Model type instance is not TextEmbeddingModel")
+        return cast(
+            EmbeddingResult,
+            self._round_robin_invoke(
+                function=self.model_type_instance.invoke,
+                model=self.model,
+                credentials=self.credentials,
+                multimodel_documents=multimodel_documents,
+                user=user,
+                input_type=input_type,
+            ),
+        )
+
     def get_text_embedding_num_tokens(self, texts: list[str]) -> list[int]:
         """
         Get number of tokens for text embedding
@@ -276,6 +304,40 @@ class ModelInstance:
             ),
         )
 
+    def invoke_multimodal_rerank(
+        self,
+        query: dict,
+        docs: list[dict],
+        score_threshold: float | None = None,
+        top_n: int | None = None,
+        user: str | None = None,
+    ) -> RerankResult:
+        """
+        Invoke rerank model
+
+        :param query: search query
+        :param docs: docs for reranking
+        :param score_threshold: score threshold
+        :param top_n: top n
+        :param user: unique user id
+        :return: rerank result
+        """
+        if not isinstance(self.model_type_instance, RerankModel):
+            raise Exception("Model type instance is not RerankModel")
+        return cast(
+            RerankResult,
+            self._round_robin_invoke(
+                function=self.model_type_instance.invoke_multimodal_rerank,
+                model=self.model,
+                credentials=self.credentials,
+                query=query,
+                docs=docs,
+                score_threshold=score_threshold,
+                top_n=top_n,
+                user=user,
+            ),
+        )
+
     def invoke_moderation(self, text: str, user: str | None = None) -> bool:
         """
         Invoke moderation model
@@ -461,6 +523,32 @@ class ModelManager:
             model=default_model_entity.model,
         )
 
+    def check_model_support_vision(self, tenant_id: str, provider: str, model: str, model_type: ModelType) -> bool:
+        """
+        Check if model supports vision
+        :param tenant_id: tenant id
+        :param provider: provider name
+        :param model: model name
+        :return: True if model supports vision, False otherwise
+        """
+        model_instance = self.get_model_instance(tenant_id, provider, model_type, model)
+        model_type_instance = model_instance.model_type_instance
+        match model_type:
+            case ModelType.LLM:
+                model_type_instance = cast(LargeLanguageModel, model_type_instance)
+            case ModelType.TEXT_EMBEDDING:
+                model_type_instance = cast(TextEmbeddingModel, model_type_instance)
+            case ModelType.RERANK:
+                model_type_instance = cast(RerankModel, model_type_instance)
+            case _:
+                raise ValueError(f"Model type {model_type} is not supported")
+        model_schema = model_type_instance.get_model_schema(model, model_instance.credentials)
+        if not model_schema:
+            return False
+        if model_schema.features and ModelFeature.VISION in model_schema.features:
+            return True
+        return False
+
 
 class LBModelManager:
     def __init__(

api/core/model_runtime/entities/text_embedding_entities.py

@@ -19,7 +19,7 @@ class EmbeddingUsage(ModelUsage):
     latency: float
 
 
-class TextEmbeddingResult(BaseModel):
+class EmbeddingResult(BaseModel):
     """
     Model class for text embedding result.
     """
@@ -27,3 +27,13 @@ class TextEmbeddingResult(BaseModel):
     model: str
     embeddings: list[list[float]]
     usage: EmbeddingUsage
+
+
+class FileEmbeddingResult(BaseModel):
+    """
+    Model class for file embedding result.
+    """
+
+    model: str
+    embeddings: list[list[float]]
+    usage: EmbeddingUsage

api/core/model_runtime/model_providers/__base/rerank_model.py

@@ -50,3 +50,43 @@ class RerankModel(AIModel):
             )
         except Exception as e:
             raise self._transform_invoke_error(e)
+
+    def invoke_multimodal_rerank(
+        self,
+        model: str,
+        credentials: dict,
+        query: dict,
+        docs: list[dict],
+        score_threshold: float | None = None,
+        top_n: int | None = None,
+        user: str | None = None,
+    ) -> RerankResult:
+        """
+        Invoke multimodal rerank model
+        :param model: model name
+        :param credentials: model credentials
+        :param query: search query
+        :param docs: docs for reranking
+        :param score_threshold: score threshold
+        :param top_n: top n
+        :param user: unique user id
+        :return: rerank result
+        """
+        try:
+            from core.plugin.impl.model import PluginModelClient
+
+            plugin_model_manager = PluginModelClient()
+            return plugin_model_manager.invoke_multimodal_rerank(
+                tenant_id=self.tenant_id,
+                user_id=user or "unknown",
+                plugin_id=self.plugin_id,
+                provider=self.provider_name,
+                model=model,
+                credentials=credentials,
+                query=query,
+                docs=docs,
+                score_threshold=score_threshold,
+                top_n=top_n,
+            )
+        except Exception as e:
+            raise self._transform_invoke_error(e)

api/core/model_runtime/model_providers/__base/text_embedding_model.py

@@ -2,7 +2,7 @@ from pydantic import ConfigDict
 
 from core.entities.embedding_type import EmbeddingInputType
 from core.model_runtime.entities.model_entities import ModelPropertyKey, ModelType
-from core.model_runtime.entities.text_embedding_entities import TextEmbeddingResult
+from core.model_runtime.entities.text_embedding_entities import EmbeddingResult
 from core.model_runtime.model_providers.__base.ai_model import AIModel
 
 
@@ -20,16 +20,18 @@ class TextEmbeddingModel(AIModel):
         self,
         model: str,
         credentials: dict,
-        texts: list[str],
+        texts: list[str] | None = None,
+        multimodel_documents: list[dict] | None = None,
         user: str | None = None,
         input_type: EmbeddingInputType = EmbeddingInputType.DOCUMENT,
-    ) -> TextEmbeddingResult:
+    ) -> EmbeddingResult:
         """
         Invoke text embedding model
 
         :param model: model name
         :param credentials: model credentials
         :param texts: texts to embed
+        :param files: files to embed
         :param user: unique user id
         :param input_type: input type
         :return: embeddings result
@@ -38,16 +40,29 @@ class TextEmbeddingModel(AIModel):
 
         try:
             plugin_model_manager = PluginModelClient()
-            return plugin_model_manager.invoke_text_embedding(
-                tenant_id=self.tenant_id,
-                user_id=user or "unknown",
-                plugin_id=self.plugin_id,
-                provider=self.provider_name,
-                model=model,
-                credentials=credentials,
-                texts=texts,
-                input_type=input_type,
-            )
+            if texts:
+                return plugin_model_manager.invoke_text_embedding(
+                    tenant_id=self.tenant_id,
+                    user_id=user or "unknown",
+                    plugin_id=self.plugin_id,
+                    provider=self.provider_name,
+                    model=model,
+                    credentials=credentials,
+                    texts=texts,
+                    input_type=input_type,
+                )
+            if multimodel_documents:
+                return plugin_model_manager.invoke_multimodal_embedding(
+                    tenant_id=self.tenant_id,
+                    user_id=user or "unknown",
+                    plugin_id=self.plugin_id,
+                    provider=self.provider_name,
+                    model=model,
+                    credentials=credentials,
+                    documents=multimodel_documents,
+                    input_type=input_type,
+                )
+            raise ValueError("No texts or files provided")
         except Exception as e:
             raise self._transform_invoke_error(e)
 

api/core/ops/tencent_trace/tencent_trace.py

@@ -521,4 +521,4 @@ class TencentDataTrace(BaseTraceInstance):
             if hasattr(self, "trace_client"):
                 self.trace_client.shutdown()
         except Exception:
-            pass
+            logger.exception("[Tencent APM] Failed to shutdown trace client during cleanup")

api/core/plugin/impl/model.py

@@ -6,7 +6,7 @@ from core.model_runtime.entities.llm_entities import LLMResultChunk
 from core.model_runtime.entities.message_entities import PromptMessage, PromptMessageTool
 from core.model_runtime.entities.model_entities import AIModelEntity
 from core.model_runtime.entities.rerank_entities import RerankResult
-from core.model_runtime.entities.text_embedding_entities import TextEmbeddingResult
+from core.model_runtime.entities.text_embedding_entities import EmbeddingResult
 from core.model_runtime.utils.encoders import jsonable_encoder
 from core.plugin.entities.plugin_daemon import (
     PluginBasicBooleanResponse,
@@ -243,14 +243,14 @@ class PluginModelClient(BasePluginClient):
         credentials: dict,
         texts: list[str],
         input_type: str,
-    ) -> TextEmbeddingResult:
+    ) -> EmbeddingResult:
         """
         Invoke text embedding
         """
         response = self._request_with_plugin_daemon_response_stream(
             method="POST",
             path=f"plugin/{tenant_id}/dispatch/text_embedding/invoke",
-            type_=TextEmbeddingResult,
+            type_=EmbeddingResult,
             data=jsonable_encoder(
                 {
                     "user_id": user_id,
@@ -275,6 +275,48 @@ class PluginModelClient(BasePluginClient):
 
         raise ValueError("Failed to invoke text embedding")
 
+    def invoke_multimodal_embedding(
+        self,
+        tenant_id: str,
+        user_id: str,
+        plugin_id: str,
+        provider: str,
+        model: str,
+        credentials: dict,
+        documents: list[dict],
+        input_type: str,
+    ) -> EmbeddingResult:
+        """
+        Invoke file embedding
+        """
+        response = self._request_with_plugin_daemon_response_stream(
+            method="POST",
+            path=f"plugin/{tenant_id}/dispatch/multimodal_embedding/invoke",
+            type_=EmbeddingResult,
+            data=jsonable_encoder(
+                {
+                    "user_id": user_id,
+                    "data": {
+                        "provider": provider,
+                        "model_type": "text-embedding",
+                        "model": model,
+                        "credentials": credentials,
+                        "documents": documents,
+                        "input_type": input_type,
+                    },
+                }
+            ),
+            headers={
+                "X-Plugin-ID": plugin_id,
+                "Content-Type": "application/json",
+            },
+        )
+
+        for resp in response:
+            return resp
+
+        raise ValueError("Failed to invoke file embedding")
+
     def get_text_embedding_num_tokens(
         self,
         tenant_id: str,
@@ -361,6 +403,51 @@ class PluginModelClient(BasePluginClient):
 
         raise ValueError("Failed to invoke rerank")
 
+    def invoke_multimodal_rerank(
+        self,
+        tenant_id: str,
+        user_id: str,
+        plugin_id: str,
+        provider: str,
+        model: str,
+        credentials: dict,
+        query: dict,
+        docs: list[dict],
+        score_threshold: float | None = None,
+        top_n: int | None = None,
+    ) -> RerankResult:
+        """
+        Invoke multimodal rerank
+        """
+        response = self._request_with_plugin_daemon_response_stream(
+            method="POST",
+            path=f"plugin/{tenant_id}/dispatch/multimodal_rerank/invoke",
+            type_=RerankResult,
+            data=jsonable_encoder(
+                {
+                    "user_id": user_id,
+                    "data": {
+                        "provider": provider,
+                        "model_type": "rerank",
+                        "model": model,
+                        "credentials": credentials,
+                        "query": query,
+                        "docs": docs,
+                        "score_threshold": score_threshold,
+                        "top_n": top_n,
+                    },
+                }
+            ),
+            headers={
+                "X-Plugin-ID": plugin_id,
+                "Content-Type": "application/json",
+            },
+        )
+        for resp in response:
+            return resp
+
+        raise ValueError("Failed to invoke multimodal rerank")
+
     def invoke_tts(
         self,
         tenant_id: str,

api/core/prompt/simple_prompt_transform.py

@@ -49,6 +49,7 @@ class SimplePromptTransform(PromptTransform):
         memory: TokenBufferMemory | None,
         model_config: ModelConfigWithCredentialsEntity,
         image_detail_config: ImagePromptMessageContent.DETAIL | None = None,
+        context_files: list["File"] | None = None,
     ) -> tuple[list[PromptMessage], list[str] | None]:
         inputs = {key: str(value) for key, value in inputs.items()}
 
@@ -64,6 +65,7 @@ class SimplePromptTransform(PromptTransform):
                 memory=memory,
                 model_config=model_config,
                 image_detail_config=image_detail_config,
+                context_files=context_files,
             )
         else:
             prompt_messages, stops = self._get_completion_model_prompt_messages(
@@ -76,6 +78,7 @@ class SimplePromptTransform(PromptTransform):
                 memory=memory,
                 model_config=model_config,
                 image_detail_config=image_detail_config,
+                context_files=context_files,
             )
 
         return prompt_messages, stops
@@ -187,6 +190,7 @@ class SimplePromptTransform(PromptTransform):
         memory: TokenBufferMemory | None,
         model_config: ModelConfigWithCredentialsEntity,
         image_detail_config: ImagePromptMessageContent.DETAIL | None = None,
+        context_files: list["File"] | None = None,
     ) -> tuple[list[PromptMessage], list[str] | None]:
         prompt_messages: list[PromptMessage] = []
 
@@ -216,9 +220,9 @@ class SimplePromptTransform(PromptTransform):
             )
 
         if query:
-            prompt_messages.append(self._get_last_user_message(query, files, image_detail_config))
+            prompt_messages.append(self._get_last_user_message(query, files, image_detail_config, context_files))
         else:
-            prompt_messages.append(self._get_last_user_message(prompt, files, image_detail_config))
+            prompt_messages.append(self._get_last_user_message(prompt, files, image_detail_config, context_files))
 
         return prompt_messages, None
 
@@ -233,6 +237,7 @@ class SimplePromptTransform(PromptTransform):
         memory: TokenBufferMemory | None,
         model_config: ModelConfigWithCredentialsEntity,
         image_detail_config: ImagePromptMessageContent.DETAIL | None = None,
+        context_files: list["File"] | None = None,
     ) -> tuple[list[PromptMessage], list[str] | None]:
         # get prompt
         prompt, prompt_rules = self._get_prompt_str_and_rules(
@@ -275,20 +280,27 @@ class SimplePromptTransform(PromptTransform):
         if stops is not None and len(stops) == 0:
             stops = None
 
-        return [self._get_last_user_message(prompt, files, image_detail_config)], stops
+        return [self._get_last_user_message(prompt, files, image_detail_config, context_files)], stops
 
     def _get_last_user_message(
         self,
         prompt: str,
         files: Sequence["File"],
         image_detail_config: ImagePromptMessageContent.DETAIL | None = None,
+        context_files: list["File"] | None = None,
     ) -> UserPromptMessage:
+        prompt_message_contents: list[PromptMessageContentUnionTypes] = []
         if files:
-            prompt_message_contents: list[PromptMessageContentUnionTypes] = []
             for file in files:
                 prompt_message_contents.append(
                     file_manager.to_prompt_message_content(file, image_detail_config=image_detail_config)
                 )
+        if context_files:
+            for file in context_files:
+                prompt_message_contents.append(
+                    file_manager.to_prompt_message_content(file, image_detail_config=image_detail_config)
+                )
+        if prompt_message_contents:
             prompt_message_contents.append(TextPromptMessageContent(data=prompt))
 
             prompt_message = UserPromptMessage(content=prompt_message_contents)

api/core/rag/data_post_processor/data_post_processor.py

@@ -2,6 +2,7 @@ from core.model_manager import ModelInstance, ModelManager
 from core.model_runtime.entities.model_entities import ModelType
 from core.model_runtime.errors.invoke import InvokeAuthorizationError
 from core.rag.data_post_processor.reorder import ReorderRunner
+from core.rag.index_processor.constant.query_type import QueryType
 from core.rag.models.document import Document
 from core.rag.rerank.entity.weight import KeywordSetting, VectorSetting, Weights
 from core.rag.rerank.rerank_base import BaseRerankRunner
@@ -30,9 +31,10 @@ class DataPostProcessor:
         score_threshold: float | None = None,
         top_n: int | None = None,
         user: str | None = None,
+        query_type: QueryType = QueryType.TEXT_QUERY,
     ) -> list[Document]:
         if self.rerank_runner:
-            documents = self.rerank_runner.run(query, documents, score_threshold, top_n, user)
+            documents = self.rerank_runner.run(query, documents, score_threshold, top_n, user, query_type)
 
         if self.reorder_runner:
             documents = self.reorder_runner.run(documents)

api/core/rag/datasource/retrieval_service.py

@@ -1,23 +1,30 @@
 import concurrent.futures
 from concurrent.futures import ThreadPoolExecutor
+from typing import Any
 
 from flask import Flask, current_app
 from sqlalchemy import select
 from sqlalchemy.orm import Session, load_only
 
 from configs import dify_config
+from core.model_manager import ModelManager
+from core.model_runtime.entities.model_entities import ModelType
 from core.rag.data_post_processor.data_post_processor import DataPostProcessor
 from core.rag.datasource.keyword.keyword_factory import Keyword
 from core.rag.datasource.vdb.vector_factory import Vector
 from core.rag.embedding.retrieval import RetrievalSegments
 from core.rag.entities.metadata_entities import MetadataCondition
-from core.rag.index_processor.constant.index_type import IndexType
+from core.rag.index_processor.constant.doc_type import DocType
+from core.rag.index_processor.constant.index_type import IndexStructureType
+from core.rag.index_processor.constant.query_type import QueryType
 from core.rag.models.document import Document
 from core.rag.rerank.rerank_type import RerankMode
 from core.rag.retrieval.retrieval_methods import RetrievalMethod
+from core.tools.signature import sign_upload_file
 from extensions.ext_database import db
-from models.dataset import ChildChunk, Dataset, DocumentSegment
+from models.dataset import ChildChunk, Dataset, DocumentSegment, SegmentAttachmentBinding
 from models.dataset import Document as DatasetDocument
+from models.model import UploadFile
 from services.external_knowledge_service import ExternalDatasetService
 
 default_retrieval_model = {
@@ -37,14 +44,15 @@ class RetrievalService:
         retrieval_method: RetrievalMethod,
         dataset_id: str,
         query: str,
-        top_k: int,
+        top_k: int = 4,
         score_threshold: float | None = 0.0,
         reranking_model: dict | None = None,
         reranking_mode: str = "reranking_model",
         weights: dict | None = None,
         document_ids_filter: list[str] | None = None,
+        attachment_ids: list | None = None,
     ):
-        if not query:
+        if not query and not attachment_ids:
             return []
         dataset = cls._get_dataset(dataset_id)
         if not dataset:
@@ -56,69 +64,52 @@ class RetrievalService:
         # Optimize multithreading with thread pools
         with ThreadPoolExecutor(max_workers=dify_config.RETRIEVAL_SERVICE_EXECUTORS) as executor:  # type: ignore
             futures = []
-            if retrieval_method == RetrievalMethod.KEYWORD_SEARCH:
+            retrieval_service = RetrievalService()
+            if query:
                 futures.append(
                     executor.submit(
-                        cls.keyword_search,
+                        retrieval_service._retrieve,
                         flask_app=current_app._get_current_object(),  # type: ignore
-                        dataset_id=dataset_id,
-                        query=query,
-                        top_k=top_k,
-                        all_documents=all_documents,
-                        exceptions=exceptions,
-                        document_ids_filter=document_ids_filter,
-                    )
-                )
-            if RetrievalMethod.is_support_semantic_search(retrieval_method):
-                futures.append(
-                    executor.submit(
-                        cls.embedding_search,
-                        flask_app=current_app._get_current_object(),  # type: ignore
-                        dataset_id=dataset_id,
+                        retrieval_method=retrieval_method,
+                        dataset=dataset,
                         query=query,
                         top_k=top_k,
                         score_threshold=score_threshold,
                         reranking_model=reranking_model,
-                        all_documents=all_documents,
-                        retrieval_method=retrieval_method,
-                        exceptions=exceptions,
+                        reranking_mode=reranking_mode,
+                        weights=weights,
                         document_ids_filter=document_ids_filter,
+                        attachment_id=None,
+                        all_documents=all_documents,
+                        exceptions=exceptions,
                     )
                 )
-            if RetrievalMethod.is_support_fulltext_search(retrieval_method):
-                futures.append(
-                    executor.submit(
-                        cls.full_text_index_search,
-                        flask_app=current_app._get_current_object(),  # type: ignore
-                        dataset_id=dataset_id,
-                        query=query,
-                        top_k=top_k,
-                        score_threshold=score_threshold,
-                        reranking_model=reranking_model,
-                        all_documents=all_documents,
-                        retrieval_method=retrieval_method,
-                        exceptions=exceptions,
-                        document_ids_filter=document_ids_filter,
+            if attachment_ids:
+                for attachment_id in attachment_ids:
+                    futures.append(
+                        executor.submit(
+                            retrieval_service._retrieve,
+                            flask_app=current_app._get_current_object(),  # type: ignore
+                            retrieval_method=retrieval_method,
+                            dataset=dataset,
+                            query=None,
+                            top_k=top_k,
+                            score_threshold=score_threshold,
+                            reranking_model=reranking_model,
+                            reranking_mode=reranking_mode,
+                            weights=weights,
+                            document_ids_filter=document_ids_filter,
+                            attachment_id=attachment_id,
+                            all_documents=all_documents,
+                            exceptions=exceptions,
+                        )
                     )
-                )
-            concurrent.futures.wait(futures, timeout=30, return_when=concurrent.futures.ALL_COMPLETED)
+
+            concurrent.futures.wait(futures, timeout=3600, return_when=concurrent.futures.ALL_COMPLETED)
 
         if exceptions:
             raise ValueError(";\n".join(exceptions))
 
-        # Deduplicate documents for hybrid search to avoid duplicate chunks
-        if retrieval_method == RetrievalMethod.HYBRID_SEARCH:
-            all_documents = cls._deduplicate_documents(all_documents)
-            data_post_processor = DataPostProcessor(
-                str(dataset.tenant_id), reranking_mode, reranking_model, weights, False
-            )
-            all_documents = data_post_processor.invoke(
-                query=query,
-                documents=all_documents,
-                score_threshold=score_threshold,
-                top_n=top_k,
-            )
-
         return all_documents
 
     @classmethod
@@ -223,6 +214,7 @@ class RetrievalService:
         retrieval_method: RetrievalMethod,
         exceptions: list,
         document_ids_filter: list[str] | None = None,
+        query_type: QueryType = QueryType.TEXT_QUERY,
     ):
         with flask_app.app_context():
             try:
@@ -231,14 +223,30 @@ class RetrievalService:
                     raise ValueError("dataset not found")
 
                 vector = Vector(dataset=dataset)
-                documents = vector.search_by_vector(
-                    query,
-                    search_type="similarity_score_threshold",
-                    top_k=top_k,
-                    score_threshold=score_threshold,
-                    filter={"group_id": [dataset.id]},
-                    document_ids_filter=document_ids_filter,
-                )
+                documents = []
+                if query_type == QueryType.TEXT_QUERY:
+                    documents.extend(
+                        vector.search_by_vector(
+                            query,
+                            search_type="similarity_score_threshold",
+                            top_k=top_k,
+                            score_threshold=score_threshold,
+                            filter={"group_id": [dataset.id]},
+                            document_ids_filter=document_ids_filter,
+                        )
+                    )
+                if query_type == QueryType.IMAGE_QUERY:
+                    if not dataset.is_multimodal:
+                        return
+                    documents.extend(
+                        vector.search_by_file(
+                            file_id=query,
+                            top_k=top_k,
+                            score_threshold=score_threshold,
+                            filter={"group_id": [dataset.id]},
+                            document_ids_filter=document_ids_filter,
+                        )
+                    )
 
                 if documents:
                     if (
@@ -250,14 +258,37 @@ class RetrievalService:
                         data_post_processor = DataPostProcessor(
                             str(dataset.tenant_id), str(RerankMode.RERANKING_MODEL), reranking_model, None, False
                         )
-                        all_documents.extend(
-                            data_post_processor.invoke(
-                                query=query,
-                                documents=documents,
-                                score_threshold=score_threshold,
-                                top_n=len(documents),
+                        if dataset.is_multimodal:
+                            model_manager = ModelManager()
+                            is_support_vision = model_manager.check_model_support_vision(
+                                tenant_id=dataset.tenant_id,
+                                provider=reranking_model.get("reranking_provider_name") or "",
+                                model=reranking_model.get("reranking_model_name") or "",
+                                model_type=ModelType.RERANK,
+                            )
+                            if is_support_vision:
+                                all_documents.extend(
+                                    data_post_processor.invoke(
+                                        query=query,
+                                        documents=documents,
+                                        score_threshold=score_threshold,
+                                        top_n=len(documents),
+                                        query_type=query_type,
+                                    )
+                                )
+                            else:
+                                # not effective, return original documents
+                                all_documents.extend(documents)
+                        else:
+                            all_documents.extend(
+                                data_post_processor.invoke(
+                                    query=query,
+                                    documents=documents,
+                                    score_threshold=score_threshold,
+                                    top_n=len(documents),
+                                    query_type=query_type,
+                                )
                             )
-                        )
                     else:
                         all_documents.extend(documents)
             except Exception as e:
@@ -339,103 +370,159 @@ class RetrievalService:
             records = []
             include_segment_ids = set()
             segment_child_map = {}
-
-            # Process documents
-            for document in documents:
-                document_id = document.metadata.get("document_id")
-                if document_id not in dataset_documents:
-                    continue
-
-                dataset_document = dataset_documents[document_id]
-                if not dataset_document:
-                    continue
-
-                if dataset_document.doc_form == IndexType.PARENT_CHILD_INDEX:
-                    # Handle parent-child documents
-                    child_index_node_id = document.metadata.get("doc_id")
-                    child_chunk_stmt = select(ChildChunk).where(ChildChunk.index_node_id == child_index_node_id)
-                    child_chunk = db.session.scalar(child_chunk_stmt)
-
-                    if not child_chunk:
+            segment_file_map = {}
+            with Session(db.engine) as session:
+                # Process documents
+                for document in documents:
+                    segment_id = None
+                    attachment_info = None
+                    child_chunk = None
+                    document_id = document.metadata.get("document_id")
+                    if document_id not in dataset_documents:
                         continue
 
-                    segment = (
-                        db.session.query(DocumentSegment)
-                        .where(
-                            DocumentSegment.dataset_id == dataset_document.dataset_id,
-                            DocumentSegment.enabled == True,
-                            DocumentSegment.status == "completed",
-                            DocumentSegment.id == child_chunk.segment_id,
-                        )
-                        .options(
-                            load_only(
-                                DocumentSegment.id,
-                                DocumentSegment.content,
-                                DocumentSegment.answer,
+                    dataset_document = dataset_documents[document_id]
+                    if not dataset_document:
+                        continue
+
+                    if dataset_document.doc_form == IndexStructureType.PARENT_CHILD_INDEX:
+                        # Handle parent-child documents
+                        if document.metadata.get("doc_type") == DocType.IMAGE:
+                            attachment_info_dict = cls.get_segment_attachment_info(
+                                dataset_document.dataset_id,
+                                dataset_document.tenant_id,
+                                document.metadata.get("doc_id") or "",
+                                session,
                             )
+                            if attachment_info_dict:
+                                attachment_info = attachment_info_dict["attchment_info"]
+                                segment_id = attachment_info_dict["segment_id"]
+                        else:
+                            child_index_node_id = document.metadata.get("doc_id")
+                            child_chunk_stmt = select(ChildChunk).where(ChildChunk.index_node_id == child_index_node_id)
+                            child_chunk = session.scalar(child_chunk_stmt)
+
+                            if not child_chunk:
+                                continue
+                            segment_id = child_chunk.segment_id
+
+                        if not segment_id:
+                            continue
+
+                        segment = (
+                            session.query(DocumentSegment)
+                            .where(
+                                DocumentSegment.dataset_id == dataset_document.dataset_id,
+                                DocumentSegment.enabled == True,
+                                DocumentSegment.status == "completed",
+                                DocumentSegment.id == segment_id,
+                            )
+                            .options(
+                                load_only(
+                                    DocumentSegment.id,
+                                    DocumentSegment.content,
+                                    DocumentSegment.answer,
+                                )
+                            )
+                            .first()
                         )
-                        .first()
-                    )
 
-                    if not segment:
-                        continue
+                        if not segment:
+                            continue
 
-                    if segment.id not in include_segment_ids:
-                        include_segment_ids.add(segment.id)
-                        child_chunk_detail = {
-                            "id": child_chunk.id,
-                            "content": child_chunk.content,
-                            "position": child_chunk.position,
-                            "score": document.metadata.get("score", 0.0),
-                        }
-                        map_detail = {
-                            "max_score": document.metadata.get("score", 0.0),
-                            "child_chunks": [child_chunk_detail],
-                        }
-                        segment_child_map[segment.id] = map_detail
-                        record = {
-                            "segment": segment,
-                        }
-                        records.append(record)
+                        if segment.id not in include_segment_ids:
+                            include_segment_ids.add(segment.id)
+                            if child_chunk:
+                                child_chunk_detail = {
+                                    "id": child_chunk.id,
+                                    "content": child_chunk.content,
+                                    "position": child_chunk.position,
+                                    "score": document.metadata.get("score", 0.0),
+                                }
+                                map_detail = {
+                                    "max_score": document.metadata.get("score", 0.0),
+                                    "child_chunks": [child_chunk_detail],
+                                }
+                                segment_child_map[segment.id] = map_detail
+                            record = {
+                                "segment": segment,
+                            }
+                            if attachment_info:
+                                segment_file_map[segment.id] = [attachment_info]
+                            records.append(record)
+                        else:
+                            if child_chunk:
+                                child_chunk_detail = {
+                                    "id": child_chunk.id,
+                                    "content": child_chunk.content,
+                                    "position": child_chunk.position,
+                                    "score": document.metadata.get("score", 0.0),
+                                }
+                                segment_child_map[segment.id]["child_chunks"].append(child_chunk_detail)
+                                segment_child_map[segment.id]["max_score"] = max(
+                                    segment_child_map[segment.id]["max_score"], document.metadata.get("score", 0.0)
+                                )
+                            if attachment_info:
+                                segment_file_map[segment.id].append(attachment_info)
                     else:
-                        child_chunk_detail = {
-                            "id": child_chunk.id,
-                            "content": child_chunk.content,
-                            "position": child_chunk.position,
-                            "score": document.metadata.get("score", 0.0),
-                        }
-                        segment_child_map[segment.id]["child_chunks"].append(child_chunk_detail)
-                        segment_child_map[segment.id]["max_score"] = max(
-                            segment_child_map[segment.id]["max_score"], document.metadata.get("score", 0.0)
-                        )
-                else:
-                    # Handle normal documents
-                    index_node_id = document.metadata.get("doc_id")
-                    if not index_node_id:
-                        continue
-                    document_segment_stmt = select(DocumentSegment).where(
-                        DocumentSegment.dataset_id == dataset_document.dataset_id,
-                        DocumentSegment.enabled == True,
-                        DocumentSegment.status == "completed",
-                        DocumentSegment.index_node_id == index_node_id,
-                    )
-                    segment = db.session.scalar(document_segment_stmt)
+                        # Handle normal documents
+                        segment = None
+                        if document.metadata.get("doc_type") == DocType.IMAGE:
+                            attachment_info_dict = cls.get_segment_attachment_info(
+                                dataset_document.dataset_id,
+                                dataset_document.tenant_id,
+                                document.metadata.get("doc_id") or "",
+                                session,
+                            )
+                            if attachment_info_dict:
+                                attachment_info = attachment_info_dict["attchment_info"]
+                                segment_id = attachment_info_dict["segment_id"]
+                                document_segment_stmt = select(DocumentSegment).where(
+                                    DocumentSegment.dataset_id == dataset_document.dataset_id,
+                                    DocumentSegment.enabled == True,
+                                    DocumentSegment.status == "completed",
+                                    DocumentSegment.id == segment_id,
+                                )
+                                segment = db.session.scalar(document_segment_stmt)
+                                if segment:
+                                    segment_file_map[segment.id] = [attachment_info]
+                        else:
+                            index_node_id = document.metadata.get("doc_id")
+                            if not index_node_id:
+                                continue
+                            document_segment_stmt = select(DocumentSegment).where(
+                                DocumentSegment.dataset_id == dataset_document.dataset_id,
+                                DocumentSegment.enabled == True,
+                                DocumentSegment.status == "completed",
+                                DocumentSegment.index_node_id == index_node_id,
+                            )
+                            segment = db.session.scalar(document_segment_stmt)
 
-                    if not segment:
-                        continue
-
-                    include_segment_ids.add(segment.id)
-                    record = {
-                        "segment": segment,
-                        "score": document.metadata.get("score"),  # type: ignore
-                    }
-                    records.append(record)
+                        if not segment:
+                            continue
+                        if segment.id not in include_segment_ids:
+                            include_segment_ids.add(segment.id)
+                            record = {
+                                "segment": segment,
+                                "score": document.metadata.get("score"),  # type: ignore
+                            }
+                            if attachment_info:
+                                segment_file_map[segment.id] = [attachment_info]
+                            records.append(record)
+                        else:
+                            if attachment_info:
+                                attachment_infos = segment_file_map.get(segment.id, [])
+                                if attachment_info not in attachment_infos:
+                                    attachment_infos.append(attachment_info)
+                                segment_file_map[segment.id] = attachment_infos
 
             # Add child chunks information to records
             for record in records:
                 if record["segment"].id in segment_child_map:
                     record["child_chunks"] = segment_child_map[record["segment"].id].get("child_chunks")  # type: ignore
                     record["score"] = segment_child_map[record["segment"].id]["max_score"]
+                if record["segment"].id in segment_file_map:
+                    record["files"] = segment_file_map[record["segment"].id]  # type: ignore[assignment]
 
             result = []
             for record in records:
@@ -447,6 +534,11 @@ class RetrievalService:
                 if not isinstance(child_chunks, list):
                     child_chunks = None
 
+                # Extract files, ensuring it's a list or None
+                files = record.get("files")
+                if not isinstance(files, list):
+                    files = None
+
                 # Extract score, ensuring it's a float or None
                 score_value = record.get("score")
                 score = (
@@ -456,10 +548,149 @@ class RetrievalService:
                 )
 
                 # Create RetrievalSegments object
-                retrieval_segment = RetrievalSegments(segment=segment, child_chunks=child_chunks, score=score)
+                retrieval_segment = RetrievalSegments(
+                    segment=segment, child_chunks=child_chunks, score=score, files=files
+                )
                 result.append(retrieval_segment)
 
             return result
         except Exception as e:
             db.session.rollback()
             raise e
+
+    def _retrieve(
+        self,
+        flask_app: Flask,
+        retrieval_method: RetrievalMethod,
+        dataset: Dataset,
+        query: str | None = None,
+        top_k: int = 4,
+        score_threshold: float | None = 0.0,
+        reranking_model: dict | None = None,
+        reranking_mode: str = "reranking_model",
+        weights: dict | None = None,
+        document_ids_filter: list[str] | None = None,
+        attachment_id: str | None = None,
+        all_documents: list[Document] = [],
+        exceptions: list[str] = [],
+    ):
+        if not query and not attachment_id:
+            return
+        with flask_app.app_context():
+            all_documents_item: list[Document] = []
+            # Optimize multithreading with thread pools
+            with ThreadPoolExecutor(max_workers=dify_config.RETRIEVAL_SERVICE_EXECUTORS) as executor:  # type: ignore
+                futures = []
+                if retrieval_method == RetrievalMethod.KEYWORD_SEARCH and query:
+                    futures.append(
+                        executor.submit(
+                            self.keyword_search,
+                            flask_app=current_app._get_current_object(),  # type: ignore
+                            dataset_id=dataset.id,
+                            query=query,
+                            top_k=top_k,
+                            all_documents=all_documents_item,
+                            exceptions=exceptions,
+                            document_ids_filter=document_ids_filter,
+                        )
+                    )
+                if RetrievalMethod.is_support_semantic_search(retrieval_method):
+                    if query:
+                        futures.append(
+                            executor.submit(
+                                self.embedding_search,
+                                flask_app=current_app._get_current_object(),  # type: ignore
+                                dataset_id=dataset.id,
+                                query=query,
+                                top_k=top_k,
+                                score_threshold=score_threshold,
+                                reranking_model=reranking_model,
+                                all_documents=all_documents_item,
+                                retrieval_method=retrieval_method,
+                                exceptions=exceptions,
+                                document_ids_filter=document_ids_filter,
+                                query_type=QueryType.TEXT_QUERY,
+                            )
+                        )
+                    if attachment_id:
+                        futures.append(
+                            executor.submit(
+                                self.embedding_search,
+                                flask_app=current_app._get_current_object(),  # type: ignore
+                                dataset_id=dataset.id,
+                                query=attachment_id,
+                                top_k=top_k,
+                                score_threshold=score_threshold,
+                                reranking_model=reranking_model,
+                                all_documents=all_documents_item,
+                                retrieval_method=retrieval_method,
+                                exceptions=exceptions,
+                                document_ids_filter=document_ids_filter,
+                                query_type=QueryType.IMAGE_QUERY,
+                            )
+                        )
+                if RetrievalMethod.is_support_fulltext_search(retrieval_method) and query:
+                    futures.append(
+                        executor.submit(
+                            self.full_text_index_search,
+                            flask_app=current_app._get_current_object(),  # type: ignore
+                            dataset_id=dataset.id,
+                            query=query,
+                            top_k=top_k,
+                            score_threshold=score_threshold,
+                            reranking_model=reranking_model,
+                            all_documents=all_documents_item,
+                            retrieval_method=retrieval_method,
+                            exceptions=exceptions,
+                            document_ids_filter=document_ids_filter,
+                        )
+                    )
+                concurrent.futures.wait(futures, timeout=300, return_when=concurrent.futures.ALL_COMPLETED)
+
+            if exceptions:
+                raise ValueError(";\n".join(exceptions))
+
+            # Deduplicate documents for hybrid search to avoid duplicate chunks
+            if retrieval_method == RetrievalMethod.HYBRID_SEARCH:
+                if attachment_id and reranking_mode == RerankMode.WEIGHTED_SCORE:
+                    all_documents.extend(all_documents_item)
+                all_documents_item = self._deduplicate_documents(all_documents_item)
+                data_post_processor = DataPostProcessor(
+                    str(dataset.tenant_id), reranking_mode, reranking_model, weights, False
+                )
+
+                query = query or attachment_id
+                if not query:
+                    return
+                all_documents_item = data_post_processor.invoke(
+                    query=query,
+                    documents=all_documents_item,
+                    score_threshold=score_threshold,
+                    top_n=top_k,
+                    query_type=QueryType.TEXT_QUERY if query else QueryType.IMAGE_QUERY,
+                )
+
+            all_documents.extend(all_documents_item)
+
+    @classmethod
+    def get_segment_attachment_info(
+        cls, dataset_id: str, tenant_id: str, attachment_id: str, session: Session
+    ) -> dict[str, Any] | None:
+        upload_file = session.query(UploadFile).where(UploadFile.id == attachment_id).first()
+        if upload_file:
+            attachment_binding = (
+                session.query(SegmentAttachmentBinding)
+                .where(SegmentAttachmentBinding.attachment_id == upload_file.id)
+                .first()
+            )
+            if attachment_binding:
+                attchment_info = {
+                    "id": upload_file.id,
+                    "name": upload_file.name,
+                    "extension": "." + upload_file.extension,
+                    "mime_type": upload_file.mime_type,
+                    "source_url": sign_upload_file(upload_file.id, upload_file.extension),
+                    "size": upload_file.size,
+                }
+                return {"attchment_info": attchment_info, "segment_id": attachment_binding.segment_id}
+        return None

api/core/rag/datasource/vdb/vector_factory.py

@@ -1,3 +1,4 @@
+import base64
 import logging
 import time
 from abc import ABC, abstractmethod
@@ -12,10 +13,13 @@ from core.rag.datasource.vdb.vector_base import BaseVector
 from core.rag.datasource.vdb.vector_type import VectorType
 from core.rag.embedding.cached_embedding import CacheEmbedding
 from core.rag.embedding.embedding_base import Embeddings
+from core.rag.index_processor.constant.doc_type import DocType
 from core.rag.models.document import Document
 from extensions.ext_database import db
 from extensions.ext_redis import redis_client
+from extensions.ext_storage import storage
 from models.dataset import Dataset, Whitelist
+from models.model import UploadFile
 
 logger = logging.getLogger(__name__)
 
@@ -203,6 +207,47 @@ class Vector:
                 self._vector_processor.create(texts=batch, embeddings=batch_embeddings, **kwargs)
             logger.info("Embedding %s texts took %s s", len(texts), time.time() - start)
 
+    def create_multimodal(self, file_documents: list | None = None, **kwargs):
+        if file_documents:
+            start = time.time()
+            logger.info("start embedding %s files %s", len(file_documents), start)
+            batch_size = 1000
+            total_batches = len(file_documents) + batch_size - 1
+            for i in range(0, len(file_documents), batch_size):
+                batch = file_documents[i : i + batch_size]
+                batch_start = time.time()
+                logger.info("Processing batch %s/%s (%s files)", i // batch_size + 1, total_batches, len(batch))
+
+                # Batch query all upload files to avoid N+1 queries
+                attachment_ids = [doc.metadata["doc_id"] for doc in batch]
+                stmt = select(UploadFile).where(UploadFile.id.in_(attachment_ids))
+                upload_files = db.session.scalars(stmt).all()
+                upload_file_map = {str(f.id): f for f in upload_files}
+
+                file_base64_list = []
+                real_batch = []
+                for document in batch:
+                    attachment_id = document.metadata["doc_id"]
+                    doc_type = document.metadata["doc_type"]
+                    upload_file = upload_file_map.get(attachment_id)
+                    if upload_file:
+                        blob = storage.load_once(upload_file.key)
+                        file_base64_str = base64.b64encode(blob).decode()
+                        file_base64_list.append(
+                            {
+                                "content": file_base64_str,
+                                "content_type": doc_type,
+                                "file_id": attachment_id,
+                            }
+                        )
+                        real_batch.append(document)
+                batch_embeddings = self._embeddings.embed_multimodal_documents(file_base64_list)
+                logger.info(
+                    "Embedding batch %s/%s took %s s", i // batch_size + 1, total_batches, time.time() - batch_start
+                )
+                self._vector_processor.create(texts=real_batch, embeddings=batch_embeddings, **kwargs)
+            logger.info("Embedding %s files took %s s", len(file_documents), time.time() - start)
+
     def add_texts(self, documents: list[Document], **kwargs):
         if kwargs.get("duplicate_check", False):
             documents = self._filter_duplicate_texts(documents)
@@ -223,6 +268,22 @@ class Vector:
         query_vector = self._embeddings.embed_query(query)
         return self._vector_processor.search_by_vector(query_vector, **kwargs)
 
+    def search_by_file(self, file_id: str, **kwargs: Any) -> list[Document]:
+        upload_file: UploadFile | None = db.session.query(UploadFile).where(UploadFile.id == file_id).first()
+
+        if not upload_file:
+            return []
+        blob = storage.load_once(upload_file.key)
+        file_base64_str = base64.b64encode(blob).decode()
+        multimodal_vector = self._embeddings.embed_multimodal_query(
+            {
+                "content": file_base64_str,
+                "content_type": DocType.IMAGE,
+                "file_id": file_id,
+            }
+        )
+        return self._vector_processor.search_by_vector(multimodal_vector, **kwargs)
+
     def search_by_full_text(self, query: str, **kwargs: Any) -> list[Document]:
         return self._vector_processor.search_by_full_text(query, **kwargs)
 

api/core/rag/datasource/vdb/weaviate/weaviate_vector.py

@@ -79,6 +79,18 @@ class WeaviateVector(BaseVector):
         self._client = self._init_client(config)
         self._attributes = attributes
 
+    def __del__(self):
+        """
+        Destructor to properly close the Weaviate client connection.
+        Prevents connection leaks and resource warnings.
+        """
+        if hasattr(self, "_client") and self._client is not None:
+            try:
+                self._client.close()
+            except Exception as e:
+                # Ignore errors during cleanup as object is being destroyed
+                logger.warning("Error closing Weaviate client %s", e, exc_info=True)
+
     def _init_client(self, config: WeaviateConfig) -> weaviate.WeaviateClient:
         """
         Initializes and returns a connected Weaviate client.

api/core/rag/docstore/dataset_docstore.py

@@ -5,9 +5,9 @@ from sqlalchemy import func, select
 
 from core.model_manager import ModelManager
 from core.model_runtime.entities.model_entities import ModelType
-from core.rag.models.document import Document
+from core.rag.models.document import AttachmentDocument, Document
 from extensions.ext_database import db
-from models.dataset import ChildChunk, Dataset, DocumentSegment
+from models.dataset import ChildChunk, Dataset, DocumentSegment, SegmentAttachmentBinding
 
 
 class DatasetDocumentStore:
@@ -120,6 +120,9 @@ class DatasetDocumentStore:
 
                 db.session.add(segment_document)
                 db.session.flush()
+                self.add_multimodel_documents_binding(
+                    segment_id=segment_document.id, multimodel_documents=doc.attachments
+                )
                 if save_child:
                     if doc.children:
                         for position, child in enumerate(doc.children, start=1):
@@ -144,6 +147,9 @@ class DatasetDocumentStore:
                 segment_document.index_node_hash = doc.metadata.get("doc_hash")
                 segment_document.word_count = len(doc.page_content)
                 segment_document.tokens = tokens
+                self.add_multimodel_documents_binding(
+                    segment_id=segment_document.id, multimodel_documents=doc.attachments
+                )
                 if save_child and doc.children:
                     # delete the existing child chunks
                     db.session.query(ChildChunk).where(
@@ -233,3 +239,15 @@ class DatasetDocumentStore:
         document_segment = db.session.scalar(stmt)
 
         return document_segment
+
+    def add_multimodel_documents_binding(self, segment_id: str, multimodel_documents: list[AttachmentDocument] | None):
+        if multimodel_documents:
+            for multimodel_document in multimodel_documents:
+                binding = SegmentAttachmentBinding(
+                    tenant_id=self._dataset.tenant_id,
+                    dataset_id=self._dataset.id,
+                    document_id=self._document_id,
+                    segment_id=segment_id,
+                    attachment_id=multimodel_document.metadata["doc_id"],
+                )
+                db.session.add(binding)

api/core/rag/embedding/cached_embedding.py

@@ -104,6 +104,88 @@ class CacheEmbedding(Embeddings):
 
         return text_embeddings
 
+    def embed_multimodal_documents(self, multimodel_documents: list[dict]) -> list[list[float]]:
+        """Embed file documents."""
+        # use doc embedding cache or store if not exists
+        multimodel_embeddings: list[Any] = [None for _ in range(len(multimodel_documents))]
+        embedding_queue_indices = []
+        for i, multimodel_document in enumerate(multimodel_documents):
+            file_id = multimodel_document["file_id"]
+            embedding = (
+                db.session.query(Embedding)
+                .filter_by(
+                    model_name=self._model_instance.model, hash=file_id, provider_name=self._model_instance.provider
+                )
+                .first()
+            )
+            if embedding:
+                multimodel_embeddings[i] = embedding.get_embedding()
+            else:
+                embedding_queue_indices.append(i)
+
+        # NOTE: avoid closing the shared scoped session here; downstream code may still have pending work
+
+        if embedding_queue_indices:
+            embedding_queue_multimodel_documents = [multimodel_documents[i] for i in embedding_queue_indices]
+            embedding_queue_embeddings = []
+            try:
+                model_type_instance = cast(TextEmbeddingModel, self._model_instance.model_type_instance)
+                model_schema = model_type_instance.get_model_schema(
+                    self._model_instance.model, self._model_instance.credentials
+                )
+                max_chunks = (
+                    model_schema.model_properties[ModelPropertyKey.MAX_CHUNKS]
+                    if model_schema and ModelPropertyKey.MAX_CHUNKS in model_schema.model_properties
+                    else 1
+                )
+                for i in range(0, len(embedding_queue_multimodel_documents), max_chunks):
+                    batch_multimodel_documents = embedding_queue_multimodel_documents[i : i + max_chunks]
+
+                    embedding_result = self._model_instance.invoke_multimodal_embedding(
+                        multimodel_documents=batch_multimodel_documents,
+                        user=self._user,
+                        input_type=EmbeddingInputType.DOCUMENT,
+                    )
+
+                    for vector in embedding_result.embeddings:
+                        try:
+                            # FIXME: type ignore for numpy here
+                            normalized_embedding = (vector / np.linalg.norm(vector)).tolist()  # type: ignore
+                            # stackoverflow best way: https://stackoverflow.com/questions/20319813/how-to-check-list-containing-nan
+                            if np.isnan(normalized_embedding).any():
+                                # for issue #11827  float values are not json compliant
+                                logger.warning("Normalized embedding is nan: %s", normalized_embedding)
+                                continue
+                            embedding_queue_embeddings.append(normalized_embedding)
+                        except IntegrityError:
+                            db.session.rollback()
+                        except Exception:
+                            logger.exception("Failed transform embedding")
+                cache_embeddings = []
+                try:
+                    for i, n_embedding in zip(embedding_queue_indices, embedding_queue_embeddings):
+                        multimodel_embeddings[i] = n_embedding
+                        file_id = multimodel_documents[i]["file_id"]
+                        if file_id not in cache_embeddings:
+                            embedding_cache = Embedding(
+                                model_name=self._model_instance.model,
+                                hash=file_id,
+                                provider_name=self._model_instance.provider,
+                                embedding=pickle.dumps(n_embedding, protocol=pickle.HIGHEST_PROTOCOL),
+                            )
+                            embedding_cache.set_embedding(n_embedding)
+                            db.session.add(embedding_cache)
+                            cache_embeddings.append(file_id)
+                    db.session.commit()
+                except IntegrityError:
+                    db.session.rollback()
+            except Exception as ex:
+                db.session.rollback()
+                logger.exception("Failed to embed documents")
+                raise ex
+
+        return multimodel_embeddings
+
     def embed_query(self, text: str) -> list[float]:
         """Embed query text."""
         # use doc embedding cache or store if not exists
@@ -146,3 +228,46 @@ class CacheEmbedding(Embeddings):
             raise ex
 
         return embedding_results  # type: ignore
+
+    def embed_multimodal_query(self, multimodel_document: dict) -> list[float]:
+        """Embed multimodal documents."""
+        # use doc embedding cache or store if not exists
+        file_id = multimodel_document["file_id"]
+        embedding_cache_key = f"{self._model_instance.provider}_{self._model_instance.model}_{file_id}"
+        embedding = redis_client.get(embedding_cache_key)
+        if embedding:
+            redis_client.expire(embedding_cache_key, 600)
+            decoded_embedding = np.frombuffer(base64.b64decode(embedding), dtype="float")
+            return [float(x) for x in decoded_embedding]
+        try:
+            embedding_result = self._model_instance.invoke_multimodal_embedding(
+                multimodel_documents=[multimodel_document], user=self._user, input_type=EmbeddingInputType.QUERY
+            )
+
+            embedding_results = embedding_result.embeddings[0]
+            # FIXME: type ignore for numpy here
+            embedding_results = (embedding_results / np.linalg.norm(embedding_results)).tolist()  # type: ignore
+            if np.isnan(embedding_results).any():
+                raise ValueError("Normalized embedding is nan please try again")
+        except Exception as ex:
+            if dify_config.DEBUG:
+                logger.exception("Failed to embed multimodal document '%s'", multimodel_document["file_id"])
+            raise ex
+
+        try:
+            # encode embedding to base64
+            embedding_vector = np.array(embedding_results)
+            vector_bytes = embedding_vector.tobytes()
+            # Transform to Base64
+            encoded_vector = base64.b64encode(vector_bytes)
+            # Transform to string
+            encoded_str = encoded_vector.decode("utf-8")
+            redis_client.setex(embedding_cache_key, 600, encoded_str)
+        except Exception as ex:
+            if dify_config.DEBUG:
+                logger.exception(
+                    "Failed to add embedding to redis for the multimodal document '%s'", multimodel_document["file_id"]
+                )
+            raise ex
+
+        return embedding_results  # type: ignore

api/core/rag/embedding/embedding_base.py

@@ -9,11 +9,21 @@ class Embeddings(ABC):
         """Embed search docs."""
         raise NotImplementedError
 
+    @abstractmethod
+    def embed_multimodal_documents(self, multimodel_documents: list[dict]) -> list[list[float]]:
+        """Embed file documents."""
+        raise NotImplementedError
+
     @abstractmethod
     def embed_query(self, text: str) -> list[float]:
         """Embed query text."""
         raise NotImplementedError
 
+    @abstractmethod
+    def embed_multimodal_query(self, multimodel_document: dict) -> list[float]:
+        """Embed multimodal query."""
+        raise NotImplementedError
+
     async def aembed_documents(self, texts: list[str]) -> list[list[float]]:
         """Asynchronous Embed search docs."""
         raise NotImplementedError

api/core/rag/embedding/retrieval.py

@@ -19,3 +19,4 @@ class RetrievalSegments(BaseModel):
     segment: DocumentSegment
     child_chunks: list[RetrievalChildChunk] | None = None
     score: float | None = None
+    files: list[dict[str, str | int]] | None = None

api/core/rag/entities/citation_metadata.py

@@ -21,3 +21,4 @@ class RetrievalSourceMetadata(BaseModel):
     page: int | None = None
     doc_metadata: dict[str, Any] | None = None
     title: str | None = None
+    files: list[dict[str, Any]] | None = None

api/core/rag/index_processor/constant/doc_type.py

@@ -0,0 +1,6 @@
+from enum import StrEnum
+
+
+class DocType(StrEnum):
+    TEXT = "text"
+    IMAGE = "image"

api/core/rag/index_processor/constant/index_type.py

@@ -1,7 +1,12 @@
 from enum import StrEnum
 
 
-class IndexType(StrEnum):
+class IndexStructureType(StrEnum):
     PARAGRAPH_INDEX = "text_model"
     QA_INDEX = "qa_model"
     PARENT_CHILD_INDEX = "hierarchical_model"
+
+
+class IndexTechniqueType(StrEnum):
+    ECONOMY = "economy"
+    HIGH_QUALITY = "high_quality"

api/core/rag/index_processor/constant/query_type.py

@@ -0,0 +1,6 @@
+from enum import StrEnum
+
+
+class QueryType(StrEnum):
+    TEXT_QUERY = "text_query"
+    IMAGE_QUERY = "image_query"

api/core/rag/index_processor/index_processor_base.py

@@ -1,20 +1,34 @@
 """Abstract interface for document loader implementations."""
 
+import cgi
+import logging
+import mimetypes
+import os
+import re
 from abc import ABC, abstractmethod
 from collections.abc import Mapping
 from typing import TYPE_CHECKING, Any, Optional
+from urllib.parse import unquote, urlparse
+
+import httpx
 
 from configs import dify_config
+from core.helper import ssrf_proxy
 from core.rag.extractor.entity.extract_setting import ExtractSetting
-from core.rag.models.document import Document
+from core.rag.index_processor.constant.doc_type import DocType
+from core.rag.models.document import AttachmentDocument, Document
 from core.rag.retrieval.retrieval_methods import RetrievalMethod
 from core.rag.splitter.fixed_text_splitter import (
     EnhanceRecursiveCharacterTextSplitter,
     FixedRecursiveCharacterTextSplitter,
 )
 from core.rag.splitter.text_splitter import TextSplitter
+from extensions.ext_database import db
+from extensions.ext_storage import storage
+from models import Account, ToolFile
 from models.dataset import Dataset, DatasetProcessRule
 from models.dataset import Document as DatasetDocument
+from models.model import UploadFile
 
 if TYPE_CHECKING:
     from core.model_manager import ModelInstance
@@ -28,11 +42,18 @@ class BaseIndexProcessor(ABC):
         raise NotImplementedError
 
     @abstractmethod
-    def transform(self, documents: list[Document], **kwargs) -> list[Document]:
+    def transform(self, documents: list[Document], current_user: Account | None = None, **kwargs) -> list[Document]:
         raise NotImplementedError
 
     @abstractmethod
-    def load(self, dataset: Dataset, documents: list[Document], with_keywords: bool = True, **kwargs):
+    def load(
+        self,
+        dataset: Dataset,
+        documents: list[Document],
+        multimodal_documents: list[AttachmentDocument] | None = None,
+        with_keywords: bool = True,
+        **kwargs,
+    ):
         raise NotImplementedError
 
     @abstractmethod
@@ -96,3 +117,178 @@ class BaseIndexProcessor(ABC):
             )
 
         return character_splitter  # type: ignore
+
+    def _get_content_files(self, document: Document, current_user: Account | None = None) -> list[AttachmentDocument]:
+        """
+        Get the content files from the document.
+        """
+        multi_model_documents: list[AttachmentDocument] = []
+        text = document.page_content
+        images = self._extract_markdown_images(text)
+        if not images:
+            return multi_model_documents
+        upload_file_id_list = []
+
+        for image in images:
+            # Collect all upload_file_ids including duplicates to preserve occurrence count
+
+            # For data before v0.10.0
+            pattern = r"/files/([a-f0-9\-]+)/image-preview(?:\?.*?)?"
+            match = re.search(pattern, image)
+            if match:
+                upload_file_id = match.group(1)
+                upload_file_id_list.append(upload_file_id)
+                continue
+
+            # For data after v0.10.0
+            pattern = r"/files/([a-f0-9\-]+)/file-preview(?:\?.*?)?"
+            match = re.search(pattern, image)
+            if match:
+                upload_file_id = match.group(1)
+                upload_file_id_list.append(upload_file_id)
+                continue
+
+            # For tools directory - direct file formats (e.g., .png, .jpg, etc.)
+            # Match URL including any query parameters up to common URL boundaries (space, parenthesis, quotes)
+            pattern = r"/files/tools/([a-f0-9\-]+)\.([a-zA-Z0-9]+)(?:\?[^\s\)\"\']*)?"
+            match = re.search(pattern, image)
+            if match:
+                if current_user:
+                    tool_file_id = match.group(1)
+                    upload_file_id = self._download_tool_file(tool_file_id, current_user)
+                    if upload_file_id:
+                        upload_file_id_list.append(upload_file_id)
+                continue
+            if current_user:
+                upload_file_id = self._download_image(image.split(" ")[0], current_user)
+                if upload_file_id:
+                    upload_file_id_list.append(upload_file_id)
+
+        if not upload_file_id_list:
+            return multi_model_documents
+
+        # Get unique IDs for database query
+        unique_upload_file_ids = list(set(upload_file_id_list))
+        upload_files = db.session.query(UploadFile).where(UploadFile.id.in_(unique_upload_file_ids)).all()
+
+        # Create a mapping from ID to UploadFile for quick lookup
+        upload_file_map = {upload_file.id: upload_file for upload_file in upload_files}
+
+        # Create a Document for each occurrence (including duplicates)
+        for upload_file_id in upload_file_id_list:
+            upload_file = upload_file_map.get(upload_file_id)
+            if upload_file:
+                multi_model_documents.append(
+                    AttachmentDocument(
+                        page_content=upload_file.name,
+                        metadata={
+                            "doc_id": upload_file.id,
+                            "doc_hash": "",
+                            "document_id": document.metadata.get("document_id"),
+                            "dataset_id": document.metadata.get("dataset_id"),
+                            "doc_type": DocType.IMAGE,
+                        },
+                    )
+                )
+        return multi_model_documents
+
+    def _extract_markdown_images(self, text: str) -> list[str]:
+        """
+        Extract the markdown images from the text.
+        """
+        pattern = r"!\[.*?\]\((.*?)\)"
+        return re.findall(pattern, text)
+
+    def _download_image(self, image_url: str, current_user: Account) -> str | None:
+        """
+        Download the image from the URL.
+        Image size must not exceed 2MB.
+        """
+        from services.file_service import FileService
+
+        MAX_IMAGE_SIZE = dify_config.ATTACHMENT_IMAGE_FILE_SIZE_LIMIT * 1024 * 1024
+        DOWNLOAD_TIMEOUT = dify_config.ATTACHMENT_IMAGE_DOWNLOAD_TIMEOUT
+
+        try:
+            # Download with timeout
+            response = ssrf_proxy.get(image_url, timeout=DOWNLOAD_TIMEOUT)
+            response.raise_for_status()
+
+            # Check Content-Length header if available
+            content_length = response.headers.get("Content-Length")
+            if content_length and int(content_length) > MAX_IMAGE_SIZE:
+                logging.warning("Image from %s exceeds 2MB limit (size: %s bytes)", image_url, content_length)
+                return None
+
+            filename = None
+
+            content_disposition = response.headers.get("content-disposition")
+            if content_disposition:
+                _, params = cgi.parse_header(content_disposition)
+                if "filename" in params:
+                    filename = params["filename"]
+                    filename = unquote(filename)
+
+            if not filename:
+                parsed_url = urlparse(image_url)
+                # unquote 处理 URL 中的中文
+                path = unquote(parsed_url.path)
+                filename = os.path.basename(path)
+
+            if not filename:
+                filename = "downloaded_image_file"
+
+            name, current_ext = os.path.splitext(filename)
+
+            content_type = response.headers.get("content-type", "").split(";")[0].strip()
+
+            real_ext = mimetypes.guess_extension(content_type)
+
+            if not current_ext and real_ext or current_ext in [".php", ".jsp", ".asp", ".html"] and real_ext:
+                filename = f"{name}{real_ext}"
+            # Download content with size limit
+            blob = b""
+            for chunk in response.iter_bytes(chunk_size=8192):
+                blob += chunk
+                if len(blob) > MAX_IMAGE_SIZE:
+                    logging.warning("Image from %s exceeds 2MB limit during download", image_url)
+                    return None
+
+            if not blob:
+                logging.warning("Image from %s is empty", image_url)
+                return None
+
+            upload_file = FileService(db.engine).upload_file(
+                filename=filename,
+                content=blob,
+                mimetype=content_type,
+                user=current_user,
+            )
+            return upload_file.id
+        except httpx.TimeoutException:
+            logging.warning("Timeout downloading image from %s after %s seconds", image_url, DOWNLOAD_TIMEOUT)
+            return None
+        except httpx.RequestError as e:
+            logging.warning("Error downloading image from %s: %s", image_url, str(e))
+            return None
+        except Exception:
+            logging.exception("Unexpected error downloading image from %s", image_url)
+            return None
+
+    def _download_tool_file(self, tool_file_id: str, current_user: Account) -> str | None:
+        """
+        Download the tool file from the ID.
+        """
+        from services.file_service import FileService
+
+        tool_file = db.session.query(ToolFile).where(ToolFile.id == tool_file_id).first()
+        if not tool_file:
+            return None
+        blob = storage.load_once(tool_file.file_key)
+        upload_file = FileService(db.engine).upload_file(
+            filename=tool_file.name,
+            content=blob,
+            mimetype=tool_file.mimetype,
+            user=current_user,
+        )
+        return upload_file.id

api/core/rag/index_processor/index_processor_factory.py

@@ -1,6 +1,6 @@
 """Abstract interface for document loader implementations."""
 
-from core.rag.index_processor.constant.index_type import IndexType
+from core.rag.index_processor.constant.index_type import IndexStructureType
 from core.rag.index_processor.index_processor_base import BaseIndexProcessor
 from core.rag.index_processor.processor.paragraph_index_processor import ParagraphIndexProcessor
 from core.rag.index_processor.processor.parent_child_index_processor import ParentChildIndexProcessor
@@ -19,11 +19,11 @@ class IndexProcessorFactory:
         if not self._index_type:
             raise ValueError("Index type must be specified.")
 
-        if self._index_type == IndexType.PARAGRAPH_INDEX:
+        if self._index_type == IndexStructureType.PARAGRAPH_INDEX:
             return ParagraphIndexProcessor()
-        elif self._index_type == IndexType.QA_INDEX:
+        elif self._index_type == IndexStructureType.QA_INDEX:
             return QAIndexProcessor()
-        elif self._index_type == IndexType.PARENT_CHILD_INDEX:
+        elif self._index_type == IndexStructureType.PARENT_CHILD_INDEX:
             return ParentChildIndexProcessor()
         else:
             raise ValueError(f"Index type {self._index_type} is not supported.")

api/core/rag/index_processor/processor/paragraph_index_processor.py

@@ -11,14 +11,17 @@ from core.rag.datasource.vdb.vector_factory import Vector
 from core.rag.docstore.dataset_docstore import DatasetDocumentStore
 from core.rag.extractor.entity.extract_setting import ExtractSetting
 from core.rag.extractor.extract_processor import ExtractProcessor
-from core.rag.index_processor.constant.index_type import IndexType
+from core.rag.index_processor.constant.doc_type import DocType
+from core.rag.index_processor.constant.index_type import IndexStructureType
 from core.rag.index_processor.index_processor_base import BaseIndexProcessor
-from core.rag.models.document import Document
+from core.rag.models.document import AttachmentDocument, Document, MultimodalGeneralStructureChunk
 from core.rag.retrieval.retrieval_methods import RetrievalMethod
 from core.tools.utils.text_processing_utils import remove_leading_symbols
 from libs import helper
+from models.account import Account
 from models.dataset import Dataset, DatasetProcessRule
 from models.dataset import Document as DatasetDocument
+from services.account_service import AccountService
 from services.entities.knowledge_entities.knowledge_entities import Rule
 
 
@@ -33,7 +36,7 @@ class ParagraphIndexProcessor(BaseIndexProcessor):
 
         return text_docs
 
-    def transform(self, documents: list[Document], **kwargs) -> list[Document]:
+    def transform(self, documents: list[Document], current_user: Account | None = None, **kwargs) -> list[Document]:
         process_rule = kwargs.get("process_rule")
         if not process_rule:
             raise ValueError("No process rule found.")
@@ -69,6 +72,11 @@ class ParagraphIndexProcessor(BaseIndexProcessor):
                     if document_node.metadata is not None:
                         document_node.metadata["doc_id"] = doc_id
                         document_node.metadata["doc_hash"] = hash
+                    multimodal_documents = (
+                        self._get_content_files(document_node, current_user) if document_node.metadata else None
+                    )
+                    if multimodal_documents:
+                        document_node.attachments = multimodal_documents
                     # delete Splitter character
                     page_content = remove_leading_symbols(document_node.page_content).strip()
                     if len(page_content) > 0:
@@ -77,10 +85,19 @@ class ParagraphIndexProcessor(BaseIndexProcessor):
             all_documents.extend(split_documents)
         return all_documents
 
-    def load(self, dataset: Dataset, documents: list[Document], with_keywords: bool = True, **kwargs):
+    def load(
+        self,
+        dataset: Dataset,
+        documents: list[Document],
+        multimodal_documents: list[AttachmentDocument] | None = None,
+        with_keywords: bool = True,
+        **kwargs,
+    ):
         if dataset.indexing_technique == "high_quality":
             vector = Vector(dataset)
             vector.create(documents)
+            if multimodal_documents and dataset.is_multimodal:
+                vector.create_multimodal(multimodal_documents)
             with_keywords = False
         if with_keywords:
             keywords_list = kwargs.get("keywords_list")
@@ -134,8 +151,9 @@ class ParagraphIndexProcessor(BaseIndexProcessor):
         return docs
 
     def index(self, dataset: Dataset, document: DatasetDocument, chunks: Any):
+        documents: list[Any] = []
+        all_multimodal_documents: list[Any] = []
         if isinstance(chunks, list):
-            documents = []
             for content in chunks:
                 metadata = {
                     "dataset_id": dataset.id,
@@ -144,26 +162,68 @@ class ParagraphIndexProcessor(BaseIndexProcessor):
                     "doc_hash": helper.generate_text_hash(content),
                 }
                 doc = Document(page_content=content, metadata=metadata)
+                attachments = self._get_content_files(doc)
+                if attachments:
+                    doc.attachments = attachments
+                    all_multimodal_documents.extend(attachments)
                 documents.append(doc)
-            if documents:
-                # save node to document segment
-                doc_store = DatasetDocumentStore(dataset=dataset, user_id=document.created_by, document_id=document.id)
-                # add document segments
-                doc_store.add_documents(docs=documents, save_child=False)
-                if dataset.indexing_technique == "high_quality":
-                    vector = Vector(dataset)
-                    vector.create(documents)
-                elif dataset.indexing_technique == "economy":
-                    keyword = Keyword(dataset)
-                    keyword.add_texts(documents)
         else:
-            raise ValueError("Chunks is not a list")
+            multimodal_general_structure = MultimodalGeneralStructureChunk.model_validate(chunks)
+            for general_chunk in multimodal_general_structure.general_chunks:
+                metadata = {
+                    "dataset_id": dataset.id,
+                    "document_id": document.id,
+                    "doc_id": str(uuid.uuid4()),
+                    "doc_hash": helper.generate_text_hash(general_chunk.content),
+                }
+                doc = Document(page_content=general_chunk.content, metadata=metadata)
+                if general_chunk.files:
+                    attachments = []
+                    for file in general_chunk.files:
+                        file_metadata = {
+                            "doc_id": file.id,
+                            "doc_hash": "",
+                            "document_id": document.id,
+                            "dataset_id": dataset.id,
+                            "doc_type": DocType.IMAGE,
+                        }
+                        file_document = AttachmentDocument(
+                            page_content=file.filename or "image_file", metadata=file_metadata
+                        )
+                        attachments.append(file_document)
+                        all_multimodal_documents.append(file_document)
+                    doc.attachments = attachments
+                else:
+                    account = AccountService.load_user(document.created_by)
+                    if not account:
+                        raise ValueError("Invalid account")
+                    doc.attachments = self._get_content_files(doc, current_user=account)
+                    if doc.attachments:
+                        all_multimodal_documents.extend(doc.attachments)
+                documents.append(doc)
+        if documents:
+            # save node to document segment
+            doc_store = DatasetDocumentStore(dataset=dataset, user_id=document.created_by, document_id=document.id)
+            # add document segments
+            doc_store.add_documents(docs=documents, save_child=False)
+            if dataset.indexing_technique == "high_quality":
+                vector = Vector(dataset)
+                vector.create(documents)
+                if all_multimodal_documents:
+                    vector.create_multimodal(all_multimodal_documents)
+            elif dataset.indexing_technique == "economy":
+                keyword = Keyword(dataset)
+                keyword.add_texts(documents)
 
     def format_preview(self, chunks: Any) -> Mapping[str, Any]:
         if isinstance(chunks, list):
             preview = []
             for content in chunks:
                 preview.append({"content": content})
-            return {"chunk_structure": IndexType.PARAGRAPH_INDEX, "preview": preview, "total_segments": len(chunks)}
+            return {
+                "chunk_structure": IndexStructureType.PARAGRAPH_INDEX,
+                "preview": preview,
+                "total_segments": len(chunks),
+            }
         else:
             raise ValueError("Chunks is not a list")

api/core/rag/index_processor/processor/parent_child_index_processor.py

@@ -13,14 +13,17 @@ from core.rag.datasource.vdb.vector_factory import Vector
 from core.rag.docstore.dataset_docstore import DatasetDocumentStore
 from core.rag.extractor.entity.extract_setting import ExtractSetting
 from core.rag.extractor.extract_processor import ExtractProcessor
-from core.rag.index_processor.constant.index_type import IndexType
+from core.rag.index_processor.constant.doc_type import DocType
+from core.rag.index_processor.constant.index_type import IndexStructureType
 from core.rag.index_processor.index_processor_base import BaseIndexProcessor
-from core.rag.models.document import ChildDocument, Document, ParentChildStructureChunk
+from core.rag.models.document import AttachmentDocument, ChildDocument, Document, ParentChildStructureChunk
 from core.rag.retrieval.retrieval_methods import RetrievalMethod
 from extensions.ext_database import db
 from libs import helper
+from models import Account
 from models.dataset import ChildChunk, Dataset, DatasetProcessRule, DocumentSegment
 from models.dataset import Document as DatasetDocument
+from services.account_service import AccountService
 from services.entities.knowledge_entities.knowledge_entities import ParentMode, Rule
 
 
@@ -35,7 +38,7 @@ class ParentChildIndexProcessor(BaseIndexProcessor):
 
         return text_docs
 
-    def transform(self, documents: list[Document], **kwargs) -> list[Document]:
+    def transform(self, documents: list[Document], current_user: Account | None = None, **kwargs) -> list[Document]:
         process_rule = kwargs.get("process_rule")
         if not process_rule:
             raise ValueError("No process rule found.")
@@ -77,6 +80,9 @@ class ParentChildIndexProcessor(BaseIndexProcessor):
                             page_content = page_content
                         if len(page_content) > 0:
                             document_node.page_content = page_content
+                            multimodel_documents = self._get_content_files(document_node, current_user)
+                            if multimodel_documents:
+                                document_node.attachments = multimodel_documents
                             # parse document to child nodes
                             child_nodes = self._split_child_nodes(
                                 document_node, rules, process_rule.get("mode"), kwargs.get("embedding_model_instance")
@@ -87,6 +93,9 @@ class ParentChildIndexProcessor(BaseIndexProcessor):
         elif rules.parent_mode == ParentMode.FULL_DOC:
             page_content = "\n".join([document.page_content for document in documents])
             document = Document(page_content=page_content, metadata=documents[0].metadata)
+            multimodel_documents = self._get_content_files(document)
+            if multimodel_documents:
+                document.attachments = multimodel_documents
             # parse document to child nodes
             child_nodes = self._split_child_nodes(
                 document, rules, process_rule.get("mode"), kwargs.get("embedding_model_instance")
@@ -104,7 +113,14 @@ class ParentChildIndexProcessor(BaseIndexProcessor):
 
         return all_documents
 
-    def load(self, dataset: Dataset, documents: list[Document], with_keywords: bool = True, **kwargs):
+    def load(
+        self,
+        dataset: Dataset,
+        documents: list[Document],
+        multimodal_documents: list[AttachmentDocument] | None = None,
+        with_keywords: bool = True,
+        **kwargs,
+    ):
         if dataset.indexing_technique == "high_quality":
             vector = Vector(dataset)
             for document in documents:
@@ -114,6 +130,8 @@ class ParentChildIndexProcessor(BaseIndexProcessor):
                         Document.model_validate(child_document.model_dump()) for child_document in child_documents
                     ]
                     vector.create(formatted_child_documents)
+            if multimodal_documents and dataset.is_multimodal:
+                vector.create_multimodal(multimodal_documents)
 
     def clean(self, dataset: Dataset, node_ids: list[str] | None, with_keywords: bool = True, **kwargs):
         # node_ids is segment's node_ids
@@ -244,6 +262,24 @@ class ParentChildIndexProcessor(BaseIndexProcessor):
                 }
                 child_documents.append(ChildDocument(page_content=child, metadata=child_metadata))
             doc = Document(page_content=parent_child.parent_content, metadata=metadata, children=child_documents)
+            if parent_child.files and len(parent_child.files) > 0:
+                attachments = []
+                for file in parent_child.files:
+                    file_metadata = {
+                        "doc_id": file.id,
+                        "doc_hash": "",
+                        "document_id": document.id,
+                        "dataset_id": dataset.id,
+                        "doc_type": DocType.IMAGE,
+                    }
+                    file_document = AttachmentDocument(page_content=file.filename or "", metadata=file_metadata)
+                    attachments.append(file_document)
+                doc.attachments = attachments
+            else:
+                account = AccountService.load_user(document.created_by)
+                if not account:
+                    raise ValueError("Invalid account")
+                doc.attachments = self._get_content_files(doc, current_user=account)
             documents.append(doc)
         if documents:
             # update document parent mode
@@ -267,12 +303,17 @@ class ParentChildIndexProcessor(BaseIndexProcessor):
             doc_store.add_documents(docs=documents, save_child=True)
             if dataset.indexing_technique == "high_quality":
                 all_child_documents = []
+                all_multimodal_documents = []
                 for doc in documents:
                     if doc.children:
                         all_child_documents.extend(doc.children)
+                    if doc.attachments:
+                        all_multimodal_documents.extend(doc.attachments)
+                vector = Vector(dataset)
                 if all_child_documents:
-                    vector = Vector(dataset)
                     vector.create(all_child_documents)
+                if all_multimodal_documents:
+                    vector.create_multimodal(all_multimodal_documents)
 
     def format_preview(self, chunks: Any) -> Mapping[str, Any]:
         parent_childs = ParentChildStructureChunk.model_validate(chunks)
@@ -280,7 +321,7 @@ class ParentChildIndexProcessor(BaseIndexProcessor):
         for parent_child in parent_childs.parent_child_chunks:
             preview.append({"content": parent_child.parent_content, "child_chunks": parent_child.child_contents})
         return {
-            "chunk_structure": IndexType.PARENT_CHILD_INDEX,
+            "chunk_structure": IndexStructureType.PARENT_CHILD_INDEX,
             "parent_mode": parent_childs.parent_mode,
             "preview": preview,
             "total_segments": len(parent_childs.parent_child_chunks),

api/core/rag/index_processor/processor/qa_index_processor.py

@@ -18,12 +18,13 @@ from core.rag.datasource.vdb.vector_factory import Vector
 from core.rag.docstore.dataset_docstore import DatasetDocumentStore
 from core.rag.extractor.entity.extract_setting import ExtractSetting
 from core.rag.extractor.extract_processor import ExtractProcessor
-from core.rag.index_processor.constant.index_type import IndexType
+from core.rag.index_processor.constant.index_type import IndexStructureType
 from core.rag.index_processor.index_processor_base import BaseIndexProcessor
-from core.rag.models.document import Document, QAStructureChunk
+from core.rag.models.document import AttachmentDocument, Document, QAStructureChunk
 from core.rag.retrieval.retrieval_methods import RetrievalMethod
 from core.tools.utils.text_processing_utils import remove_leading_symbols
 from libs import helper
+from models.account import Account
 from models.dataset import Dataset
 from models.dataset import Document as DatasetDocument
 from services.entities.knowledge_entities.knowledge_entities import Rule
@@ -41,7 +42,7 @@ class QAIndexProcessor(BaseIndexProcessor):
         )
         return text_docs
 
-    def transform(self, documents: list[Document], **kwargs) -> list[Document]:
+    def transform(self, documents: list[Document], current_user: Account | None = None, **kwargs) -> list[Document]:
         preview = kwargs.get("preview")
         process_rule = kwargs.get("process_rule")
         if not process_rule:
@@ -116,7 +117,7 @@ class QAIndexProcessor(BaseIndexProcessor):
 
         try:
             # Skip the first row
-            df = pd.read_csv(file)
+            df = pd.read_csv(file)  # type: ignore
             text_docs = []
             for _, row in df.iterrows():
                 data = Document(page_content=row.iloc[0], metadata={"answer": row.iloc[1]})
@@ -128,10 +129,19 @@ class QAIndexProcessor(BaseIndexProcessor):
             raise ValueError(str(e))
         return text_docs
 
-    def load(self, dataset: Dataset, documents: list[Document], with_keywords: bool = True, **kwargs):
+    def load(
+        self,
+        dataset: Dataset,
+        documents: list[Document],
+        multimodal_documents: list[AttachmentDocument] | None = None,
+        with_keywords: bool = True,
+        **kwargs,
+    ):
         if dataset.indexing_technique == "high_quality":
             vector = Vector(dataset)
             vector.create(documents)
+            if multimodal_documents and dataset.is_multimodal:
+                vector.create_multimodal(multimodal_documents)
 
     def clean(self, dataset: Dataset, node_ids: list[str] | None, with_keywords: bool = True, **kwargs):
         vector = Vector(dataset)
@@ -197,7 +207,7 @@ class QAIndexProcessor(BaseIndexProcessor):
         for qa_chunk in qa_chunks.qa_chunks:
             preview.append({"question": qa_chunk.question, "answer": qa_chunk.answer})
         return {
-            "chunk_structure": IndexType.QA_INDEX,
+            "chunk_structure": IndexStructureType.QA_INDEX,
             "qa_preview": preview,
             "total_segments": len(qa_chunks.qa_chunks),
         }

api/core/rag/models/document.py

@@ -4,6 +4,8 @@ from typing import Any
 
 from pydantic import BaseModel, Field
 
+from core.file import File
+
 
 class ChildDocument(BaseModel):
     """Class for storing a piece of text and associated metadata."""
@@ -15,7 +17,19 @@ class ChildDocument(BaseModel):
     """Arbitrary metadata about the page content (e.g., source, relationships to other
         documents, etc.).
     """
-    metadata: dict = Field(default_factory=dict)
+    metadata: dict[str, Any] = Field(default_factory=dict)
+
+
+class AttachmentDocument(BaseModel):
+    """Class for storing a piece of text and associated metadata."""
+
+    page_content: str
+
+    provider: str | None = "dify"
+
+    vector: list[float] | None = None
+
+    metadata: dict[str, Any] = Field(default_factory=dict)
 
 
 class Document(BaseModel):
@@ -28,12 +42,31 @@ class Document(BaseModel):
     """Arbitrary metadata about the page content (e.g., source, relationships to other
         documents, etc.).
     """
-    metadata: dict = Field(default_factory=dict)
+    metadata: dict[str, Any] = Field(default_factory=dict)
 
     provider: str | None = "dify"
 
     children: list[ChildDocument] | None = None
 
+    attachments: list[AttachmentDocument] | None = None
+
+
+class GeneralChunk(BaseModel):
+    """
+    General Chunk.
+    """
+
+    content: str
+    files: list[File] | None = None
+
+
+class MultimodalGeneralStructureChunk(BaseModel):
+    """
+    Multimodal General Structure Chunk.
+    """
+
+    general_chunks: list[GeneralChunk]
+
 
 class GeneralStructureChunk(BaseModel):
     """
@@ -50,6 +83,7 @@ class ParentChildChunk(BaseModel):
 
     parent_content: str
     child_contents: list[str]
+    files: list[File] | None = None
 
 
 class ParentChildStructureChunk(BaseModel):

api/core/rag/rerank/rerank_base.py

@@ -1,5 +1,6 @@
 from abc import ABC, abstractmethod
 
+from core.rag.index_processor.constant.query_type import QueryType
 from core.rag.models.document import Document
 
 
@@ -12,6 +13,7 @@ class BaseRerankRunner(ABC):
         score_threshold: float | None = None,
         top_n: int | None = None,
         user: str | None = None,
+        query_type: QueryType = QueryType.TEXT_QUERY,
     ) -> list[Document]:
         """
         Run rerank model

api/core/rag/rerank/rerank_model.py

@@ -1,6 +1,15 @@
-from core.model_manager import ModelInstance
+import base64
+
+from core.model_manager import ModelInstance, ModelManager
+from core.model_runtime.entities.model_entities import ModelType
+from core.model_runtime.entities.rerank_entities import RerankResult
+from core.rag.index_processor.constant.doc_type import DocType
+from core.rag.index_processor.constant.query_type import QueryType
 from core.rag.models.document import Document
 from core.rag.rerank.rerank_base import BaseRerankRunner
+from extensions.ext_database import db
+from extensions.ext_storage import storage
+from models.model import UploadFile
 
 
 class RerankModelRunner(BaseRerankRunner):
@@ -14,6 +23,7 @@ class RerankModelRunner(BaseRerankRunner):
         score_threshold: float | None = None,
         top_n: int | None = None,
         user: str | None = None,
+        query_type: QueryType = QueryType.TEXT_QUERY,
     ) -> list[Document]:
         """
         Run rerank model
@@ -24,6 +34,56 @@ class RerankModelRunner(BaseRerankRunner):
         :param user: unique user id if needed
         :return:
         """
+        model_manager = ModelManager()
+        is_support_vision = model_manager.check_model_support_vision(
+            tenant_id=self.rerank_model_instance.provider_model_bundle.configuration.tenant_id,
+            provider=self.rerank_model_instance.provider,
+            model=self.rerank_model_instance.model,
+            model_type=ModelType.RERANK,
+        )
+        if not is_support_vision:
+            if query_type == QueryType.TEXT_QUERY:
+                rerank_result, unique_documents = self.fetch_text_rerank(query, documents, score_threshold, top_n, user)
+            else:
+                return documents
+        else:
+            rerank_result, unique_documents = self.fetch_multimodal_rerank(
+                query, documents, score_threshold, top_n, user, query_type
+            )
+
+        rerank_documents = []
+        for result in rerank_result.docs:
+            if score_threshold is None or result.score >= score_threshold:
+                # format document
+                rerank_document = Document(
+                    page_content=result.text,
+                    metadata=unique_documents[result.index].metadata,
+                    provider=unique_documents[result.index].provider,
+                )
+                if rerank_document.metadata is not None:
+                    rerank_document.metadata["score"] = result.score
+                    rerank_documents.append(rerank_document)
+
+        rerank_documents.sort(key=lambda x: x.metadata.get("score", 0.0), reverse=True)
+        return rerank_documents[:top_n] if top_n else rerank_documents
+
+    def fetch_text_rerank(
+        self,
+        query: str,
+        documents: list[Document],
+        score_threshold: float | None = None,
+        top_n: int | None = None,
+        user: str | None = None,
+    ) -> tuple[RerankResult, list[Document]]:
+        """
+        Fetch text rerank
+        :param query: search query
+        :param documents: documents for reranking
+        :param score_threshold: score threshold
+        :param top_n: top n
+        :param user: unique user id if needed
+        :return:
+        """
         docs = []
         doc_ids = set()
         unique_documents = []
@@ -33,33 +93,99 @@ class RerankModelRunner(BaseRerankRunner):
                 and document.metadata is not None
                 and document.metadata["doc_id"] not in doc_ids
             ):
-                doc_ids.add(document.metadata["doc_id"])
-                docs.append(document.page_content)
-                unique_documents.append(document)
+                if not document.metadata.get("doc_type") or document.metadata.get("doc_type") == DocType.TEXT:
+                    doc_ids.add(document.metadata["doc_id"])
+                    docs.append(document.page_content)
+                    unique_documents.append(document)
             elif document.provider == "external":
                 if document not in unique_documents:
                     docs.append(document.page_content)
                     unique_documents.append(document)
 
-        documents = unique_documents
-
         rerank_result = self.rerank_model_instance.invoke_rerank(
             query=query, docs=docs, score_threshold=score_threshold, top_n=top_n, user=user
         )
+        return rerank_result, unique_documents
 
-        rerank_documents = []
+    def fetch_multimodal_rerank(
+        self,
+        query: str,
+        documents: list[Document],
+        score_threshold: float | None = None,
+        top_n: int | None = None,
+        user: str | None = None,
+        query_type: QueryType = QueryType.TEXT_QUERY,
+    ) -> tuple[RerankResult, list[Document]]:
+        """
+        Fetch multimodal rerank
+        :param query: search query
+        :param documents: documents for reranking
+        :param score_threshold: score threshold
+        :param top_n: top n
+        :param user: unique user id if needed
+        :param query_type: query type
+        :return: rerank result
+        """
+        docs = []
+        doc_ids = set()
+        unique_documents = []
+        for document in documents:
+            if (
+                document.provider == "dify"
+                and document.metadata is not None
+                and document.metadata["doc_id"] not in doc_ids
+            ):
+                if document.metadata.get("doc_type") == DocType.IMAGE:
+                    # Query file info within db.session context to ensure thread-safe access
+                    upload_file = (
+                        db.session.query(UploadFile).where(UploadFile.id == document.metadata["doc_id"]).first()
+                    )
+                    if upload_file:
+                        blob = storage.load_once(upload_file.key)
+                        document_file_base64 = base64.b64encode(blob).decode()
+                        document_file_dict = {
+                            "content": document_file_base64,
+                            "content_type": document.metadata["doc_type"],
+                        }
+                        docs.append(document_file_dict)
+                else:
+                    document_text_dict = {
+                        "content": document.page_content,
+                        "content_type": document.metadata.get("doc_type") or DocType.TEXT,
+                    }
+                    docs.append(document_text_dict)
+                doc_ids.add(document.metadata["doc_id"])
+                unique_documents.append(document)
+            elif document.provider == "external":
+                if document not in unique_documents:
+                    docs.append(
+                        {
+                            "content": document.page_content,
+                            "content_type": document.metadata.get("doc_type") or DocType.TEXT,
+                        }
+                    )
+                    unique_documents.append(document)
 
-        for result in rerank_result.docs:
-            if score_threshold is None or result.score >= score_threshold:
-                # format document
-                rerank_document = Document(
-                    page_content=result.text,
-                    metadata=documents[result.index].metadata,
-                    provider=documents[result.index].provider,
+        documents = unique_documents
+        if query_type == QueryType.TEXT_QUERY:
+            rerank_result, unique_documents = self.fetch_text_rerank(query, documents, score_threshold, top_n, user)
+            return rerank_result, unique_documents
+        elif query_type == QueryType.IMAGE_QUERY:
+            # Query file info within db.session context to ensure thread-safe access
+            upload_file = db.session.query(UploadFile).where(UploadFile.id == query).first()
+            if upload_file:
+                blob = storage.load_once(upload_file.key)
+                file_query = base64.b64encode(blob).decode()
+                file_query_dict = {
+                    "content": file_query,
+                    "content_type": DocType.IMAGE,
+                }
+                rerank_result = self.rerank_model_instance.invoke_multimodal_rerank(
+                    query=file_query_dict, docs=docs, score_threshold=score_threshold, top_n=top_n, user=user
                 )
-                if rerank_document.metadata is not None:
-                    rerank_document.metadata["score"] = result.score
-                    rerank_documents.append(rerank_document)
+                return rerank_result, unique_documents
+            else:
+                raise ValueError(f"Upload file not found for query: {query}")
 
-        rerank_documents.sort(key=lambda x: x.metadata.get("score", 0.0), reverse=True)
-        return rerank_documents[:top_n] if top_n else rerank_documents
+        else:
+            raise ValueError(f"Query type {query_type} is not supported")

api/core/rag/rerank/weight_rerank.py

@@ -7,6 +7,8 @@ from core.model_manager import ModelManager
 from core.model_runtime.entities.model_entities import ModelType
 from core.rag.datasource.keyword.jieba.jieba_keyword_table_handler import JiebaKeywordTableHandler
 from core.rag.embedding.cached_embedding import CacheEmbedding
+from core.rag.index_processor.constant.doc_type import DocType
+from core.rag.index_processor.constant.query_type import QueryType
 from core.rag.models.document import Document
 from core.rag.rerank.entity.weight import VectorSetting, Weights
 from core.rag.rerank.rerank_base import BaseRerankRunner
@@ -24,6 +26,7 @@ class WeightRerankRunner(BaseRerankRunner):
         score_threshold: float | None = None,
         top_n: int | None = None,
         user: str | None = None,
+        query_type: QueryType = QueryType.TEXT_QUERY,
     ) -> list[Document]:
         """
         Run rerank model
@@ -43,8 +46,10 @@ class WeightRerankRunner(BaseRerankRunner):
                 and document.metadata is not None
                 and document.metadata["doc_id"] not in doc_ids
             ):
-                doc_ids.add(document.metadata["doc_id"])
-                unique_documents.append(document)
+                # weight rerank only support text documents
+                if not document.metadata.get("doc_type") or document.metadata.get("doc_type") == DocType.TEXT:
+                    doc_ids.add(document.metadata["doc_id"])
+                    unique_documents.append(document)
             else:
                 if document not in unique_documents:
                     unique_documents.append(document)

api/core/rag/retrieval/dataset_retrieval.py

@@ -8,6 +8,7 @@ from typing import Any, Union, cast
 
 from flask import Flask, current_app
 from sqlalchemy import and_, or_, select
+from sqlalchemy.orm import Session
 
 from core.app.app_config.entities import (
     DatasetEntity,
@@ -19,6 +20,7 @@ from core.app.entities.app_invoke_entities import InvokeFrom, ModelConfigWithCre
 from core.callback_handler.index_tool_callback_handler import DatasetIndexToolCallbackHandler
 from core.entities.agent_entities import PlanningStrategy
 from core.entities.model_entities import ModelStatus
+from core.file import File, FileTransferMethod, FileType
 from core.memory.token_buffer_memory import TokenBufferMemory
 from core.model_manager import ModelInstance, ModelManager
 from core.model_runtime.entities.llm_entities import LLMResult, LLMUsage
@@ -37,7 +39,9 @@ from core.rag.datasource.retrieval_service import RetrievalService
 from core.rag.entities.citation_metadata import RetrievalSourceMetadata
 from core.rag.entities.context_entities import DocumentContext
 from core.rag.entities.metadata_entities import Condition, MetadataCondition
-from core.rag.index_processor.constant.index_type import IndexType
+from core.rag.index_processor.constant.doc_type import DocType
+from core.rag.index_processor.constant.index_type import IndexStructureType, IndexTechniqueType
+from core.rag.index_processor.constant.query_type import QueryType
 from core.rag.models.document import Document
 from core.rag.rerank.rerank_type import RerankMode
 from core.rag.retrieval.retrieval_methods import RetrievalMethod
@@ -52,10 +56,12 @@ from core.rag.retrieval.template_prompts import (
     METADATA_FILTER_USER_PROMPT_2,
     METADATA_FILTER_USER_PROMPT_3,
 )
+from core.tools.signature import sign_upload_file
 from core.tools.utils.dataset_retriever.dataset_retriever_base_tool import DatasetRetrieverBaseTool
 from extensions.ext_database import db
 from libs.json_in_md_parser import parse_and_check_json_markdown
-from models.dataset import ChildChunk, Dataset, DatasetMetadata, DatasetQuery, DocumentSegment
+from models import UploadFile
+from models.dataset import ChildChunk, Dataset, DatasetMetadata, DatasetQuery, DocumentSegment, SegmentAttachmentBinding
 from models.dataset import Document as DatasetDocument
 from services.external_knowledge_service import ExternalDatasetService
 
@@ -99,7 +105,8 @@ class DatasetRetrieval:
         message_id: str,
         memory: TokenBufferMemory | None = None,
         inputs: Mapping[str, Any] | None = None,
-    ) -> str | None:
+        vision_enabled: bool = False,
+    ) -> tuple[str | None, list[File] | None]:
         """
         Retrieve dataset.
         :param app_id: app_id
@@ -118,7 +125,7 @@ class DatasetRetrieval:
         """
         dataset_ids = config.dataset_ids
         if len(dataset_ids) == 0:
-            return None
+            return None, []
         retrieve_config = config.retrieve_config
 
         # check model is support tool calling
@@ -136,7 +143,7 @@ class DatasetRetrieval:
         )
 
         if not model_schema:
-            return None
+            return None, []
 
         planning_strategy = PlanningStrategy.REACT_ROUTER
         features = model_schema.features
@@ -182,8 +189,8 @@ class DatasetRetrieval:
                 tenant_id,
                 user_id,
                 user_from,
-                available_datasets,
                 query,
+                available_datasets,
                 model_instance,
                 model_config,
                 planning_strategy,
@@ -213,6 +220,7 @@ class DatasetRetrieval:
         dify_documents = [item for item in all_documents if item.provider == "dify"]
         external_documents = [item for item in all_documents if item.provider == "external"]
         document_context_list: list[DocumentContext] = []
+        context_files: list[File] = []
         retrieval_resource_list: list[RetrievalSourceMetadata] = []
         # deal with external documents
         for item in external_documents:
@@ -248,6 +256,31 @@ class DatasetRetrieval:
                                 score=record.score,
                             )
                         )
+                    if vision_enabled:
+                        attachments_with_bindings = db.session.execute(
+                            select(SegmentAttachmentBinding, UploadFile)
+                            .join(UploadFile, UploadFile.id == SegmentAttachmentBinding.attachment_id)
+                            .where(
+                                SegmentAttachmentBinding.segment_id == segment.id,
+                            )
+                        ).all()
+                        if attachments_with_bindings:
+                            for _, upload_file in attachments_with_bindings:
+                                attchment_info = File(
+                                    id=upload_file.id,
+                                    filename=upload_file.name,
+                                    extension="." + upload_file.extension,
+                                    mime_type=upload_file.mime_type,
+                                    tenant_id=segment.tenant_id,
+                                    type=FileType.IMAGE,
+                                    transfer_method=FileTransferMethod.LOCAL_FILE,
+                                    remote_url=upload_file.source_url,
+                                    related_id=upload_file.id,
+                                    size=upload_file.size,
+                                    storage_key=upload_file.key,
+                                    url=sign_upload_file(upload_file.id, upload_file.extension),
+                                )
+                                context_files.append(attchment_info)
                 if show_retrieve_source:
                     for record in records:
                         segment = record.segment
@@ -288,8 +321,10 @@ class DatasetRetrieval:
             hit_callback.return_retriever_resource_info(retrieval_resource_list)
         if document_context_list:
             document_context_list = sorted(document_context_list, key=lambda x: x.score or 0.0, reverse=True)
-            return str("\n".join([document_context.content for document_context in document_context_list]))
-        return ""
+            return str(
+                "\n".join([document_context.content for document_context in document_context_list])
+            ), context_files
+        return "", context_files
 
     def single_retrieve(
         self,
@@ -297,8 +332,8 @@ class DatasetRetrieval:
         tenant_id: str,
         user_id: str,
         user_from: str,
-        available_datasets: list,
         query: str,
+        available_datasets: list,
         model_instance: ModelInstance,
         model_config: ModelConfigWithCredentialsEntity,
         planning_strategy: PlanningStrategy,
@@ -336,7 +371,7 @@ class DatasetRetrieval:
             dataset_id, router_usage = function_call_router.invoke(query, tools, model_config, model_instance)
 
         self._record_usage(router_usage)
-
+        timer = None
         if dataset_id:
             # get retrieval model config
             dataset_stmt = select(Dataset).where(Dataset.id == dataset_id)
@@ -406,10 +441,19 @@ class DatasetRetrieval:
                             weights=retrieval_model_config.get("weights", None),
                             document_ids_filter=document_ids_filter,
                         )
-                self._on_query(query, [dataset_id], app_id, user_from, user_id)
+                self._on_query(query, None, [dataset_id], app_id, user_from, user_id)
 
                 if results:
-                    self._on_retrieval_end(results, message_id, timer)
+                    thread = threading.Thread(
+                        target=self._on_retrieval_end,
+                        kwargs={
+                            "flask_app": current_app._get_current_object(),  # type: ignore
+                            "documents": results,
+                            "message_id": message_id,
+                            "timer": timer,
+                        },
+                    )
+                    thread.start()
 
                 return results
         return []
@@ -421,7 +465,7 @@ class DatasetRetrieval:
         user_id: str,
         user_from: str,
         available_datasets: list,
-        query: str,
+        query: str | None,
         top_k: int,
         score_threshold: float,
         reranking_mode: str,
@@ -431,10 +475,11 @@ class DatasetRetrieval:
         message_id: str | None = None,
         metadata_filter_document_ids: dict[str, list[str]] | None = None,
         metadata_condition: MetadataCondition | None = None,
+        attachment_ids: list[str] | None = None,
     ):
         if not available_datasets:
             return []
-        threads = []
+        all_threads = []
         all_documents: list[Document] = []
         dataset_ids = [dataset.id for dataset in available_datasets]
         index_type_check = all(
@@ -467,131 +512,226 @@ class DatasetRetrieval:
                         0
                     ].embedding_model_provider
                     weights["vector_setting"]["embedding_model_name"] = available_datasets[0].embedding_model
-
-        for dataset in available_datasets:
-            index_type = dataset.indexing_technique
-            document_ids_filter = None
-            if dataset.provider != "external":
-                if metadata_condition and not metadata_filter_document_ids:
-                    continue
-                if metadata_filter_document_ids:
-                    document_ids = metadata_filter_document_ids.get(dataset.id, [])
-                    if document_ids:
-                        document_ids_filter = document_ids
-                    else:
-                        continue
-            retrieval_thread = threading.Thread(
-                target=self._retriever,
-                kwargs={
-                    "flask_app": current_app._get_current_object(),  # type: ignore
-                    "dataset_id": dataset.id,
-                    "query": query,
-                    "top_k": top_k,
-                    "all_documents": all_documents,
-                    "document_ids_filter": document_ids_filter,
-                    "metadata_condition": metadata_condition,
-                },
-            )
-            threads.append(retrieval_thread)
-            retrieval_thread.start()
-        for thread in threads:
-            thread.join()
-
         with measure_time() as timer:
-            if reranking_enable:
-                # do rerank for searched documents
-                data_post_processor = DataPostProcessor(tenant_id, reranking_mode, reranking_model, weights, False)
-
-                all_documents = data_post_processor.invoke(
-                    query=query, documents=all_documents, score_threshold=score_threshold, top_n=top_k
+            if query:
+                query_thread = threading.Thread(
+                    target=self._multiple_retrieve_thread,
+                    kwargs={
+                        "flask_app": current_app._get_current_object(),  # type: ignore
+                        "available_datasets": available_datasets,
+                        "metadata_condition": metadata_condition,
+                        "metadata_filter_document_ids": metadata_filter_document_ids,
+                        "all_documents": all_documents,
+                        "tenant_id": tenant_id,
+                        "reranking_enable": reranking_enable,
+                        "reranking_mode": reranking_mode,
+                        "reranking_model": reranking_model,
+                        "weights": weights,
+                        "top_k": top_k,
+                        "score_threshold": score_threshold,
+                        "query": query,
+                        "attachment_id": None,
+                    },
                 )
-            else:
-                if index_type == "economy":
-                    all_documents = self.calculate_keyword_score(query, all_documents, top_k)
-                elif index_type == "high_quality":
-                    all_documents = self.calculate_vector_score(all_documents, top_k, score_threshold)
-                else:
-                    all_documents = all_documents[:top_k] if top_k else all_documents
-
-        self._on_query(query, dataset_ids, app_id, user_from, user_id)
+                all_threads.append(query_thread)
+                query_thread.start()
+            if attachment_ids:
+                for attachment_id in attachment_ids:
+                    attachment_thread = threading.Thread(
+                        target=self._multiple_retrieve_thread,
+                        kwargs={
+                            "flask_app": current_app._get_current_object(),  # type: ignore
+                            "available_datasets": available_datasets,
+                            "metadata_condition": metadata_condition,
+                            "metadata_filter_document_ids": metadata_filter_document_ids,
+                            "all_documents": all_documents,
+                            "tenant_id": tenant_id,
+                            "reranking_enable": reranking_enable,
+                            "reranking_mode": reranking_mode,
+                            "reranking_model": reranking_model,
+                            "weights": weights,
+                            "top_k": top_k,
+                            "score_threshold": score_threshold,
+                            "query": None,
+                            "attachment_id": attachment_id,
+                        },
+                    )
+                    all_threads.append(attachment_thread)
+                    attachment_thread.start()
+            for thread in all_threads:
+                thread.join()
+        self._on_query(query, attachment_ids, dataset_ids, app_id, user_from, user_id)
 
         if all_documents:
-            self._on_retrieval_end(all_documents, message_id, timer)
-
-        return all_documents
-
-    def _on_retrieval_end(self, documents: list[Document], message_id: str | None = None, timer: dict | None = None):
-        """Handle retrieval end."""
-        dify_documents = [document for document in documents if document.provider == "dify"]
-        for document in dify_documents:
-            if document.metadata is not None:
-                dataset_document_stmt = select(DatasetDocument).where(
-                    DatasetDocument.id == document.metadata["document_id"]
-                )
-                dataset_document = db.session.scalar(dataset_document_stmt)
-                if dataset_document:
-                    if dataset_document.doc_form == IndexType.PARENT_CHILD_INDEX:
-                        child_chunk_stmt = select(ChildChunk).where(
-                            ChildChunk.index_node_id == document.metadata["doc_id"],
-                            ChildChunk.dataset_id == dataset_document.dataset_id,
-                            ChildChunk.document_id == dataset_document.id,
-                        )
-                        child_chunk = db.session.scalar(child_chunk_stmt)
-                        if child_chunk:
-                            _ = (
-                                db.session.query(DocumentSegment)
-                                .where(DocumentSegment.id == child_chunk.segment_id)
-                                .update(
-                                    {DocumentSegment.hit_count: DocumentSegment.hit_count + 1},
-                                    synchronize_session=False,
-                                )
-                            )
-                    else:
-                        query = db.session.query(DocumentSegment).where(
-                            DocumentSegment.index_node_id == document.metadata["doc_id"]
-                        )
-
-                        # if 'dataset_id' in document.metadata:
-                        if "dataset_id" in document.metadata:
-                            query = query.where(DocumentSegment.dataset_id == document.metadata["dataset_id"])
-
-                        # add hit count to document segment
-                        query.update(
-                            {DocumentSegment.hit_count: DocumentSegment.hit_count + 1}, synchronize_session=False
-                        )
-
-                    db.session.commit()
-
-        # get tracing instance
-        trace_manager: TraceQueueManager | None = (
-            self.application_generate_entity.trace_manager if self.application_generate_entity else None
-        )
-        if trace_manager:
-            trace_manager.add_trace_task(
-                TraceTask(
-                    TraceTaskName.DATASET_RETRIEVAL_TRACE, message_id=message_id, documents=documents, timer=timer
-                )
+            # add thread to call _on_retrieval_end
+            retrieval_end_thread = threading.Thread(
+                target=self._on_retrieval_end,
+                kwargs={
+                    "flask_app": current_app._get_current_object(),  # type: ignore
+                    "documents": all_documents,
+                    "message_id": message_id,
+                    "timer": timer,
+                },
             )
+            retrieval_end_thread.start()
+        retrieval_resource_list = []
+        doc_ids_filter = []
+        for document in all_documents:
+            if document.provider == "dify":
+                doc_id = document.metadata.get("doc_id")
+                if doc_id and doc_id not in doc_ids_filter:
+                    doc_ids_filter.append(doc_id)
+                    retrieval_resource_list.append(document)
+            elif document.provider == "external":
+                retrieval_resource_list.append(document)
+        return retrieval_resource_list
 
-    def _on_query(self, query: str, dataset_ids: list[str], app_id: str, user_from: str, user_id: str):
+    def _on_retrieval_end(
+        self, flask_app: Flask, documents: list[Document], message_id: str | None = None, timer: dict | None = None
+    ):
+        """Handle retrieval end."""
+        with flask_app.app_context():
+            dify_documents = [document for document in documents if document.provider == "dify"]
+            segment_ids = []
+            segment_index_node_ids = []
+            with Session(db.engine) as session:
+                for document in dify_documents:
+                    if document.metadata is not None:
+                        dataset_document_stmt = select(DatasetDocument).where(
+                            DatasetDocument.id == document.metadata["document_id"]
+                        )
+                        dataset_document = session.scalar(dataset_document_stmt)
+                        if dataset_document:
+                            if dataset_document.doc_form == IndexStructureType.PARENT_CHILD_INDEX:
+                                segment_id = None
+                                if (
+                                    "doc_type" not in document.metadata
+                                    or document.metadata.get("doc_type") == DocType.TEXT
+                                ):
+                                    child_chunk_stmt = select(ChildChunk).where(
+                                        ChildChunk.index_node_id == document.metadata["doc_id"],
+                                        ChildChunk.dataset_id == dataset_document.dataset_id,
+                                        ChildChunk.document_id == dataset_document.id,
+                                    )
+                                    child_chunk = session.scalar(child_chunk_stmt)
+                                    if child_chunk:
+                                        segment_id = child_chunk.segment_id
+                                elif (
+                                    "doc_type" in document.metadata
+                                    and document.metadata.get("doc_type") == DocType.IMAGE
+                                ):
+                                    attachment_info_dict = RetrievalService.get_segment_attachment_info(
+                                        dataset_document.dataset_id,
+                                        dataset_document.tenant_id,
+                                        document.metadata.get("doc_id") or "",
+                                        session,
+                                    )
+                                    if attachment_info_dict:
+                                        segment_id = attachment_info_dict["segment_id"]
+                                if segment_id:
+                                    if segment_id not in segment_ids:
+                                        segment_ids.append(segment_id)
+                                        _ = (
+                                            session.query(DocumentSegment)
+                                            .where(DocumentSegment.id == segment_id)
+                                            .update(
+                                                {DocumentSegment.hit_count: DocumentSegment.hit_count + 1},
+                                                synchronize_session=False,
+                                            )
+                                        )
+                            else:
+                                query = None
+                                if (
+                                    "doc_type" not in document.metadata
+                                    or document.metadata.get("doc_type") == DocType.TEXT
+                                ):
+                                    if document.metadata["doc_id"] not in segment_index_node_ids:
+                                        segment = (
+                                            session.query(DocumentSegment)
+                                            .where(DocumentSegment.index_node_id == document.metadata["doc_id"])
+                                            .first()
+                                        )
+                                        if segment:
+                                            segment_index_node_ids.append(document.metadata["doc_id"])
+                                            segment_ids.append(segment.id)
+                                            query = session.query(DocumentSegment).where(
+                                                DocumentSegment.id == segment.id
+                                            )
+                                elif (
+                                    "doc_type" in document.metadata
+                                    and document.metadata.get("doc_type") == DocType.IMAGE
+                                ):
+                                    attachment_info_dict = RetrievalService.get_segment_attachment_info(
+                                        dataset_document.dataset_id,
+                                        dataset_document.tenant_id,
+                                        document.metadata.get("doc_id") or "",
+                                        session,
+                                    )
+                                    if attachment_info_dict:
+                                        segment_id = attachment_info_dict["segment_id"]
+                                        if segment_id not in segment_ids:
+                                            segment_ids.append(segment_id)
+                                        query = session.query(DocumentSegment).where(DocumentSegment.id == segment_id)
+                                if query:
+                                    # if 'dataset_id' in document.metadata:
+                                    if "dataset_id" in document.metadata:
+                                        query = query.where(
+                                            DocumentSegment.dataset_id == document.metadata["dataset_id"]
+                                        )
+
+                                    # add hit count to document segment
+                                    query.update(
+                                        {DocumentSegment.hit_count: DocumentSegment.hit_count + 1},
+                                        synchronize_session=False,
+                                    )
+
+                            db.session.commit()
+
+            # get tracing instance
+            trace_manager: TraceQueueManager | None = (
+                self.application_generate_entity.trace_manager if self.application_generate_entity else None
+            )
+            if trace_manager:
+                trace_manager.add_trace_task(
+                    TraceTask(
+                        TraceTaskName.DATASET_RETRIEVAL_TRACE, message_id=message_id, documents=documents, timer=timer
+                    )
+                )
+
+    def _on_query(
+        self,
+        query: str | None,
+        attachment_ids: list[str] | None,
+        dataset_ids: list[str],
+        app_id: str,
+        user_from: str,
+        user_id: str,
+    ):
         """
         Handle query.
         """
-        if not query:
+        if not query and not attachment_ids:
             return
         dataset_queries = []
         for dataset_id in dataset_ids:
-            dataset_query = DatasetQuery(
-                dataset_id=dataset_id,
-                content=query,
-                source="app",
-                source_app_id=app_id,
-                created_by_role=user_from,
-                created_by=user_id,
-            )
-            dataset_queries.append(dataset_query)
-        if dataset_queries:
-            db.session.add_all(dataset_queries)
+            contents = []
+            if query:
+                contents.append({"content_type": QueryType.TEXT_QUERY, "content": query})
+            if attachment_ids:
+                for attachment_id in attachment_ids:
+                    contents.append({"content_type": QueryType.IMAGE_QUERY, "content": attachment_id})
+            if contents:
+                dataset_query = DatasetQuery(
+                    dataset_id=dataset_id,
+                    content=json.dumps(contents),
+                    source="app",
+                    source_app_id=app_id,
+                    created_by_role=user_from,
+                    created_by=user_id,
+                )
+                dataset_queries.append(dataset_query)
+            if dataset_queries:
+                db.session.add_all(dataset_queries)
         db.session.commit()
 
     def _retriever(
@@ -603,6 +743,7 @@ class DatasetRetrieval:
         all_documents: list,
         document_ids_filter: list[str] | None = None,
         metadata_condition: MetadataCondition | None = None,
+        attachment_ids: list[str] | None = None,
     ):
         with flask_app.app_context():
             dataset_stmt = select(Dataset).where(Dataset.id == dataset_id)
@@ -611,7 +752,7 @@ class DatasetRetrieval:
             if not dataset:
                 return []
 
-            if dataset.provider == "external":
+            if dataset.provider == "external" and query:
                 external_documents = ExternalDatasetService.fetch_external_knowledge_retrieval(
                     tenant_id=dataset.tenant_id,
                     dataset_id=dataset_id,
@@ -663,6 +804,7 @@ class DatasetRetrieval:
                             reranking_mode=retrieval_model.get("reranking_mode") or "reranking_model",
                             weights=retrieval_model.get("weights", None),
                             document_ids_filter=document_ids_filter,
+                            attachment_ids=attachment_ids,
                         )
 
                         all_documents.extend(documents)
@@ -1222,3 +1364,86 @@ class DatasetRetrieval:
             usage = LLMUsage.empty_usage()
 
         return full_text, usage
+
+    def _multiple_retrieve_thread(
+        self,
+        flask_app: Flask,
+        available_datasets: list,
+        metadata_condition: MetadataCondition | None,
+        metadata_filter_document_ids: dict[str, list[str]] | None,
+        all_documents: list[Document],
+        tenant_id: str,
+        reranking_enable: bool,
+        reranking_mode: str,
+        reranking_model: dict | None,
+        weights: dict[str, Any] | None,
+        top_k: int,
+        score_threshold: float,
+        query: str | None,
+        attachment_id: str | None,
+    ):
+        with flask_app.app_context():
+            threads = []
+            all_documents_item: list[Document] = []
+            index_type = None
+            for dataset in available_datasets:
+                index_type = dataset.indexing_technique
+                document_ids_filter = None
+                if dataset.provider != "external":
+                    if metadata_condition and not metadata_filter_document_ids:
+                        continue
+                    if metadata_filter_document_ids:
+                        document_ids = metadata_filter_document_ids.get(dataset.id, [])
+                        if document_ids:
+                            document_ids_filter = document_ids
+                        else:
+                            continue
+                retrieval_thread = threading.Thread(
+                    target=self._retriever,
+                    kwargs={
+                        "flask_app": flask_app,
+                        "dataset_id": dataset.id,
+                        "query": query,
+                        "top_k": top_k,
+                        "all_documents": all_documents_item,
+                        "document_ids_filter": document_ids_filter,
+                        "metadata_condition": metadata_condition,
+                        "attachment_ids": [attachment_id] if attachment_id else None,
+                    },
+                )
+                threads.append(retrieval_thread)
+                retrieval_thread.start()
+            for thread in threads:
+                thread.join()
+
+            if reranking_enable:
+                # do rerank for searched documents
+                data_post_processor = DataPostProcessor(tenant_id, reranking_mode, reranking_model, weights, False)
+                if query:
+                    all_documents_item = data_post_processor.invoke(
+                        query=query,
+                        documents=all_documents_item,
+                        score_threshold=score_threshold,
+                        top_n=top_k,
+                        query_type=QueryType.TEXT_QUERY,
+                    )
+                if attachment_id:
+                    all_documents_item = data_post_processor.invoke(
+                        documents=all_documents_item,
+                        score_threshold=score_threshold,
+                        top_n=top_k,
+                        query_type=QueryType.IMAGE_QUERY,
+                        query=attachment_id,
+                    )
+            else:
+                if index_type == IndexTechniqueType.ECONOMY:
+                    if not query:
+                        all_documents_item = []
+                    else:
+                        all_documents_item = self.calculate_keyword_score(query, all_documents_item, top_k)
+                elif index_type == IndexTechniqueType.HIGH_QUALITY:
+                    all_documents_item = self.calculate_vector_score(all_documents_item, top_k, score_threshold)
+                else:
+                    all_documents_item = all_documents_item[:top_k] if top_k else all_documents_item
+            if all_documents_item:
+                all_documents.extend(all_documents_item)

api/core/schemas/builtin/schemas/v1/multimodal_general_structure.json

@@ -0,0 +1,65 @@
+{
+  "$id": "https://dify.ai/schemas/v1/multimodal_general_structure.json",
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "version": "1.0.0",
+  "type": "array",
+  "title": "Multimodal General Structure",
+  "description": "Schema for multimodal general structure (v1) - array of objects",
+  "properties": {
+    "general_chunks": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "properties": {
+          "content": {
+            "type": "string",
+            "description": "The content"
+          },
+          "files": {
+            "type": "array",
+            "items": {
+              "type": "object",
+              "properties": {
+                "name": {
+                  "type": "string",
+                  "description": "file name"
+                },
+                "size": {
+                  "type": "number",
+                  "description": "file size"
+                },
+                "extension": {
+                  "type": "string",
+                  "description": "file extension"
+                },
+                "type": {
+                  "type": "string",
+                  "description": "file type"
+                },
+                "mime_type": {
+                  "type": "string",
+                  "description": "file mime type"
+                },
+                "transfer_method": {
+                  "type": "string",
+                  "description": "file transfer method"
+                },
+                "url": {
+                  "type": "string",
+                  "description": "file url"
+                },
+                "related_id": {
+                  "type": "string",
+                  "description": "file related id"
+                }
+            },
+            "description": "List of files"
+          }
+        }
+        },
+        "required": ["content"]
+      },
+      "description": "List of content and files"
+    }
+  }
+}

api/core/schemas/builtin/schemas/v1/multimodal_parent_child_structure.json

@@ -0,0 +1,78 @@
+{
+  "$id": "https://dify.ai/schemas/v1/multimodal_parent_child_structure.json",
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "version": "1.0.0",
+  "type": "object",
+  "title": "Multimodal Parent-Child Structure",
+  "description": "Schema for multimodal parent-child structure (v1)",
+  "properties": {
+    "parent_mode": {
+      "type": "string",
+      "description": "The mode of parent-child relationship"
+    },
+    "parent_child_chunks": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "properties": {
+          "parent_content": {
+            "type": "string",
+            "description": "The parent content"
+          },
+          "files": {
+            "type": "array",
+            "items": {
+              "type": "object",
+              "properties": {
+                "name": {
+                  "type": "string",
+                  "description": "file name"
+                },
+                "size": {
+                  "type": "number",
+                  "description": "file size"
+                },
+                "extension": {
+                  "type": "string",
+                  "description": "file extension"
+                },
+                "type": {
+                  "type": "string",
+                  "description": "file type"
+                },
+                "mime_type": {
+                  "type": "string",
+                  "description": "file mime type"
+                },
+                "transfer_method": {
+                  "type": "string",
+                  "description": "file transfer method"
+                },
+                "url": {
+                  "type": "string",
+                  "description": "file url"
+                },
+                "related_id": {
+                  "type": "string",
+                  "description": "file related id"
+                }
+              },
+              "required": ["name", "size", "extension", "type", "mime_type", "transfer_method", "url", "related_id"]
+            },
+            "description": "List of files"
+          },
+          "child_contents": {
+            "type": "array",
+            "items": {
+              "type": "string"
+            },
+            "description": "List of child contents"
+          }
+        },
+        "required": ["parent_content", "child_contents"]
+      },
+      "description": "List of parent-child chunk pairs"
+    }
+  },
+  "required": ["parent_mode", "parent_child_chunks"]
+}

api/core/tools/signature.py

@@ -25,6 +25,24 @@ def sign_tool_file(tool_file_id: str, extension: str) -> str:
     return f"{file_preview_url}?timestamp={timestamp}&nonce={nonce}&sign={encoded_sign}"
 
 
+def sign_upload_file(upload_file_id: str, extension: str) -> str:
+    """
+    sign file to get a temporary url for plugin access
+    """
+    # Use internal URL for plugin/tool file access in Docker environments
+    base_url = dify_config.INTERNAL_FILES_URL or dify_config.FILES_URL
+    file_preview_url = f"{base_url}/files/{upload_file_id}/image-preview"
+
+    timestamp = str(int(time.time()))
+    nonce = os.urandom(16).hex()
+    data_to_sign = f"image-preview|{upload_file_id}|{timestamp}|{nonce}"
+    secret_key = dify_config.SECRET_KEY.encode() if dify_config.SECRET_KEY else b""
+    sign = hmac.new(secret_key, data_to_sign.encode(), hashlib.sha256).digest()
+    encoded_sign = base64.urlsafe_b64encode(sign).decode()
+
+    return f"{file_preview_url}?timestamp={timestamp}&nonce={nonce}&sign={encoded_sign}"
+
+
 def verify_tool_file_signature(file_id: str, timestamp: str, nonce: str, sign: str) -> bool:
     """
     verify signature

api/core/tools/tool_manager.py

@@ -5,7 +5,7 @@ import time
 from collections.abc import Generator, Mapping
 from os import listdir, path
 from threading import Lock
-from typing import TYPE_CHECKING, Any, Literal, Optional, Union, cast
+from typing import TYPE_CHECKING, Any, Literal, Optional, TypedDict, Union, cast
 
 import sqlalchemy as sa
 from sqlalchemy import select
@@ -67,6 +67,11 @@ if TYPE_CHECKING:
 logger = logging.getLogger(__name__)
 
 
+class ApiProviderControllerItem(TypedDict):
+    provider: ApiToolProvider
+    controller: ApiToolProviderController
+
+
 class ToolManager:
     _builtin_provider_lock = Lock()
     _hardcoded_providers: dict[str, BuiltinToolProviderController] = {}
@@ -655,9 +660,10 @@ class ToolManager:
         else:
             filters.append(typ)
 
-        with db.session.no_autoflush:
+        # Use a single session for all database operations to reduce connection overhead
+        with Session(db.engine) as session:
             if "builtin" in filters:
-                builtin_providers = cls.list_builtin_providers(tenant_id)
+                builtin_providers = list(cls.list_builtin_providers(tenant_id))
 
                 # key: provider name, value: provider
                 db_builtin_providers = {
@@ -688,57 +694,74 @@ class ToolManager:
 
             # get db api providers
             if "api" in filters:
-                db_api_providers = db.session.scalars(
+                db_api_providers = session.scalars(
                     select(ApiToolProvider).where(ApiToolProvider.tenant_id == tenant_id)
                 ).all()
 
-                api_provider_controllers: list[dict[str, Any]] = [
-                    {"provider": provider, "controller": ToolTransformService.api_provider_to_controller(provider)}
-                    for provider in db_api_providers
-                ]
+                # Batch create controllers
+                api_provider_controllers: list[ApiProviderControllerItem] = []
+                for api_provider in db_api_providers:
+                    try:
+                        controller = ToolTransformService.api_provider_to_controller(api_provider)
+                        api_provider_controllers.append({"provider": api_provider, "controller": controller})
+                    except Exception:
+                        # Skip invalid providers but continue processing others
+                        logger.warning("Failed to create controller for API provider %s", api_provider.id)
 
-                # get labels
-                labels = ToolLabelManager.get_tools_labels([x["controller"] for x in api_provider_controllers])
-
-                for api_provider_controller in api_provider_controllers:
-                    user_provider = ToolTransformService.api_provider_to_user_provider(
-                        provider_controller=api_provider_controller["controller"],
-                        db_provider=api_provider_controller["provider"],
-                        decrypt_credentials=False,
-                        labels=labels.get(api_provider_controller["controller"].provider_id, []),
+                # Batch get labels for all API providers
+                if api_provider_controllers:
+                    controllers = cast(
+                        list[ToolProviderController], [item["controller"] for item in api_provider_controllers]
                     )
-                    result_providers[f"api_provider.{user_provider.name}"] = user_provider
+                    labels = ToolLabelManager.get_tools_labels(controllers)
+
+                    for item in api_provider_controllers:
+                        provider_controller = item["controller"]
+                        db_provider = item["provider"]
+                        provider_labels = labels.get(provider_controller.provider_id, [])
+                        user_provider = ToolTransformService.api_provider_to_user_provider(
+                            provider_controller=provider_controller,
+                            db_provider=db_provider,
+                            decrypt_credentials=False,
+                            labels=provider_labels,
+                        )
+                        result_providers[f"api_provider.{user_provider.name}"] = user_provider
 
             if "workflow" in filters:
                 # get workflow providers
-                workflow_providers = db.session.scalars(
+                workflow_providers = session.scalars(
                     select(WorkflowToolProvider).where(WorkflowToolProvider.tenant_id == tenant_id)
                 ).all()
 
                 workflow_provider_controllers: list[WorkflowToolProviderController] = []
                 for workflow_provider in workflow_providers:
                     try:
-                        workflow_provider_controllers.append(
+                        workflow_controller: WorkflowToolProviderController = (
                             ToolTransformService.workflow_provider_to_controller(db_provider=workflow_provider)
                         )
+                        workflow_provider_controllers.append(workflow_controller)
                     except Exception:
                         # app has been deleted
-                        pass
+                        logger.exception("Failed to transform workflow provider %s to controller", workflow_provider.id)
+                        continue
+                # Batch get labels for workflow providers
+                if workflow_provider_controllers:
+                    workflow_controllers: list[ToolProviderController] = [
+                        cast(ToolProviderController, controller) for controller in workflow_provider_controllers
+                    ]
+                    labels = ToolLabelManager.get_tools_labels(workflow_controllers)
 
-                labels = ToolLabelManager.get_tools_labels(
-                    [cast(ToolProviderController, controller) for controller in workflow_provider_controllers]
-                )
+                    for workflow_provider_controller in workflow_provider_controllers:
+                        provider_labels = labels.get(workflow_provider_controller.provider_id, [])
+                        user_provider = ToolTransformService.workflow_provider_to_user_provider(
+                            provider_controller=workflow_provider_controller,
+                            labels=provider_labels,
+                        )
+                        result_providers[f"workflow_provider.{user_provider.name}"] = user_provider
 
-                for provider_controller in workflow_provider_controllers:
-                    user_provider = ToolTransformService.workflow_provider_to_user_provider(
-                        provider_controller=provider_controller,
-                        labels=labels.get(provider_controller.provider_id, []),
-                    )
-                    result_providers[f"workflow_provider.{user_provider.name}"] = user_provider
             if "mcp" in filters:
-                with Session(db.engine) as session:
-                    mcp_service = MCPToolManageService(session=session)
-                    mcp_providers = mcp_service.list_providers(tenant_id=tenant_id, for_list=True)
+                mcp_service = MCPToolManageService(session=session)
+                mcp_providers = mcp_service.list_providers(tenant_id=tenant_id, for_list=True)
                 for mcp_provider in mcp_providers:
                     result_providers[f"mcp_provider.{mcp_provider.name}"] = mcp_provider
 

api/core/tools/utils/text_processing_utils.py

@@ -13,5 +13,5 @@ def remove_leading_symbols(text: str) -> str:
     """
     # Match Unicode ranges for punctuation and symbols
     # FIXME this pattern is confused quick fix for #11868 maybe refactor it later
-    pattern = r"^[\u2000-\u206F\u2E00-\u2E7F\u3000-\u303F!\"#$%&'()*+,./:;<=>?@^_`~]+"
+    pattern = r"^[\u2000-\u206F\u2E00-\u2E7F\u3000-\u303F\"#$%&'()*+,./:;<=>?@^_`~]+"
     return re.sub(pattern, "", text)

api/core/workflow/graph_engine/event_management/event_manager.py

@@ -2,6 +2,7 @@
 Unified event manager for collecting and emitting events.
 """
 
+import logging
 import threading
 import time
 from collections.abc import Generator
@@ -12,6 +13,8 @@ from core.workflow.graph_events import GraphEngineEvent
 
 from ..layers.base import GraphEngineLayer
 
+_logger = logging.getLogger(__name__)
+
 
 @final
 class ReadWriteLock:
@@ -180,5 +183,4 @@ class EventManager:
             try:
                 layer.on_event(event)
             except Exception:
-                # Silently ignore layer errors during collection
-                pass
+                _logger.exception("Error in layer on_event, layer_type=%s", type(layer))