dify/api/services/conversation_service.py

import contextlib
import logging
from collections.abc import Callable, Sequence
from typing import Any

from sqlalchemy import asc, desc, func, or_, select
from sqlalchemy.orm import Session

from clients.agent_backend import AgentBackendSessionCleanupPayload
from configs import dify_config
from core.app.apps.agent_app.session_store import AgentAppRuntimeSessionStore
from core.app.entities.app_invoke_entities import InvokeFrom
from core.llm_generator.llm_generator import LLMGenerator
from factories import variable_factory
from graphon.variables.types import SegmentType
from libs.datetime_utils import naive_utc_now
from libs.infinite_scroll_pagination import InfiniteScrollPagination
from models import Account, ConversationVariable
from models.model import App, Conversation, EndUser, Message
from services.errors.conversation import (
    ConversationNotExistsError,
    ConversationVariableNotExistsError,
    ConversationVariableTypeMismatchError,
    LastConversationNotExistsError,
)
from services.errors.message import MessageNotExistsError
from tasks.agent_backend_session_cleanup_task import cleanup_conversation_agent_runtime_session
from tasks.delete_conversation_task import delete_conversation_related_data

logger = logging.getLogger(__name__)


class ConversationService:
    @classmethod
    def pagination_by_last_id(
        cls,
        *,
        session: Session,
        app_model: App,
        user: Account | EndUser | None,
        last_id: str | None,
        limit: int,
        invoke_from: InvokeFrom,
        include_ids: Sequence[str] | None = None,
        exclude_ids: Sequence[str] | None = None,
        sort_by: str = "-updated_at",
    ) -> InfiniteScrollPagination:
        if not user:
            return InfiniteScrollPagination(data=[], limit=limit, has_more=False)

        stmt = select(Conversation).where(
            Conversation.is_deleted == False,
            Conversation.app_id == app_model.id,
            Conversation.from_source == ("api" if isinstance(user, EndUser) else "console"),
            Conversation.from_end_user_id == (user.id if isinstance(user, EndUser) else None),
            Conversation.from_account_id == (user.id if isinstance(user, Account) else None),
            or_(Conversation.invoke_from.is_(None), Conversation.invoke_from == invoke_from.value),
        )
        # Check if include_ids is not None to apply filter
        if include_ids is not None:
            if len(include_ids) == 0:
                # If include_ids is empty, return empty result
                return InfiniteScrollPagination(data=[], limit=limit, has_more=False)
            stmt = stmt.where(Conversation.id.in_(include_ids))
        # Check if exclude_ids is not None to apply filter
        if exclude_ids is not None:
            if len(exclude_ids) > 0:
                stmt = stmt.where(~Conversation.id.in_(exclude_ids))

        # define sort fields and directions
        sort_field, sort_direction = cls._get_sort_params(sort_by)

        if last_id:
            last_conversation = session.scalar(stmt.where(Conversation.id == last_id))
            if not last_conversation:
                raise LastConversationNotExistsError()

            # build filters based on sorting
            filter_condition = cls._build_filter_condition(
                sort_field=sort_field,
                sort_direction=sort_direction,
                reference_conversation=last_conversation,
            )
            stmt = stmt.where(filter_condition)
        query_stmt = stmt.order_by(sort_direction(getattr(Conversation, sort_field))).limit(limit)
        conversations = session.scalars(query_stmt).all()

        has_more = False
        if len(conversations) == limit:
            current_page_last_conversation = conversations[-1]
            rest_filter_condition = cls._build_filter_condition(
                sort_field=sort_field,
                sort_direction=sort_direction,
                reference_conversation=current_page_last_conversation,
            )
            count_stmt = select(func.count()).select_from(stmt.where(rest_filter_condition).subquery())
            rest_count = session.scalar(count_stmt) or 0
            if rest_count > 0:
                has_more = True

        return InfiniteScrollPagination(data=conversations, limit=limit, has_more=has_more)

    @classmethod
    def _get_sort_params(cls, sort_by: str):
        if sort_by.startswith("-"):
            return sort_by[1:], desc
        return sort_by, asc

    @classmethod
    def _build_filter_condition(cls, sort_field: str, sort_direction: Callable, reference_conversation: Conversation):
        field_value = getattr(reference_conversation, sort_field)
        if sort_direction is desc:
            return getattr(Conversation, sort_field) < field_value

        return getattr(Conversation, sort_field) > field_value

    @classmethod
    def rename(
        cls,
        app_model: App,
        conversation_id: str,
        user: Account | EndUser | None,
        name: str | None,
        auto_generate: bool,
        *,
        session: Session,
    ):
        conversation = cls.get_conversation(app_model, conversation_id, user, session=session)

        if auto_generate:
            return cls.auto_generate_name(app_model, conversation, session=session)
        else:
            if name is None:
                raise ValueError("name is required when auto_generate is false")
            conversation.name = name
            conversation.updated_at = naive_utc_now()
            session.commit()

        return conversation

    @classmethod
    def auto_generate_name(cls, app_model: App, conversation: Conversation, *, session: Session):
        # get conversation first message
        message = session.scalar(
            select(Message)
            .where(Message.app_id == app_model.id, Message.conversation_id == conversation.id)
            .order_by(Message.created_at.asc())
            .limit(1)
        )

        if not message:
            raise MessageNotExistsError()

        # generate conversation name
        with contextlib.suppress(Exception):
            name = LLMGenerator.generate_conversation_name(
                app_model.tenant_id, message.query, conversation.id, app_model.id
            )
            conversation.name = name

        session.commit()

        return conversation

    @classmethod
    def get_conversation(
        cls, app_model: App, conversation_id: str, user: Account | EndUser | None, *, session: Session
    ):
        conversation = session.scalar(
            select(Conversation)
            .where(
                Conversation.id == conversation_id,
                Conversation.app_id == app_model.id,
                Conversation.from_source == ("api" if isinstance(user, EndUser) else "console"),
                Conversation.from_end_user_id == (user.id if isinstance(user, EndUser) else None),
                Conversation.from_account_id == (user.id if isinstance(user, Account) else None),
                Conversation.is_deleted == False,
            )
            .limit(1)
        )

        if not conversation:
            raise ConversationNotExistsError()

        return conversation

    @classmethod
    def delete(cls, app_model: App, conversation_id: str, user: Account | EndUser | None, *, session: Session):
        """
        Delete a conversation only if it belongs to the given user and app context.

        Before removing the conversation row, this best-effort lifecycle path
        enumerates any ACTIVE conversation-owned Agent backend runtime sessions,
        enqueues asynchronous backend cleanup for rows with persisted runtime
        layer specs, and then retires the local session rows even if enqueueing
        fails. Conversation deletion and related-data cleanup scheduling still
        proceed when that lifecycle bookkeeping only partially succeeds.

        Raises:
            ConversationNotExistsError: When the conversation is not visible to the current user.
        """
        conversation = cls.get_conversation(app_model, conversation_id, user, session=session)
        session_store = AgentAppRuntimeSessionStore()
        stored_sessions = session_store.list_active_sessions_for_conversation(
            tenant_id=app_model.tenant_id,
            app_id=app_model.id,
            conversation_id=conversation.id,
        )

        try:
            logger.info(
                "Initiating conversation deletion for app_name %s, conversation_id: %s",
                app_model.name,
                conversation_id,
            )
            for stored_session in stored_sessions:
                try:
                    if stored_session.runtime_layer_specs:
                        payload = AgentBackendSessionCleanupPayload(
                            session_snapshot=stored_session.session_snapshot,
                            runtime_layer_specs=stored_session.runtime_layer_specs,
                            idempotency_key=(
                                f"{stored_session.scope.tenant_id}:{stored_session.scope.app_id}:"
                                f"{stored_session.scope.conversation_id}:agent-runtime-session-cleanup:"
                                f"{stored_session.scope.agent_id}:"
                                f"{stored_session.scope.agent_config_snapshot_id or 'no-config'}:"
                                f"{stored_session.backend_run_id or 'no-run'}"
                            ),
                            metadata={
                                "tenant_id": stored_session.scope.tenant_id,
                                "app_id": stored_session.scope.app_id,
                                "conversation_id": stored_session.scope.conversation_id,
                                "agent_id": stored_session.scope.agent_id,
                                "agent_config_snapshot_id": stored_session.scope.agent_config_snapshot_id,
                                "previous_agent_backend_run_id": stored_session.backend_run_id,
                            },
                        )
                        cleanup_conversation_agent_runtime_session.delay(payload.model_dump(mode="json"))
                except Exception:
                    logger.warning(
                        "Failed to enqueue Agent backend cleanup for conversation deletion: "
                        "tenant_id=%s app_id=%s conversation_id=%s agent_id=%s backend_run_id=%s",
                        stored_session.scope.tenant_id,
                        stored_session.scope.app_id,
                        stored_session.scope.conversation_id,
                        stored_session.scope.agent_id,
                        stored_session.backend_run_id,
                        exc_info=True,
                    )
                finally:
                    try:
                        session_store.mark_cleaned(
                            scope=stored_session.scope,
                            backend_run_id=stored_session.backend_run_id,
                        )
                    except Exception:
                        logger.warning(
                            "Failed to retire Agent App runtime session for conversation deletion: "
                            "tenant_id=%s app_id=%s conversation_id=%s agent_id=%s backend_run_id=%s",
                            stored_session.scope.tenant_id,
                            stored_session.scope.app_id,
                            stored_session.scope.conversation_id,
                            stored_session.scope.agent_id,
                            stored_session.backend_run_id,
                            exc_info=True,
                        )

            session.delete(conversation)
            session.commit()

            delete_conversation_related_data.delay(conversation.id)

        except Exception:
            session.rollback()
            raise

    @classmethod
    def get_conversational_variable(
        cls,
        app_model: App,
        conversation_id: str,
        user: Account | EndUser | None,
        limit: int,
        last_id: str | None,
        variable_name: str | None = None,
        *,
        session: Session,
    ) -> InfiniteScrollPagination:
        conversation = cls.get_conversation(app_model, conversation_id, user, session=session)

        stmt = (
            select(ConversationVariable)
            .where(ConversationVariable.app_id == app_model.id)
            .where(ConversationVariable.conversation_id == conversation.id)
            .order_by(ConversationVariable.created_at)
        )

        # Apply variable_name filter if provided
        if variable_name:
            # Filter using JSON extraction to match variable names case-insensitively
            from libs.helper import escape_like_pattern

            escaped_variable_name = escape_like_pattern(variable_name)
            # Filter using JSON extraction to match variable names case-insensitively
            if dify_config.DB_TYPE in ["mysql", "oceanbase", "seekdb"]:
                stmt = stmt.where(
                    func.json_extract(ConversationVariable.data, "$.name").ilike(
                        f"%{escaped_variable_name}%", escape="\\"
                    )
                )
            elif dify_config.DB_TYPE == "postgresql":
                stmt = stmt.where(
                    func.json_extract_path_text(ConversationVariable.data, "name").ilike(
                        f"%{escaped_variable_name}%", escape="\\"
                    )
                )

        if last_id:
            last_variable = session.scalar(stmt.where(ConversationVariable.id == last_id))
            if not last_variable:
                raise ConversationVariableNotExistsError()

            # Filter for variables created after the last_id
            stmt = stmt.where(ConversationVariable.created_at > last_variable.created_at)

        # Apply limit to query: fetch one extra row to determine has_more
        query_stmt = stmt.limit(limit + 1)
        rows = session.scalars(query_stmt).all()

        has_more = False
        if len(rows) > limit:
            has_more = True
            rows = rows[:limit]  # Remove the extra item

        variables = [
            {
                "created_at": row.created_at,
                "updated_at": row.updated_at,
                **row.to_variable().model_dump(),
            }
            for row in rows
        ]

        return InfiniteScrollPagination(variables, limit, has_more)

    @classmethod
    def update_conversation_variable(
        cls,
        app_model: App,
        conversation_id: str,
        variable_id: str,
        user: Account | EndUser | None,
        new_value: Any,
        *,
        session: Session,
    ):
        """
        Update a conversation variable's value.

        Args:
            app_model: The app model
            conversation_id: The conversation ID
            variable_id: The variable ID to update
            user: The user (Account or EndUser)
            new_value: The new value for the variable

        Returns:
            Dictionary containing the updated variable information

        Raises:
            ConversationNotExistsError: If the conversation doesn't exist
            ConversationVariableNotExistsError: If the variable doesn't exist
            ConversationVariableTypeMismatchError: If the new value type doesn't match the variable's expected type
        """
        # Verify conversation exists and user has access
        conversation = cls.get_conversation(app_model, conversation_id, user, session=session)

        # Get the existing conversation variable
        stmt = (
            select(ConversationVariable)
            .where(ConversationVariable.app_id == app_model.id)
            .where(ConversationVariable.conversation_id == conversation.id)
            .where(ConversationVariable.id == variable_id)
        )

        existing_variable = session.scalar(stmt)
        if not existing_variable:
            raise ConversationVariableNotExistsError()

        # Convert existing variable to Variable object
        current_variable = existing_variable.to_variable()

        # Validate that the new value type matches the expected variable type
        expected_type = SegmentType(current_variable.value_type)

        # There is showing number in web ui but int in db
        if expected_type == SegmentType.INTEGER:
            expected_type = SegmentType.NUMBER

        if not expected_type.is_valid(new_value):
            inferred_type = SegmentType.infer_segment_type(new_value)
            raise ConversationVariableTypeMismatchError(
                f"Type mismatch: variable '{current_variable.name}' expects {expected_type.value}, "
                f"but got {inferred_type.value if inferred_type else 'unknown'} type"
            )

        # Create updated variable with new value only, preserving everything else
        updated_variable_dict = {
            "id": current_variable.id,
            "name": current_variable.name,
            "description": current_variable.description,
            "value_type": current_variable.value_type,
            "value": new_value,
            "selector": current_variable.selector,
        }

        updated_variable = variable_factory.build_conversation_variable_from_mapping(updated_variable_dict)
        existing_variable.data = updated_variable.model_dump_json()
        session.commit()

        return {
            "created_at": existing_variable.created_at,
            "updated_at": naive_utc_now(),  # Update timestamp
            **updated_variable.model_dump(),
        }