feat: pull variable add sandbox file support

api/core/workflow/nodes/llm/llm_utils.py

@@ -1,5 +1,5 @@
 from collections.abc import Sequence
-from typing import cast
+from typing import Any, cast
 
 from sqlalchemy import select, update
 from sqlalchemy.orm import Session
@@ -216,6 +216,7 @@ def build_context(
     prompt_messages: Sequence[PromptMessage],
     assistant_response: str,
     generation_data: LLMGenerationData | None = None,
+    files: Sequence[Any] | None = None,
 ) -> list[PromptMessage]:
     """
     Build context from prompt messages and assistant response.
@@ -229,24 +230,58 @@ def build_context(
         prompt_messages: Initial prompt messages (user query, etc.)
         assistant_response: Final assistant response text
         generation_data: Optional generation data containing trace for tool-enabled runs
+        files: Optional list of File objects generated during execution
     """
 
     context_messages: list[PromptMessage] = [
         _truncate_multimodal_content(m) for m in prompt_messages if m.role != PromptMessageRole.SYSTEM
     ]
 
+    # Build file description suffix if files were generated
+    file_suffix = ""
+    if files:
+        file_descriptions = _build_file_descriptions(files)
+        if file_descriptions:
+            file_suffix = f"\n\n{file_descriptions}"
+
     # For tool-enabled runs, reconstruct messages from trace
     if generation_data and generation_data.trace:
-        context_messages.extend(_build_messages_from_trace(generation_data, assistant_response))
+        context_messages.extend(_build_messages_from_trace(generation_data, assistant_response, file_suffix))
     else:
-        context_messages.append(AssistantPromptMessage(content=assistant_response))
+        context_messages.append(AssistantPromptMessage(content=assistant_response + file_suffix))
 
     return context_messages
 
 
+def _build_file_descriptions(files: Sequence[Any]) -> str:
+    """
+    Build a text description of generated files for inclusion in context.
+
+    The description includes file_id which can be used by subsequent nodes
+    to reference the files via structured output.
+    """
+    if not files:
+        return ""
+
+    descriptions: list[str] = ["[Generated Files]"]
+    for file in files:
+        # Get file attributes (File is a Pydantic model)
+        file_id = getattr(file, "id", None) or getattr(file, "related_id", None)
+        filename = getattr(file, "filename", "unknown")
+        file_type = getattr(file, "type", "unknown")
+        if hasattr(file_type, "value"):
+            file_type = file_type.value
+
+        if file_id:
+            descriptions.append(f"- {filename} (id: {file_id}, type: {file_type})")
+
+    return "\n".join(descriptions)
+
+
 def _build_messages_from_trace(
     generation_data: LLMGenerationData,
     assistant_response: str,
+    file_suffix: str = "",
 ) -> list[PromptMessage]:
     """
     Build assistant and tool messages from trace segments.
@@ -254,7 +289,7 @@ def _build_messages_from_trace(
     Processes trace in order to reconstruct the conversation flow:
     - Model segments with tool_calls -> AssistantPromptMessage with tool_calls
     - Tool segments -> ToolPromptMessage with result
-    - Final response -> AssistantPromptMessage with assistant_response
+    - Final response -> AssistantPromptMessage with assistant_response (with optional file_suffix)
     """
     from core.workflow.nodes.llm.entities import ModelTraceSegment, ToolTraceSegment
 
@@ -290,8 +325,8 @@ def _build_messages_from_trace(
                 )
             )
 
-    # Add final assistant response as the authoritative text
-    messages.append(AssistantPromptMessage(content=assistant_response))
+    # Add final assistant response as the authoritative text (with file info if present)
+    messages.append(AssistantPromptMessage(content=assistant_response + file_suffix))
 
     return messages
 

api/core/workflow/nodes/llm/node.py

@@ -383,16 +383,7 @@ class LLMNode(Node[LLMNodeData]):
                     if tool.enabled
                 ]
 
-            # Unified outputs building
-            outputs = {
-                "text": clean_text,
-                "reasoning_content": reasoning_content,
-                "usage": jsonable_encoder(usage),
-                "finish_reason": finish_reason,
-                "context": llm_utils.build_context(prompt_messages, clean_text, generation_data),
-            }
-
-            # Build generation field
+            # Build generation field and determine files_to_output first
             if generation_data:
                 # Use generation_data from tool invocation (supports multi-turn)
                 generation = {
@@ -420,6 +411,15 @@ class LLMNode(Node[LLMNodeData]):
                 }
                 files_to_output = self._file_outputs
 
+            # Unified outputs building (files passed to context for subsequent node reference)
+            outputs = {
+                "text": clean_text,
+                "reasoning_content": reasoning_content,
+                "usage": jsonable_encoder(usage),
+                "finish_reason": finish_reason,
+                "context": llm_utils.build_context(prompt_messages, clean_text, generation_data, files=files_to_output),
+            }
+
             outputs["generation"] = generation
             if files_to_output:
                 outputs["files"] = ArrayFileSegment(value=files_to_output)
@@ -1921,6 +1921,7 @@ class LLMNode(Node[LLMNodeData]):
         tool_dependencies: ToolDependencies | None,
     ) -> Generator[NodeEventBase, None, LLMGenerationData]:
         result: LLMGenerationData | None = None
+        sandbox_output_files: list[File] = []
 
         # FIXME(Mairuis): Async processing for bash session.
         with SandboxBashSession(sandbox=sandbox, node_id=self.id, tools=tool_dependencies) as session:
@@ -1961,9 +1962,26 @@ class LLMNode(Node[LLMNodeData]):
 
             result = yield from self._process_tool_outputs(outputs)
 
+            # Collect output files from sandbox before session ends
+            # Files are saved as ToolFiles with valid tool_file_id for later reference
+            sandbox_output_files = session.collect_output_files()
+
         if result is None:
             raise LLMNodeError("SandboxSession exited unexpectedly")
 
+        # Merge sandbox output files into result
+        if sandbox_output_files:
+            result = LLMGenerationData(
+                text=result.text,
+                reasoning_contents=result.reasoning_contents,
+                tool_calls=result.tool_calls,
+                sequence=result.sequence,
+                usage=result.usage,
+                finish_reason=result.finish_reason,
+                files=result.files + sandbox_output_files,
+                trace=result.trace,
+            )
+
         return result
 
     def _get_model_features(self, model_instance: ModelInstance) -> list[ModelFeature]: