fix: Gracefully handle missing pipeline records in executor

- Check if pipeline record exists before creating S3 final output
- Fall back to basic response if no pipeline record found
- This allows executor to work with or without pipeline manager
- Fixes 'Pipeline not found' error during execution

services/pipeline_executor.py

@@ -303,59 +303,58 @@ Execute ONLY the {len(components_list)} component(s) listed above, then provide
             if "Final Answer:" in assistant_message or "final answer" in assistant_message.lower():
                 # Done!
                 if tool_results:
-                    # Compile component results for pipeline manager
+                    # Try to compile component results for pipeline manager (V3 architecture)
+                    # Note: This requires a pipeline record to exist in MongoDB
+                    # If not found, we'll fall back to basic response without S3 storage
+                    structured_result = {
+                        "status": "completed",
+                        "components_executed": tool_results,
+                        "summary": {
+                            "total_tools_called": len(tool_results),
+                            "tools": list(tool_results.keys())
+                        },
+                        "final_output": assistant_message
+                    }
+                    
                     if PIPELINE_MANAGER_AVAILABLE and session_id:
                         try:
                             pipeline_mgr = get_pipeline_manager()
-                            components_results = []
-                            for comp_name, comp_output in tool_results.items():
-                                components_results.append({
-                                    "component_name": comp_name,
-                                    "status": "completed",
-                                    "result": comp_output,
-                                    "success_message": f"Successfully executed {comp_name}"
-                                })
                             
-                            # Create final output in S3
-                            final_output_data = pipeline_mgr.mark_pipeline_completed(
-                                execution_id=session_id,
-                                components_results=components_results,
-                                executor="bedrock"
-                            )
+                            # Check if pipeline record exists first
+                            pipeline_record = pipeline_mgr.get_pipeline(session_id)
                             
-                            structured_result = {
-                                "status": "completed",
-                                "components_executed": tool_results,
-                                "summary": {
-                                    "total_tools_called": len(tool_results),
-                                    "tools": list(tool_results.keys())
-                                },
-                                "final_output_url": final_output_data.get("final_output_url"),
-                                "final_output_expires_at": final_output_data.get("final_output_expires_at"),
-                                "last_node_output": final_output_data.get("last_node_output"),
-                                "workflow_status": final_output_data.get("workflow_status", "completed")
-                            }
+                            if pipeline_record:
+                                # Pipeline record exists, create S3 final output
+                                components_results = []
+                                for comp_name, comp_output in tool_results.items():
+                                    components_results.append({
+                                        "component_name": comp_name,
+                                        "status": "completed",
+                                        "result": comp_output,
+                                        "success_message": f"Successfully executed {comp_name}"
+                                    })
+                                
+                                # Create final output in S3
+                                final_output_data = pipeline_mgr.mark_pipeline_completed(
+                                    execution_id=session_id,
+                                    components_results=components_results,
+                                    executor="bedrock"
+                                )
+                                
+                                # Add S3 URLs to structured result
+                                structured_result["final_output_url"] = final_output_data.get("final_output_url")
+                                structured_result["final_output_expires_at"] = final_output_data.get("final_output_expires_at")
+                                structured_result["last_node_output"] = final_output_data.get("last_node_output")
+                                structured_result["workflow_status"] = final_output_data.get("workflow_status", "completed")
+                            else:
+                                # No pipeline record - this is OK for direct executor calls
+                                # Just continue with basic response
+                                print(f"ℹ️ No pipeline record found for session {session_id} - using basic response")
+                                
                         except Exception as e:
-                            print(f"⚠️ Failed to create final output in S3: {e}")
-                            structured_result = {
-                                "status": "completed",
-                                "components_executed": tool_results,
-                                "summary": {
-                                    "total_tools_called": len(tool_results),
-                                    "tools": list(tool_results.keys())
-                                },
-                                "final_output": assistant_message
-                            }
-                    else:
-                        structured_result = {
-                            "status": "completed",
-                            "components_executed": tool_results,
-                            "summary": {
-                                "total_tools_called": len(tool_results),
-                                "tools": list(tool_results.keys())
-                            },
-                            "final_output": assistant_message
-                        }
+                            # Pipeline manager failed - continue with basic response
+                            print(f"⚠️ Failed to create S3 final output: {str(e)}")
+                            # structured_result already has basic fields, just continue
                     
                     yield {
                         "type": "final",