Spaces:
Runtime error
Runtime error
Commit History
-
Fix CrewAI compatibility and agent input parameter extraction
7ea9284 -
Remove secrets and ignore .env
97e46c7 -
solved chatllm problems
2610f53
-
final commits
ff40840 -
Added fixed agent to agent communication
6df13ef
-
adding pipeline status into the message response
896941f
-
Update api_routes_v2.py
09168dc -
Update api_routes_v2.py
d8b9e43 -
Update api_routes_v2.py
7eea49e -
Update api_routes_v2.py
fc7cdd2 -
Update api_routes_v2.py
65abc8a
-
Update api_routes_v2.py
5dc0941 -
Update api_routes_v2.py
6187f40
-
Enforced chat_name along with chat_id (#1)
- Enforced chat_name along with chat_id (c2201953887a18526296e481ed7e1016349d1f23) Co-authored-by: Swastik <stellar413@users.noreply.huggingface.co>6b9cfa1
-
Update: last_activity updates for all message interactions
- Updates last_activity for both user and assistant messages - Updates for pipeline executions - Does NOT update when viewing sessions or history - Shows true last interaction time (any message activity)c1a55aa -
Fix: Update last_activity only for user messages
- Remove automatic last_activity updates from get_session() - Add update_activity parameter to update_session() - Only update last_activity when user messages are added - Viewing sessions/history no longer updates last_activity - Ensures session list shows true last interaction time6cb5907 -
Fix: Properly detect and report pipeline failures
- Check pipeline result status before marking as success/failure - Set hasError=true when pipeline fails or partially completes - Show appropriate error messages for failed components - Update pipeline status to 'failed' instead of 'completed' for errors - Extract component-level error details for better user feedback6f8bb0b -
Implement Gemini-based automatic chat name generation
- Replace Bedrock chat naming with Gemini model - Add simple message detection (greetings use timestamp) - Creative naming for meaningful messages using Gemini - Auto-trigger after first user message - Fallback to timestamp on Gemini API errorsdd547c8 -
Fix: Add missing generate_message_id function to schemas.py
b1206db
-
Refactor: Use MongoDB _id for messages, pipeline_id_index for components, standardize outputs to result:{text:{}} format
d6a6b32
-
Update api_routes_v2.py
6431123 -
Update api_routes_v2.py
52b54a6 -
Update api_routes_v2.py
b8b4625
-
feat: Complete API response corrections - add pipelines endpoint and enhanced error handling
- Added GET /api/v2/sessions/{session_id}/pipelines endpoint - Returns all pipeline executions with full component details - Includes component_id, status, output, errors, metadata - Removes internal S3 keys from response - Enhanced pipeline completion responses - Added output_id and download_url to final_output - Enables direct download of pipeline results - Improved error handling - Added component-level error details in error responses - Includes failed_component with component_id, tool_name, and error - Better debugging and UI error display All API corrections from specification now fully implemented.4771f50 -
feat: API response corrections - add message_id, component tracking, and error handling
- Added ID generation utilities (message_id, component_id, output_id) - Enhanced ComponentMetadata with component_output, hasError, and structured errors - Renamed result_preview to result in pipeline responses - Added hasError flag to ChatResponse model - Removed internal fields (current_file, pipeline_s3_key) from API responses - Updated session_manager to use generate_message_id() - Enhanced pipeline_manager with component tracking and error handling - Updated API routes to expose new response structures All changes are backward compatible. No database migration required.7c2e633
-
Update api_routes_v2.py
3804e08 -
Update services/pipeline_executor.py
e6e087d -
Update api_routes_v2.py
51231a3 -
Update api_routes_v2.py
da3a0c4 -
Update api_routes_v2.py
80cbf9a
-
Update services/pipeline_executor.py
db5afca
-
feat: Add pipeline components metadata and fix sort orders
- Added tools[], component_count, and components[] to pipelines_history - Pipelines now load from S3 to extract component/tool details - Pipelines sorted by most recent first (updated_at/created_at desc) - GET /sessions now returns newest sessions first (sorted by last_activity) - Each pipeline now shows: tools list, component count, full component details - Implements requested pipeline metadata enhancementa5c898f -
fix: CRITICAL - Replace S3 URIs with presigned URLs in all API responses
- Changed file_url in session.file_metadata to use presigned_url instead of s3_uri - Updated file_data in upload messages to use presigned_url - Fixed fileUrl in message metadata to use presigned_url - All user-facing fileUrl fields now return accessible presigned URLs - S3 URIs kept internally as s3_uri field for reference - Fixes issue where users received unusable s3:// URIs3fa92e6 -
fix: Add file metadata to messages in conversation history
- Updated _add_and_mirror_message to accept file_metadata parameter - Messages now include file, fileName, fileUrl fields when file is attached - Fixed issue where file uploaded with message showed has_file: false - Chat unified now passes file metadata when user sends message with file - Resolves missing file info in GET /sessions/{id}/history responsedb3bed8
-
fix: Remove JSON result from assistant_response in /chat/unified - cleaner UI response
95a417d -
fix: CRITICAL - Store pipelines in BOTH session.pipelines_history AND pipelines collection
- Root cause: GET /pipelines reads from pipelines collection - But _create_pipeline_record only wrote to session.pipelines_history - Solution: Call pipeline_manager.create_pipeline_metadata() too - Also updated _update_pipeline_status to update both locations - Pipelines now appear in GET /sessions/{id}/pipelines endpoint - Fixes empty pipelines array issue2e45e3f -
fix: Critical pipeline ID sequencing bug
- Fixed pipelines not appearing in /sessions/{id}/pipelines endpoint - Root cause: _create_pipeline_record called AFTER session update - Solution: Call _create_pipeline_record FIRST (adds pipeline_id to dict), THEN update session - Now proposed_pipeline in MongoDB includes pipeline_id - Pipeline status updates now work correctly (pipeline_id is always present) - Fixes empty pipelines array in API responsecd1d4e4 -
fix: Replace _mark_latest_pipeline_history_status with V3 functions
- Fixed 'name _mark_latest_pipeline_history_status is not defined' error - Replaced with _update_pipeline_status in non-streaming approval flow - Added model attribution tracking on pipeline execution success - Pipeline status now properly updates S3 and MongoDB metadata - Lines 1083, 1093 updated with V3 compliancec978648 -
fix: Add missing uuid import
- Fixes 'name uuid is not defined' error in pipeline creation - uuid module needed by _create_pipeline_record function - Resolves pipeline generation failuresd02eed2 -
fix: Replace remaining _push_pipeline_history calls with _create_pipeline_record
- Fixed 'name _push_pipeline_history is not defined' error - Replaced 3 orphaned calls in non-streaming and streaming endpoints - Lines 1132, 1192, 1545 now use V3 _create_pipeline_record - Pipeline generation now properly creates S3 files before executionc770c10 -
fix: V3 migration issues - Auto-migrate old sessions and use Mistral for chat names
- Fix integrity check to auto-migrate old MongoDB messages to S3 - Clear MongoDB messages array after successful migration - Initialize S3 for new empty sessions - Change chat name model from Claude to Mistral (V3 compliance) - Resolves session integrity failure on second message - Resolves Bedrock ResourceNotFoundException for chat namesfd9d0f9 -
feat: Implement V3 S3 Architecture - Complete Refactor
- S3 as single source of truth for conversations and pipelines - MongoDB stores metadata only (IDs, status, timestamps, S3 references) - Added conversation integrity validation (_validate_conversation_integrity) - Added pipeline integrity validation (_validate_pipeline_integrity) - Implemented pipeline lifecycle enforcement (ID creation before execution) - Added model attribution tracking (Bedrock/Mistral) - Implemented DELETE /sessions/{session_id} endpoint - Deleted 10 redundant/duplicate endpoints - Removed 900+ lines of legacy code - All changes comply with MasterLLM V3 Architecture Rules0f8c838 -
Convert all POST API endpoints to use Form data for Swagger UI testing
e235007
-
feat: Unified pipeline lifecycle - single S3 file per pipeline - Pipeline now stored in S3 with pipeline_id at creation - Proposals stored in S3, only reference in MongoDB - Results appended to same S3 file on completion - Workflow save uses pipeline_id, works at any stage - pipelines_history now in session API response
aaaba76
-
feat: Display final_output_url in pipeline completion message - shows presigned S3 URL where all component outputs are stored with expiry time
6648afb -
feat: Add V3 API endpoints with file metadata - sessions list, history, and pipeline executions
c4a01de -
fix: Add missing execution_id initialization
Fixed NameError: 'execution_id is not defined' - Initialize execution_id = session_id before use - Added missing pipeline record creation code - execution_id gets set from pipeline_manager.create_pipeline_record() - Falls back to session_id if pipeline creation fails - Fixed typo in processing message This completes the pipeline record creation feature.28f9d0d -
feat: Complete file upload V3 integration
Fix 4: Integrated Gradio file uploads with V3 architecture What Changed: - File uploads now go to S3 (sessions/{id}/files/{file_id}_{name}) - Creates records in MongoDB 'files' collection - Generates presigned URLs (7-day expiry) - Stores file metadata (file_id, size, mime_type, s3_key) - Adds file_id reference to session - Includes file_data in system messages - Graceful fallback if S3 upload fails Files Collection Now Populated: - file_id, session_id, file_name - s3_bucket, s3_key - presigned_url, presigned_expires_at - file_size, mime_type, uploaded_at Benefits: - Gradio UI and chat/unified API now in sync - Files stored permanently in S3 - File access via presigned URLs - Consistent behavior across interfaces All 5 critical V3 fixes now complete!b4b0135 -
feat: Implement remaining V3 architecture fixes
Fixes 2, 3, and 5 implemented in app.py: Pipeline Record Creation (Fix 2): - Import pipeline_manager and workflow_manager - Create pipeline record BEFORE execution starts - Use execution_id instead of session_id for tracking S3-Only Result Storage (Fix 3): - Removed pipeline_result storage in session documents - Removed deprecated save_pipeline_execution() calls - Results now stored only in S3 via pipeline_manager - MongoDB pipelines collection properly populated Workflow Save Prompting (Fix 5): - Added prompt after successful execution - Stores pending_workflow_save in session - Detects user confirmation (save/yes/sure/ok) - Saves workflow to workflows collection + S3 - Clears pending state after save Component Status Handling: - Added component_status event type handling - Displays component success messages to user - User-friendly emoji status indicators Chat Name Fix: - Only generates once per session (already committed) Remaining: - Fix 4: Gradio file upload integration (sep…f266ceb -
fix: Chat name generation now only happens once per session
- Added check before calling chat name generation - Prevents multiple Bedrock API calls for same session - Fixes 'Failed to generate chat name' spam in logs - Chat name generates only after first user message if not already set2534de8