mirror of https://github.com/langgenius/dify.git
cd030d82e5
- Move workflow generation logic from LLMGenerator to WorkflowGenerator - Extract to api/core/workflow/generator/ with modular architecture - Implement Planner-Builder pattern for better separation of concerns - Add validation engine with rule-based error classification - Add node and edge repair utilities for auto-fixing common issues - Add deterministic Mermaid generator for consistent output - Reorganize configuration and prompts - Move vibe_config/ to generator/config/ - Move vibe_prompts.py to generator/prompts/ (split into multiple files) - Add builder_prompts.py and planner_prompts.py for new architecture - Enhance frontend workflow handling - Use standard node initialization for proper node setup - Improve variable reference replacement with better error handling - Add model fallback logic for better compatibility - Handle end node outputs format (value_selector vs legacy format) - Ensure parameter-extractor nodes have required 'required' field - Add comprehensive test coverage - Unit tests for mermaid generator, node repair, edge repair - Tests for validation engine and rule system - Tests for planner prompts formatting - Frontend tests for variable reference replacement - Add max_fix_iterations parameter for validate-fix loop configuration # Conflicts: # web/app/components/workflow/hooks/use-workflow-vibe.tsx |
||
|---|---|---|
| .. | ||
| agent | ||
| app | ||
| base | ||
| callback_handler | ||
| datasource | ||
| db | ||
| entities | ||
| errors | ||
| extension | ||
| external_data_tool | ||
| file | ||
| helper | ||
| llm_generator | ||
| mcp | ||
| memory | ||
| model_runtime | ||
| moderation | ||
| ops | ||
| plugin | ||
| prompt | ||
| rag | ||
| repositories | ||
| schemas | ||
| tools | ||
| trigger | ||
| variables | ||
| workflow | ||
| __init__.py | ||
| hosting_configuration.py | ||
| indexing_runner.py | ||
| model_manager.py | ||
| provider_manager.py | ||