Introduce app_asset_contents table as a read-through cache over S3 for
text-like asset files (e.g. .md skill documents). This eliminates N
individual S3 fetches during SkillBuilder builds — bulk_load pulls all
content in a single SQL query with S3 fallback on miss.
Key components:
- CachedContentAccessor: DB-first read / dual-write / S3 fallback
- AssetContentService: static DB operations (get, get_many, upsert, delete)
- should_mirror(): single source of truth for extension-based policy
- Alembic migration for app_asset_contents table
Modified callers:
- SkillBuilder uses accessor.bulk_load() instead of per-node S3 reads
- AppAssetService.get/update_file_content route through accessor
- delete_node cleans both DB cache and S3
- draft_app_assets_initializer uses should_mirror() instead of hardcoded .md
Refactor skill compilation around mergeable bundle patches so dynamic skill updates no longer require full rebuilds. Keep dependency closures accurate by recomputing affected nodes from direct dependency data.
- Introduced CachedPresignStorage to cache presigned download URLs, reducing repeated API calls.
- Updated AppAssetService to utilize CachedPresignStorage for improved performance in asset download URL generation.
- Refactored asset builders and packagers to support the new storage mechanism.
- Removed unused AppAssetsAttrsInitializer to streamline initialization processes.
- Added unit tests for CachedPresignStorage to ensure functionality and reliability.
- Replaced SkillArtifactSet with SkillBundle across various components, enhancing the organization of skill dependencies and references.
- Updated SkillManager methods to load and save bundles instead of artifacts, improving clarity in asset management.
- Refactored SkillCompiler to compile skills into bundles, streamlining the dependency resolution process.
- Adjusted DifyCli and SandboxBashSession to utilize ToolDependencies, ensuring consistent handling of tool references.
- Introduced AssetReferences for better management of file dependencies within skill bundles.
- Introduced SandboxManager.delete_storage method for improved storage management.
- Refactored skill loading and tool artifact handling in DifyCliInitializer and SandboxBashSession.
- Updated LLMNode to extract and compile tool artifacts, enhancing integration with skills.
- Improved attribute management in AttrMap for better error handling and retrieval methods.
- Introduced threading for loading skills and uploading compiled content to improve performance.
- Added data classes for better structure and clarity in handling loaded and compiled skills.
- Refactored the skill compilation process to separate loading and uploading, enhancing maintainability.
