fix: invalidate memory_document_map cache after file tool operations #916
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This fixes a bug where cat() would return stale content after file modifications because the IDE buffer cache (memory_document_map) was not being invalidated when tools wrote to disk. Changes: - auxiliary.rs: Invalidate cache after fs::write() in write_file() (covers create_textdoc, update_textdoc, update_textdoc_by_lines, update_textdoc_regex) - tool_rm.rs: Invalidate cache after file/directory deletion (with prefix removal for directories) - tool_mv.rs: Invalidate cache for source and destination paths after rename, overwrite, and cross-device copy operations Root cause: memory_document_map is an LSP cache for files open in IDE. Tools were writing to disk but not invalidating this cache, causing subsequent cat() calls to return stale IDE buffer content instead of fresh disk content.
- Replace cloud GraphQL knowledge backend with local markdown files - Store knowledge in .refact_knowledge/ folder with YAML frontmatter - Add markdown-aware file splitter for VecDB indexing - Add create_agents_md tool to generate project guide - Add filenames parameter to create_knowledge tool - Save trajectories to .refact_knowledge/trajectories/ - Update prompts with knowledge management instructions - Remove cloud module and related code completely - Allowlist .refact_knowledge/ folder for indexing despite being hidden New tools: - knowledge(search_key) - search local knowledge - create_knowledge(tags, filenames, content) - create knowledge entry - create_agents_md() - generate AGENTS.md from all knowledge Files changed: - New: src/vecdb/vdb_markdown_splitter.rs - New: src/tools/tool_create_agents_md.rs - Rewritten: src/memories.rs - Deleted: src/cloud/ directory - Updated: prompts, tools, file filtering
… commits Replace the existing commit message prompts with comprehensive guidance following the Conventional Commits specification: - Add all standard commit types (feat, fix, refactor, perf, docs, etc.) - Enforce imperative mood and 50/72 character rules - Support scopes for better organization - Add breaking change detection with ! notation and BREAKING CHANGE footer - Include issue reference handling (Fixes #, Closes #, Refs #) - Provide clear examples for each type of commit - Add guidelines for synthesizing user intent with diff analysis
- Add petgraph-based knowledge graph (kg_structs, kg_builder, kg_query) - Extended frontmatter: UUID, kind, status, links, review_after, superseded_by - Subchat-powered create_knowledge: auto-enriches metadata, deprecates outdated docs - Graph-enhanced search: expands results via shared tags/files/entities - Weekly background cleanup: auto-archives old trajectories and deprecated docs - Entity extraction from backticked identifiers - Staleness detection: orphan refs, past review, deprecated ready to archive New files: - src/knowledge_graph/mod.rs - src/knowledge_graph/kg_structs.rs - src/knowledge_graph/kg_builder.rs - src/knowledge_graph/kg_query.rs - src/knowledge_graph/kg_staleness.rs - src/knowledge_graph/kg_subchat.rs - src/knowledge_graph/kg_cleanup.rs Modified: - Cargo.toml: added petgraph - memories.rs: new frontmatter format, deprecate/archive functions - tool_create_knowledge.rs: 2-phase subchat pipeline - tool_knowledge.rs: graph-enhanced search - vdb_markdown_splitter.rs: uses KnowledgeFrontmatter - background_tasks.rs: added cleanup task
- New HTTP endpoint GET /v1/knowledge-graph returns JSON:
- nodes: [{id, node_type, label}] for docs, tags, files, entities
- edges: [{source, target, edge_type}] for all relationships
- stats: doc_count, tag_count, file_count, entity_count, edge_count, etc.
- Enhanced stats logging when knowledge graph is built:
- Total docs (active, deprecated, trajectories, code)
- Tags, files, entities counts
- Graph edge count
New file: src/http/routers/v1/knowledge_graph.rs
- Add EnrichmentParams struct and memories_add_enriched() to memories.rs - Refactor tool_create_knowledge.rs to use memories_add_enriched (254→93 lines) - Add enriched memory creation to tool_deep_research.rs - Add enriched memory creation to tool_strategic_planning.rs - All knowledge entries now get AI-enriched metadata and auto-deprecation
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This fixes a bug where cat() would return stale content after file modifications because the IDE buffer cache (memory_document_map) was not being invalidated when tools wrote to disk.
Changes:
Root cause: memory_document_map is an LSP cache for files open in IDE. Tools were writing to disk but not invalidating this cache, causing subsequent cat() calls to return stale IDE buffer content instead of fresh disk content.