v1.7.2 — local index directory fix

Bug Fixes

• local-index: Ensure .context-mode/ parent directory is created before opening code-index.db. better-sqlite3 creates the file but not intermediate directories, causing ctx_local_index / ctx_local_search to fail with "Cannot open database because the directory does not exist" on first use.

Verification

• Build passes (npm run build)
• All 16 local-indexer + searcher tests pass
• CLI local-index, local-search, local-repos verified end-to-end

context-mode v1.7.1 — Build fix

What's New in 1.7.1

Fix

• esbuild externals — added onnxruntime-node to esbuild external dependencies list. Prevents build failures caused by esbuild attempting to bundle native .node bindings from onnxruntime-node.

1.7.0 Features (see full release notes)

• Local code indexing with Merkle diff + hybrid search
• Semantic chunking via tree-sitter + tiktoken fallback
• FTS5 + vector hybrid search with ONNX/Ollama embeddings
• CLI and MCP tools for local code indexing

Migration

No breaking changes. Install/update via:

npm install -g ShmidtS/context-mode

context-mode v1.7.0 — Local Code Indexing

What's New

Local Code Indexing (Merkle diff + hybrid search)

• Merkle-style incremental indexing — SHA-256 + mtime diff. Re-indexes only changed files.
• Semantic chunking — tree-sitter AST extraction (functions, classes, interfaces, methods) with tiktoken fallback for unsupported languages.
• FTS5 + vector hybrid search — BM25 full-text ranking fused with cosine similarity from local embeddings.
• ONNX / Ollama / zero-vector fallback — all-MiniLM-L6-v2 via onnxruntime-node, Ollama API, or graceful FTS5-only degradation.
• Cross-encoder reranking — optional ms-marco-MiniLM-L-6-v2 ONNX reranker + heuristic scoring fallback.
• File watcher — fs.watch with 1-second debounce; persists watch list to ~/.context-mode/watched.json.
• CLI commands: local-index, local-search, local-repos, local-status.
• MCP tools: ctx_local_index, ctx_local_search, ctx_local_status, ctx_local_repos, ctx_local_watch, ctx_local_unwatch.
• WAL-safe SQLite schema — dedicated code-index.db with FTS5 virtual table, vectors BLOB store, and jobs tracking.

Technical Details

• 384-d embeddings, batch size 32.
• Tiktoken truncation for result formatting.
• Lazy model download via .url markers in ~/.context-mode/models/.
• WAL mode + mmap_size = 256MB for read-heavy search workloads.

v1.6.2 — Graph analytics fix & graphify-style report

What's Changed

Fix

• Eliminated N+1 SQLite queries in ctx_graph_analyze. Analytics now builds an in-memory snapshot in a single batch (3 queries), preventing timeouts on large vaults.

Feature

• Graphify-style report: ctx_graph_analyze now returns markdownReport (human-readable report) and tokenEstimate (approximate token reduction vs reading raw notes).
• Plain-English explanations: each surprising connection includes explanation (cross-module, cross-file-type, no shared tags).

Documentation

• Vault graph tools (ctx_vault_graph, ctx_graph_analyze) are now documented as active by default in skills/context-mode/SKILL.md.
• skills/improve-codebase-architecture/SKILL.md references vault graph analysis before codebase exploration.

Technical

• Added getAllNodes() and getNodeTagMap() batch read methods to VaultGraphStore.
• Union-Find clustering replaced per-node BFS in community detection for O(E α(V)) complexity.

context-mode v1.6.1

Full Changelog: v1.6.0...v1.6.1

v1.6.0

What's Changed

Refactoring & Architecture

• Decomposed monolithic modules following DRY/SRP principles
• shared.ts (970+ lines) → stats.ts, batch-helpers.ts, paths.ts, security-helpers.ts, snippet.ts, vault-lifecycle.ts with barrel re-exports
• store.ts (1609 lines) → schema.ts, chunking.ts, reranking.ts, vocabulary.ts, search-helpers.ts, content-store.ts, types.ts with barrel re-exports
• Extracted single reciprocalRankFuse implementation from 3 duplicate RRF copies

New Features

• Graphify integration: confidence-tagged edges (EXTRACTED/INFERRED/AMBIGUOUS) in vault graph analytics
• Composite surprise scoring and filterEdgesByConfidence for vault graph queries
• Semantic search enabled by default (removed env-var gates)
• Per-platform skill files: SKILL-claude.md, SKILL-codex.md, SKILL-cursor.md, SKILL-gemini.md, SKILL-copilot.md

Fixes

• Updated all test paths for new module decomposition (2457 tests pass)
• Added vendor type declarations for @clack/prompts and picocolors
• Cleaned VS Code settings for workspace TypeScript SDK

Full Changelog: v1.5.0...v1.6.0

v1.5.0

Full Changelog: v1.4.4...v1.5.0

v1.4.4

Fixes

• ctx_dead_code falsely flagged reachable entry points on Windows due to backslash path separator mismatch in basename extraction.
• Added language-agnostic ENTRY_STEMS so auto-detection works for any source file (Python, Go, Rust, PHP, etc.), not only .ts/.tsx/.js/.jsx.
• Added pi.extensions and openclaw.extensions parsing to detectPackageJsonEntryPoints, preventing plugin source files from being marked dead.

Constraint: MCP server restart required for the fix to take effect in live sessions.

v1.4.3

Bug Fixes

• Fix ESM dynamic require crash in tree-sitter loader by switching from require() to await import() in src/vault/ast-parser.ts and src/vault/symbol-graph.ts. The esbuild ESM bundle replaced require() with __require() shim that threw "Dynamic require is not supported" on Node.js, causing MCP connection closed on plugin install/update.
• Add auto-rebuild fallback in start.mjs: if server.bundle.mjs fails to load, trigger npm run build before falling back.

Internal

• chore(release): bump version to 1.4.3

v1.4.2

Bug Fixes

• Fixed false "outdated" version warnings — __pkg_dir resolution walked up from src/tools/ and failed to find package.json, causing VERSION to always be "unknown". Now walks up the directory tree looking for a package.json with name: "context-mode".
• Fixed semverNewer with invalid versions — semverNewer("1.4.1", "unknown") returned true due to NaN handling. Added guards for "unknown" and non-numeric version strings.

Docs

• Removed badges and improved source install instructions in README.