chore(main): release 0.1.0

[aeneasr]

🤖 I have created a release beep boop

0.1.0 (2026-04-07)

⚠ BREAKING CHANGES

• lumen install and lumen uninstall commands removed. Use claude plugin install lumen or claude --plugin-dir . instead.

Features

• add cwd parameter to MCP tools for reliable index root detection (b622e62)
• add go/ast chunker for function/method/type/interface/const/var extraction (8efb510)
• add index orchestrator with merkle-based incremental indexing (da1fa90)
• add internal/config package for shared configuration (234ca1e)
• add LUMEN_EMBED_DIMS/LUMEN_EMBED_CTX env vars for custom models (#80) (2b124b1)
• add Markdown/MDX chunker splitting by ATX headings (81863f8)
• add merkle tree change detection for file hashing and diffing (05de922)
• add model registry mapping Ollama model names to specs (332e51e)
• add MultiChunker and DefaultLanguages for all supported file types (4c9d237)
• add ollama embedder with batching and retry (a489800)
• add PHP tree-sitter chunker (f109b8b)
• add PTerm-based TUI progress output for index command (#38) (7fee513)
• add sqlite + sqlite-vec store with vector search and batch inserts (e12bb2d)
• add StructuredChunker for YAML/JSON files with TDD (7d78504)
• add TreeSitterChunker for multi-language AST parsing (cf25133)
• add YAML+JSON data chunker splitting by top-level keys (466e9fd)
• align Lumen packaging across Codex, Cursor, and OpenCode (#73) (665c8db)
• auto-reset vec_chunks on embedding dimension mismatch (ff3279a)
• bench-swe: add PTerm-based TUI progress component (#43) (edb13dc)
• better benchmarks with swe-bench approach and improved chunker (#34) (c722d66)
• change default limit to 50, default min_score to 0.5; handle min_score=-1 as no-filter (becc603)
• chunker: add C# language support (915e917)
• chunker: add Dart language support (#69) (00a985b)
• cmd: add search subcommand with --trace diagnostic flag (e1f4bf2)
• cmd: Close() waits for background reindex goroutines (d7b7cad)
• cmd: non-blocking ensureIndexed with 15s timeout and background reindex (fa883d7)
• cmd: render StaleWarning in semantic_search output (ce2503b)
• config: add LUMEN_EMBED_DIMS/LUMEN_EMBED_CTX escape hatch for custom models (#88) (29ad682)
• config: resolve ModelAliases before KnownModels lookup (08989a4)
• distribute lumen binary via npm optional dependencies (#47) (6c382af)
• embedder: add manutic/nomic-embed-code:7b to known models (#42) (b2b8e40)
• go-ast + sqlite-vec + mcp server (8b13f64)
• hook,stdio: background index pre-warming and TTL pre-population (338fdc9)
• hook: spawn detached background indexer on session start (d943ebf)
• ignore known vendor dirs and gitignore (67668f1)
• improved benchmarks and further tree sitter fixes (67f9dd3)
• index: acquire flock before indexing, thread ctx, cancel on SIGTERM (9409971)
• index: add root hashes and worktree info to indexing plan log (bd237e3)
• index: add slog to background indexer and enrich Stats with change breakdown (919eef1)
• indexlock: add flock-based advisory lock for index coordination (2a55737)
• index: reuse ancestor index for non-git subdirectories (3f89427)
• index: seed CLI index command from sibling worktree on first use (ff55769)
• index: seed worktree indexes from sibling to skip full re-embedding (#52) (76809bd)
• install: show only supported models with local availability status (6d5a5a0)
• LM Studio backend, chunk deduplication, and quality improvements (0bd3fe3)
• make embedding dimensions configurable via AGENT_INDEX_EMBED_DIMS (87944cd)
• merkle: load global gitignore (core.excludesFile) (04e7bcb)
• merkle: support linguist-vendored in .gitattributes (c7492d4)
• pass num_ctx to Ollama based on model's context length (d5a4540)
• replace key-based DataChunker with plain text splitting for JSON and YAML (4cff686)
• reuse parent index for subdirectory searches (7e9f803)
• split oversized chunks at line boundaries before embedding (cae8a90)
• stdio: group search results by file with score-based ranking (967e23e)
• stdio: skip EnsureFresh when background indexer holds lock (64a8d03)
• switch default model to qwen3-embedding:8b, rename module (0df840a)
• use XML-tagged output format for search results (5e34a89)
• wire MCP server with semantic_search and index_status tools (a57bebc)
• wire MultiChunker and MakeExtSkip for multi-language indexing (0caed1f)
• wire StructuredChunker into DefaultLanguages for yaml/yml/json files (91ebc82)

Bug Fixes

• broken json (5d4496e)
• build: fix cross-platform build on Apple Silicon (597821d)
• C pointer-return query; add test coverage for all languages (4446d1f)
• chunker: cap leading comments to 10 lines (34f56eb)
• chunker: deduplicate overlapping tree-sitter chunks (0581cf0)
• ci: chain goreleaser into release-please workflow (8ceb5d6)
• ci: remove redundant test/vet/lint steps from release job (2f33014)
• cmd: add TestMain to prevent fork-bomb in TestSpawnBackgroundIndexer (19849aa)
• cmd: close indexers on shutdown and set up signals before DB open (859a274)
• cmd: eliminate reindex fragmentation causing constant cpu usage (c7eb27e)
• cmd: lint cleanup in search subcommand (736e2c6)
• cmd: remove lumberjack rotation; add discardLog to test fixtures (502f454)
• convert cosine distance to similarity score, explicit ORDER BY score DESC (85bc4eb)
• e2e: add LUMEN_REINDEX_TIMEOUT to prevent premature search during indexing (1131bbb)
• e2e: increase reindex timeout to 10m for PHP fixture set (19e6b83)
• explicitly ignore f.Close() error in parseLinguistGenerated (f95ca35)
• fix 5 pre-existing E2E test failures (64377cc)
• harden concurrent indexer lifecycle (close cancellation, reindex dedup, reader consistency) (e943a79)
• hook: add license headers, smoke test, and Windows stub guidance (2cffb63)
• index: cap parent index walk at git repo boundary + store perf improvements (#58) (c5074e2)
• index: cap parent index walk at git repository boundary (#56) (10e9635)
• index: defer UpsertFile until chunks are flushed (9daac4f)
• index: exclude internal git worktrees from project index (#54) (4890908)
• index: force reindex removes deleted files and purges stale extensions (b3b3b48)
• index: prevent triple indexing when non-git parent contains git repos (#81) (adf4e54)
• index: purge stale unsupported-extension records from donor seeding + freshness TTL cache + debug logging (cea7687)
• index: serialize concurrent Index/EnsureFresh calls to prevent duplicate key errors (6f27677)
• index: simplify isBinaryContent using slices.Contains (d04216a)
• index: skip binary files based on NUL-byte detection (4fef1d2)
• index: write skip message to stderr, document signal-race trade-off (bbc63f1)
• lint: wrap deferred Close calls to satisfy errcheck (5a200a2)
• log: use lumberjack for log rotation; remove repro test; fix imports (24d6906)
• make all e2e and unit tests pass (57db1ea)
• mcp: rename n_results to limit in semantic_search tool (ca74157)
• merkle: skip symlinks and files >10MB during tree walk (3f3de05)
• opencode: invoke run.cmd directly instead of wrapping with sh (f3fd945)
• plugin: align plugin files with Claude Code plugin reference (89cdb08)
• plugin: remove invalid manifest fields and bump version to 0.0.8 (34ddfba)
• prevent Search() from blocking during indexing by using separate read connection (bdb837a), closes #94
• reduce cyclomatic complexity of major functions (8faef25)
• reindex stability — sentinel resume, donor safety, worktree subdirs, progress bar (#67) (07782d6)
• remove package-level chunk from Go AST chunker (e53d95b)
• remove worktrees (61384a7)
• replace custom minInt with builtin min in structured_test.go (8aa1269)
• replace custom time.Sleep backoff with go-retry for context-aware retries (91a14c1)
• resolve import and style issues in merkle package (fd8c1a8)
• scripts: correct jsonpath field name in release-please extra-files config (528d6a3)
• scripts: pin binary download to manifest version (9b1f90d)
• scripts: point to ory/lumen and download raw binaries (fef815d)
• scripts: resolve version from release-please manifest before GitHub API (42a4e4a)
• search: pass db file path instead of directory to setupIndexer (#75) (81cb4db)
• set MaxOpenConns(1) for SQLite, fix pragma test to scan int (7f82484)
• skip permission-denied files instead of aborting index/search (#84) (fde97b4)
• stdio: default to git repo root when no existing index found (929b4c9)
• stdio: increase scanner buffer to 1MB for long lines (36c0435)
• stdio: only adopt cwd as index root when a DB already exists there (80fe5d4)
• stdio: propagate LUMEN_FRESHNESS_TTL from config to indexerCache (cb4029b)
• stdio: resolve symlinks in search input paths before processing (f1caa70)
• stdio: reuse parent index when search path points into internal worktree (e96b464)
• stdio: skip force reindex when background indexer holds lock (a867910)
• store,index: auto-recover from SQLite database corruption (652b418)
• store: increase busy_timeout to 120s and use INSERT OR REPLACE for vec_chunks (25a0e01)
• store: increase busy_timeout to 30s for slow embedding batches (91a2883)
• store: wrap dimension-reset deletes in a transaction (33ea835)
• test: resolve symlinks in e2e test paths for macOS compatibility (fca3aec)
• tests: prevent fork bomb and close indexer FDs in cmd tests (90c3d94)
• tool description to increase invokation (5f92ca4)
• tui: disable ShowElapsedTime to prevent pterm ticker goroutine race (cec32c5)
• update guidance for reindexing the Lumen index (f08c3b9)
• update MinScoreFilter test to use -1 for all-results and 0.5 threshold (562e474)
• use query_only pragma instead of read-only mode for WAL compatibility (34e5123)
• version to 0.0.1-alpha.4 (816452a)

Performance Improvements

• add SQLite write pragmas and chunk column indexes (5bac05f)
• avoid double Merkle tree build by sharing tree between EnsureFresh and indexWithTree (91f3cae)
• combine Stats() into one query, add GetMetaBatch for Status() (80300f9)
• e2e: parallelize all E2E tests and fix openIndexDB race (#45) (1ae2993)
• make Status() DB-only by storing total_files in metadata, remove StaleFiles (59826c2)
• parallelize file reads in merkle.BuildTree with worker pool (02e262e)
• stdio: skip merkle walk within 30s freshness TTL (0e2c7a4)
• stream chunk embed+insert in batches of 256 to bound memory (7441642)
• use RWMutex with double-checked locking in indexer cache (5d1c2b6)

Reverts

• distribute lumen binary via npm optional dependencies (#47) (#50) (4ea2314)

Code Refactoring

• migrate from install/uninstall to Claude Code plugin system (0920f3d)

---

This PR was generated with Release Please. See documentation.