yvgude/lean-ctx v3.5.12

on GitHub

Improved

• RAM optimization: eliminate double tokenization — extract_chunks in bm25_index.rs, artifact_index.rs, and chunks_ts.rs no longer allocates a tokens: Vec<String> per chunk. Token count is computed inline; the vector is set to Vec::new(). add_chunk tokenizes from content once for the inverted index and overwrites token_count from the fresh result. This eliminates one redundant allocation + tokenization pass per chunk during index build.
• MemoryProfile fully wired — The MemoryProfile enum (low / balanced / performance) now actively controls runtime behavior:
  • max_bm25_cache_bytes() respects profile limits (64 / 128 / 512 MB), with explicit user config taking precedence.
  • Semantic cache (SemanticCacheIndex) is skipped entirely when memory_profile = low.
  • Embedding engine loading is skipped in ctx_semantic_search and ctx_knowledge when memory_profile = low.
• Doctor shows active memory profile — lean-ctx doctor now displays the effective memory profile (low / balanced / performance), its source (env / config / default), and what it controls (cache limits, embedding status). Helps users understand and debug RAM behavior.
• MCP manifest regenerated — Updated mcp-tools.json to reflect current tool count (57 granular tools).

Upgrade

lean-ctx update                 # recommended (auto-downloads + refreshes shell hooks)
cargo install lean-ctx          # or
npm update -g lean-ctx-bin      # or
brew upgrade lean-ctx

Note: After upgrading via cargo/npm/brew, run lean-ctx setup to refresh shell aliases. lean-ctx update does this automatically.

Full Changelog: v3.5.12...v3.5.12