pomazanbohdan/memory-mcp-1file

What's Changed

B1: Composite index + safe parameter binding for symbol search

• Added idx_symbols_type_name composite index on code_symbols(symbol_type, name) in the schema. Filtered symbol lookups (e.g. search by type) now use the index instead of a full table scan.
• Changed search_symbols in surrealdb.rs to bind limit and offset as named query parameters ($limit, $offset) rather than interpolating them directly into the SQL string. This follows the same safe pattern already used for project_id, symbol_type, and path_prefix.

B2: Remove problematic fulltext index on code_chunks

• Removed idx_chunks_fts (FULLTEXT / BM25) from schema.surql. This index was not used in any hot path and caused startup errors on existing databases.
• Added an explicit REMOVE INDEX IF EXISTS idx_chunks_fts ON code_chunks at startup (before schema application) so existing installations are fixed on first run without requiring a manual database reset.

B3: Skip noise chunks during AST-based indexing

• Added MIN_OTHER_CHUNK_LINES = 3 constant in chunker.rs.
• AST nodes classified as ChunkType::Other that span fewer than 3 lines are now skipped during chunking. This eliminates trivial single-line or two-line noise nodes (punctuation, use statements, stray semicolons) from the indexed corpus, improving the signal-to-noise ratio of code search results.

Full Changelog

https://github.com/pomazanbohdan/memory-mcp-1file/compare/v0.4.9...v0.4.10