unified-memory-indexer

Hybrid semantic search combines keyword BM25 and vector search to index code, notes, and conversations into a single SQLite database, enabling low-latency retrieval of relevant information across all sources simultaneously.

Yes, SQLite supports fast vector and full-text search by utilizing sqlite-vec for semantic retrieval and FTS5 for keyword matching, delivering query times under 20ms for personal knowledge bases and code repositories.

Build a unified memory index for multiple data sources by running the command 'unified-memory build --all' via the Python API or CLI, which synchronizes conversations, code, and notes into one searchable database.

The best way to search across conversation logs and code snippets is using a unified semantic memory indexer, which cross-references fragmented data sources to find specific concepts quickly through combined vector and keyword retrieval.

Yes, incremental synchronization works with SQLite-vec and FTS5 by updating the unified memory index incrementally, ensuring cross-reference boosting and fast retrieval without rebuilding the entire database from scratch.