010 GitHub stars
02Zero external dependencies, embedding LanceDB and using local ONNX embeddings via Transformers.js for local embeddings.
03Ensures persistent storage of memories in a configurable local directory by default.
04Implements hybrid BM25 full-text and vector semantic search with Reciprocal Rank Fusion for optimal recall.
05Automatically downloads and reuses the embedding model on first run, requiring no manual setup.
06Offers 6 core memory management tools: save, search, get recent, get by ID, delete, and list tags.