01LLM-driven consolidation that merges duplicates, detects contradictions, promotes facts, and generates summaries during idle times.
02Graph-primary data model using typed edges to represent complex relationships between memories.
030 GitHub stars
04Local-first operation with configurable LLM backends including Ollama, Claude Code, and OpenAI-compatible endpoints.
05RAM-resident, single binary with zero runtime dependencies for sub-5ms latency.
06Hybrid retrieval pipeline combining lexical, vector, RRF, graph expansion, and cross-encoder reranking.