01MCP-native interface for zero-code agent integration
02Single PostgreSQL 17 database with pgvector and Apache AGE for ACID transactions across semantic, relational, and graph data
03Hybrid search combining vector similarity, BM25 full-text, and knowledge graph traversal
04Token-budgeted context output, returning compiled documents instead of raw memory lists
05Local embedding model support by default, eliminating API key dependencies
061 GitHub stars