Engrava introduces local graph memory database for AI agents
TL;DR. Engrava released a standalone, embedded graph memory database built on SQLite for AI agents, offering local and tamper-evident storage. - The database provides thought CRUD, knowledge graphs, and hybrid search capabilities without external service dependencies. - It supports various embedding backends, including local models and OpenAI-compatible APIs for semantic search. - Benchmark results are published and reproducible, demonstrating its architecture without LLMs in the memory layer.
- Engrava is a standalone, embedded graph memory database for AI agents, built on SQLite.
- It features thought CRUD, edge-based knowledge graphs, hybrid search (embedding-based and full-text), and a tamper-evident journal.
- The system operates with zero external service dependencies and includes options for various local and API-based embedding models.
- Performance benchmarks are publicly available and reproducible, confirming its design without LLMs in the core memory architecture.