Composable, open-source data infrastructure built in Rust. Each product stands on its own — and they share a common philosophy: fast, embeddable, and built for the AI era.
The database that speaks ORM natively
A relational database engine that understands your entities, relations, and constraints natively — no N+1 queries, no impedance mismatch. Fetch entire object graphs in a single round-trip.
For backend & full-stack engineers
$ cargo install ormdb-server The SQLite for AI agents
Programmable memory that agents query with sandboxed Lua. Instead of a fixed vector-search API, agents write their own retrieval, ranking, and filtering logic — safely.
For AI engineers & agent builders
$ pip install liath They're independent, and they compose: ORMDB for your application's data, Liath for your agents' memory. Use this matrix to pick.
| If you need… | Reach for | Because |
|---|---|---|
| Your application data layer | ORMDB | Relational entities, graph fetches without N+1, safe migrations |
| Per-agent memory | Liath | Embedded store the agent queries with sandboxed Lua |
| Semantic / vector search over app data | ORMDB | Vector, geo, and full-text search built into the engine |
| Retrieval logic more expressive than search(query, k) | Liath | The agent writes its own filtering, ranking, and re-ranking |
| Real-time features & cache invalidation | ORMDB | Change streams (CDC) with version tokens |
| Local-first / edge with zero infrastructure | Liath | Single dependency, no server to run |
See the FAQ for comparisons against Postgres and hosted vector databases.