hydradb vs latticedb
Object-store-native distributed graph DB in Rust: graph on SlateDB over S3, snapshot-consistent OpenCypher, GraphBLAS traversal, Neo4j Bolt 5.x compatibility, disaggregated data nodes and indexers. — versus — Embedded single-file graph database in Zig: graph traversal, HNSW vector search and BM25 full-text in one query language, plus a durable event log — built for Graph RAG and local agent memory.
Opposite ends of the same need: latticedb is an embedded single-file graph+vector+BM25 store, hydradb is a distributed S3-backed cluster — pick by whether the graph outgrows one process.
| hydradb | latticedb | |
|---|---|---|
| Stars | 3.1k | 667 |
| Forks | 660 | 30 |
| Language | Rust | Zig |
| License | AGPL-3.0 | MIT |
| Last activity | 23 days ago | 11 days ago |
| Topics | knowledge-graphs, storage | knowledge-graphs, rag, local |
| Curated connections | 4 | 5 |
hydradb — the curator's take
Consider it when you want a Neo4j-shaped graph (Cypher, Bolt drivers) but with cloud-native durability — S3 is the only copy that matters, and compute nodes are disposable and scale independently. That makes it a fit for large agent-memory or knowledge graphs on a budget where you'd rather not run a stateful Neo4j cluster. Avoid it for embedded or single-process use (latticedb, pggraph or plain SQLite are simpler), for early-stage projects that need battle-tested drivers and tooling, and note the AGPL-3.0 licence if you embed it in a product. Still 0.x; verify with a round-tripped write, as its own README insists.
latticedb — the curator's take
SQLite's shape applied to the three retrieval modes agents actually need. One file, no server, one query layer where a Cypher-style MATCH can filter on vector distance and a full-text match in the same statement, and one WAL that graph writes and named event streams share — so your changefeed is transactional with the graph. Quoted 0.13µs node lookups and 0.83ms vector search at 1M vectors, with CLI, Python, TypeScript, Java and Go bindings. Reach for it when relationships are the point and you were about to bolt a vector DB onto a graph DB. Do not reach for it for multi-writer or multi-machine work: it is explicitly single-writer, one owning process, one machine. Also young — Zig, 567 stars, bundled native libs per platform, and the built-in `hash_embed` helper is a deterministic placeholder, not an embedding model.