Minigraf 2.0.3 was released on crates.io, bringing a tiny, self‑contained graph database that supports Datalog queries and bi‑temporal time travel. The core library can be added with cargo add minigraf and opens a file via Minigraf::open("data.graph") with zero configuration.
Core capabilities
- Datalog queries – recursive rules let you traverse relationships without joins. Window functions such as
sum,count, andrankare available directly in:findclauses. - Bi‑temporal model – each fact stores a transaction timestamp and a valid‑time interval, enabling "as‑of" queries that reconstruct the exact knowledge state at any past moment.
- Embedded everywhere – the library compiles to native Rust, WebAssembly (browser and WASI), Android, iOS, Python, Node.js, Java, and C. All bindings use a single
.graphfile and require no external server.
Performance and limits
Benchmarks show a point query on 1 M facts takes ~4.4 s, opening the database takes 1.31 s, and peak heap stays around 1 GB. Fact size is capped at 4 080 bytes for file‑backed stores; in‑memory stores have no hard limit. Writes are fsync‑ed by default, but SyncMode::Normal can boost bulk‑load throughput at the cost of durability guarantees.
Known caveats
The current 2.x line has a documented bug where writing two values for the same attribute in a single transaction can collapse into a single value (see issue #371). The fix lands in the upcoming 3.0.0 release, after which v2.x will receive only security patches for 12 months. Also, Minigraf is deliberately not a distributed system – it does not provide clustering, replication, or a client‑server protocol, and is optimized for <1 M nodes.
Licensing and cost
Minigraf is dual‑licensed under the Apache 2.0 (http://www.apache.org/licenses/LICENSE-2.0) and MIT (http://opensource.org/licenses/MIT) licenses. There is no pricing tier; the software is free to use, but the repository does not publish any commercial support options.
When to try it
If your startup needs an embedded graph store that can audit every change – for example, an AI agent that must replay past beliefs or a mobile app that corrects user data retroactively – clone the repo at https://github.com/project-minigraf/minigraf and run the browser visualizer to explore time‑travel before wiring it into your code.
