// the find
Pometry/Raphtory
Scalable graph analytics database powered by a multithreaded, vectorized temporal engine, written in Rust
Raphtory is an in-memory temporal graph database written in Rust with Python bindings, built specifically for time-aware graph analysis — querying how a graph looked at time T, rolling windows, motif detection across time slices. It targets data scientists doing network analysis on things like transaction graphs, communication networks, or social graphs where the order and timing of edges actually matter.
The temporal model is first-class, not bolted on — `graph.at(t)`, `graph.rolling(window=1)`, and `graph.layers()` are core primitives, not afterthoughts. The vectorized Rust core means hundreds of millions of edges fit in a laptop with real performance (there's a benchmark page that runs on every push). The Python API is clean enough that loading a CSV and running PageRank is genuinely five lines. Out-of-disk persistence via the subscription tier uses a custom WAL + segment store (the `db4-storage` crate) that avoids the usual mmap pitfalls.
628 stars for a project with this scope suggests it hasn't broken out of the academic graph-analysis niche — the citation is a JOSS paper, and the examples lean heavily on Lord of the Rings and Enron, which signals research-tool positioning more than production use. The on-disk persistence and clustering story is behind a commercial subscription with no public pricing, so the open-source version is bounded to what fits in RAM. The GraphQL server is convenient for exploration but thin — no authentication, no access control, and the schema is essentially a thin wrapper over the graph API rather than something you'd expose externally. The GPL v3 license will block most commercial adopters from embedding it.