Skip to content

Velr database

The graph engine behind AI-native knowledge layers

Velr is a lightweight, blazing-fast Rust database that unifies openCypher graphs, SQL, vectors, and time-series on top of SQLite. It is built for edge systems, desktop apps, mobile runtimes, cloud services, and back-office intelligence layers.

Use Velr directly as an embedded database, or build domain verticals on top when the application needs graph-native semantics, fast local queries, and agent-ready context.

Learn More Velr on crates.io Velr on PyPI 

velr = Velr.open("file:PlantA.db")

velr.subscribe(
    Subscription.topic("claims.anomaly")
        .profile("velr:profile:industry4")
        .major(1)  # "1.x"
)

while True:
    velr.sync()
    df = velr.to_polars("""
MATCH
  (c:Claim {kind:'anomaly'})-[:ABOUT]->(a:Asset)
WHERE c.ts > timestamp() - 5 * 60 * 1000
RETURN c.claim_id AS id,
  a.asset_id AS asset,
  c.confidence AS conf
ORDER BY conf DESC
LIMIT 20
    """)
    handle_anomaly(df)

One database, several intelligence shapes

Combine graph traversals, relational queries, vector retrieval, and time-series context without turning every app into an integration project.

Built close to the workload

Run at the edge, inside desktop tools, on mobile, in cloud services, or as the knowledge layer behind operational back-office systems.

Agent-ready by design

Give agents fast local context, structured relationships, and domain-specific retrieval without hiding the data behind slow glue code.