FederIQ

Federated SQL over any source, in a single Rust binary.

FederIQ is an open-source federated query engine built on top of DuckDB. One command attaches a Postgres database, a Parquet directory on S3, a CSV on disk, and an HTTP JSON API — and lets you join them in plain SQL.

SELECT u.name, SUM(o.total) AS revenue
FROM postgres_users.public.users u
JOIN events.orders o ON o.user_id = u.id
GROUP BY u.name
ORDER BY revenue DESC;

Why not Trino?

Trino is the incumbent. It's also ~4 GB of JVM, a cluster to deploy, and a DBA to operate. FederIQ is a single ~20 MB binary:

Why not DuckDB alone?

DuckDB is incredible but single-node and single-process. FederIQ adds:

• Catalog-driven federation — declare sources in YAML, query from any client.
• Policy engine — column masking, row filters, region pinning enforced at query-rewrite time.
• HTTP service mode — federiq serve with bearer auth + TLS + Prometheus metrics.
• Cache layer — memory, disk, and Redis backends with hooks for a predictive prefetcher.
• First-party Python and TypeScript clients.

License

Apache 2.0. The core will always be Apache 2.0 — hosted and enterprise features may be licensed separately in the future.