Serverless SQL over open formats.
Run SELECT queries over Parquet, CSV, JSON and Iceberg in object storage, on DuckDB or Trino, and pay only for the bytes you scan.
Async query executions
Submit a query, poll its status and stats, then page results or stream them as Parquet/CSV — a managed submit→poll→results lifecycle.
Interactive sync path
Run queries inline for SQL editors and dashboards, returning the first result page immediately through the same metered executor.
Pluggable engines
Dispatch per query to embedded DuckDB, a distributed DuckDB worker pool, or Trino — one SQL surface, three execution models.
Workgroups with quotas
Group queries by team with concurrency limits and a monthly bytes-scanned cap, enforced at submit time.
Bytes & rows scanned metering
Actual scanned volume is measured per query and reported to VXBilling — you pay per query, not for idle compute.
Results in object storage
Every execution materializes result.parquet and result.csv under a tenant/month/id prefix in a results bucket.
Open formats & catalog
Query read_parquet, read_csv, read_json and iceberg_scan sources with hive-partition pruning, managed via a source catalog.
Explain, cancel, history
Inspect plans, cancel running queries, and browse history and saved queries.