Skip to content
Pivot is in early development and is not production ready. See Project status.

Introduction

Pivot is a high-performance analytics engine that runs on open data formats. It delivers low-latency queries and high concurrency without replicating data into a dedicated database or proprietary format.

Pivot architecture Object storage holds pivotlake tables in Delta Lake and Iceberg formats. A Pivot cluster reads and writes them and serves SQL clients over the Postgres wire. Agents, each embedding its own pivot open, and third-party engines such as Snowflake, read and write the same files directly without going through the cluster. Object storage S3 · GCS pivotlake events Delta Lake format part-00000-3f7a….parquet part-00001-9c21….parquet orders Delta Lake format part-00000-3f7a….parquet part-00001-9c21….parquet sessions Iceberg format part-00000-3f7a….parquet part-00001-9c21….parquet read + write read + write read + write Pivot cluster node 1 node 2 node 3 Agents Bash(pivot open s3://pivotlake) └ opened 3 tables Bash(pivot open s3://pivotlake) └ opened 3 tables Other engines Snowflake Databricks DuckDB Spark Trino Polars Postgres wire SQL clients backend psql BI tools

Key features

Section titled “Key features”
  • Fast - Written in Rust and built on state-of-the-art columnar execution techniques, including morsel-driven parallelism, SIMD, NUMA-aware execution, and cache-conscious aggregation and joins, in addition to a few novel additions (Why is Pivot fast?).
  • Scalable - With object storage as its backing store, Pivot can be scaled up, down, or to zero almost instantly.
  • Portable - Pivot can run both as a server serving backends and clients, or as a local engine where users and agents query the source of truth directly, allowing local ad-hoc and agentic analytics to share the same engine and architecture as traditional dashboards and in-app analytics
  • Open - Pivot is open source and built on open data formats (Delta Lake and Apache Iceberg). This means you can use Pivot with data already stored in your data warehouse, while data ingested by Pivot remains accessible to other query engines.

Project status

Section titled “Project status”

Pivot is in early development and is not production ready. It is meant for evaluation, experiments and local analysis today.

  • Breaking changes. SQL surface, configuration keys, wire behaviour and the on-disk layout change between releases, with no migration path and no deprecation period. Expect to recreate a deployment rather than upgrade it.
  • Gaps in coverage. Whole areas of SQL are unimplemented. What is supported today is listed under SQL reference; assume anything not listed there is missing.

Where to start

Section titled “Where to start”
  • Quickstart runs a server and issues a first query.
  • Architecture explains how a query becomes work across the Dispatch pool.
  • Roadmap shows what is available, in progress, and planned.
  • SQL reference documents commands, functions, operators, and data types with examples.
  • CLI & server reference covers CLI commands, configuration, datastores, authentication, and system tables.