# ProvSQL

> ProvSQL is a free, open-source (MIT) PostgreSQL extension that adds
> semiring provenance and uncertainty management to SQL queries. It
> transparently rewrites queries to track which input tuples contribute
> to each result, then evaluates the provenance in any user-chosen
> semiring: Boolean reachability, counting, why/how/where-provenance,
> probability (probabilistic databases, including continuous random
> variables), Shapley and Banzhaf values, temporal validity, and more.
> A companion web UI, ProvSQL Studio, provides interactive provenance
> inspection; the ProvSQL Playground runs the entire system in the
> browser via WebAssembly.

Capabilities have grown substantially since the 2018 VLDB demo paper:
current ProvSQL supports inner and outer joins, subqueries, aggregation
with HAVING, set operations, CTEs including WITH RECURSIVE, data
modifications (INSERT/UPDATE/DELETE), window-function pass-through,
continuous probability distributions, and latent-variable Bayesian
inference (observing evidence conditions latent distribution parameters
into their posterior by likelihood weighting). The authoritative,
current reference is the documentation below.

## Documentation

- [Introduction](https://provsql.org/docs/user/introduction.html): concepts and capabilities
- [Installation](https://provsql.org/docs/user/getting-provsql.html): from source, PGXN, Docker; license
- [Tutorial](https://provsql.org/docs/user/tutorial.html): hands-on first steps
- [Querying](https://provsql.org/docs/user/querying.html): supported SQL features
- [Semirings](https://provsql.org/docs/user/semirings.html): available provenance semirings
- [Probabilities](https://provsql.org/docs/user/probabilities.html): probabilistic-database evaluation methods
- [Continuous distributions](https://provsql.org/docs/user/continuous-distributions.html): random-variable columns
- [Shapley and Banzhaf values](https://provsql.org/docs/user/shapley.html)
- [Where-provenance](https://provsql.org/docs/user/where-provenance.html)
- [Aggregation](https://provsql.org/docs/user/aggregation.html): aggregate provenance
- [ProvSQL Studio](https://provsql.org/docs/user/studio.html): web UI and notebooks
- [Case studies](https://provsql.org/docs/user/casestudies.html): extended worked examples

## Software

- [GitHub repository](https://github.com/PierreSenellart/provsql): source code, README, issues
- [ProvSQL Playground](https://provsql.org/playground/): full system in the browser, no install
- [Docker image](https://hub.docker.com/r/inriavalda/provsql)
- [PGXN distribution](https://pgxn.org/dist/provsql/)
- [provsql-studio on PyPI](https://pypi.org/project/provsql-studio/)

## Research

- [ProvSQL system paper (ICDE 2026)](https://arxiv.org/abs/2504.12058): current description of the system
- [Publications](https://provsql.org/publications/): full list of related research
- [Lean formalization](https://provsql.org/lean/): machine-checked theory behind the system (m-semirings, annotated-database semantics, query rewriting, HAVING provenance, circuits, probability, complexity)
- [Lean API documentation](https://provsql.org/lean-docs/Provenance.html): the formalization itself, from the companion [provenance-lean](https://github.com/PierreSenellart/provenance-lean) library

## Optional

- [Developer guide](https://provsql.org/docs/dev/introduction.html): internal architecture
- [SQL API reference](https://provsql.org/doxygen-sql/html/index.html)
- [C/C++ API reference](https://provsql.org/doxygen-c/html/index.html)
- [Releases](https://provsql.org/releases/): version history