Skip to content

PGEverything

One PostgreSQL image with everything you need to build a modern AI Foundry.

github.com/MEGALab/pgeverything

Nine databases and services is a lot to run. Postgres, MongoDB, Neo4j, InfluxDB, Redis, Pinecone, Elasticsearch, Auth0, plus a cache. PGEverything is one PostgreSQL 16 server, one port, nine workloads.

Everything is a Postgres extension, so it's all transactional, backed up together, and queried in one SQL dialect.

  • --one-port

    SQL, documents, graph, time-series, queues, vectors, cache, full-text search, and auth. One container, one connection string.

  • --transactional

    Every workload lives in the same database. One backup. One restore. No syncing between systems.

  • --self-hosted

    make build && make up. Runs on your hardware, offline, with your data.

what's inside

Capability Powered by
SQL Core PostgreSQL 16
NoSQL / document Native JSONB + GIN
Graph Apache AGE (openCypher)
Time-series TimescaleDB
Pub/Sub pgmq + LISTEN/NOTIFY
Vector pgvector + pgvectorscale
Key-value cache pgcache (jsonb values, TTL)
Full-text search core tsvector + GIN
Auth / RLS pgauth (JWT) + Row-Level Security

Plus pg_cron, pg_trgm, pgcrypto, pgjwt, pg_stat_statements, and optional PostGIS.

quick start

make build      # build the image
make up         # start it, wait for healthy
make smoke      # verify all nine capabilities
make shell      # psql in

Also included: per-tenant database provisioning, streaming read-replicas, an encrypted secrets vault, and an expert agent for Claude Code. Full docs are in the repo.

status

Early and in active development.

$ view-source $ pay-it-forward