
DecisionBox
Autonomous AI Discovery For Your Data — Open Source
100 followers
Autonomous AI Discovery For Your Data — Open Source
100 followers
DecisionBox for PostgreSQL





100 followers
100 followers







Launched on April 29th, 2026

Launched on April 24th, 2026
DecisionBox
Hey PH!
DecisionBox is an autonomous AI discovery agent; connect it to a database, point it at a domain pack, and it writes its own SQL, validates every finding against the data, and ships a ranked backlog of insights and recommendations. No prompting, no question-writing.
Today we're adding PostgreSQL to the list of supported databases.
Same agent, same loop, same domain packs as our BigQuery, Redshift, Snowflake, and Databricks integrations. What's specific to this provider:
▸ Read-only by Postgres role. USAGE on the schema, SELECT on the tables you opt in to, nothing more. The agent cannot reach what the role does not grant.
▸ TLS required by default. sslmode=require out of the box; verify-ca and verify-full both supported when you provide a CA bundle.
▸ Two ways in. Field-by-field (host, port, db, user, schema, sslmode) or a full DSN. The DSN path works with whatever your managed Postgres handed you — RDS, Supabase, Neon, Railway, Fly.io, Aiven, Crunchy, all of them.
▸ Capped pool, per-query timeout. Sensible defaults out of the box, configurable per project.
▸ No writes, ever. No DDL, no DML. The role you give the agent is the second layer of enforcement.
Every line of the Postgres provider is in the public repo, AGPL v3.
Same agent runs against BigQuery, Redshift, Snowflake, Databricks, and MSSQL. If your data moves, your DecisionBox install moves with it.
Happy to dig into the role grants, the auth options, or how the agent loop produces a discovery in the comments.