I automated most of my DB Admin job — then open-sourced it!

2am. Checkout's down. I'm on the floor, laptop on a pizza box, praying to EXPLAIN ANALYZE. Again.

8 years I did this. Every slow query, every "how many signups last week," every migration waiting on my blessing. People called me indispensable. I knew the truth — I was just the last guy who remembered how it all fit together. A single point of failure with a heartbeat.

Then one night it hit me: the bottleneck was never SQL. Anyone can write SQL. The bottleneck was context — what "active customer" means, that revenue is really revenue minus refunds. All of it lived in one place. My head.

So I stopped hoarding the map. I built something to carry it instead — watches the workload, rewrites the slow queries, answers the BI questions, and checks every AI-written migration before it nukes a 41M-row table.

The pager went quiet. I got my nights back.

Two kinds of people, boss — one hoards it for job security, the other gives it away and goes build bigger. I open-sourced the whole thing. I call it deepsql.

The bits this crowd will actually care about:

• Everything runs in your own infra — your VPC, your credentials, connects read-only to a replica.
• BYO model — point it at OpenAI, Azure, or any OpenAI-compatible endpoint (including a local one). No forced vendor.
• Apache-2.0, full source. Docker Compose, up in ~15 min.
• Runs on any Linux box / EC2 / container. RBAC, query policies, & audit logs built in.

Install (clone + compose, so you can read it before you trust it):

git clone https://github.com/DeepSQLAI/deepsql
cd deepsql && cp .env.example .env
docker compose up -d

Link: https://deepsql.ai/
Code: github.com/DeepSQLAI/deepsql

Stop doing the same work twice. Get the map out of your head. Then go live. 🔥

Roast it, break it, PRs welcome — that's why I'm here.