Scott Campbell
Assembly line analytics, text to SQL console

Automotive Analyst

Ask the assembly plant warehouse a question in plain English. Your model writes PostgreSQL, the server validates it through read-only guardrails and runs it, and you always see the query behind the answer.

Read only, liveFactory dashboard ↗
The question
Can a plain English question become SQL over a factory warehouse without trusting the model?
The data
The synthetic warehouse behind the factory dashboard. No proprietary or employer data.
The result
In testing the guardrail blocked 43 of 43 attack queries and accepted 16 of 16 legitimate ones. Read the case study.
Bring your own key

Connect a model to run queries

Pick a provider and paste an API key. It is stored only in this browser tab and sent directly to the provider, never to this site's server, and it is gone when you close the tab.

Need one? Get a Claude key ↗

Ask the warehouse

Generate
Guardrail
Execute

How a question becomes an answer

  1. 01Your key
    Your provider key stays in this browser tab and calls Claude, OpenAI or Gemini directly, never our server.
  2. 02Generate
    Your model writes one PostgreSQL SELECT, grounded in the star schema served by the API.
  3. 03Guardrail
    The server validates it: SELECT-only, single statement, allow-listed tables, LIMIT injected.
  4. 04Execute
    Run as a read-only role inside a read-only transaction with a statement timeout. Self-corrects once on error.

Stack: FastAPI, PostgreSQL, Next.js and Recharts.