The questions everyone asks before connecting.

Your database never does — the DBMind cloud never receives credentials or a connection. In cloud mode each question sends the schema metadata of your approved tables and, after your own server has run the guarded query locally, its capped, sanitised result rows so the model can phrase the answer. In local mode nothing leaves at all.

By default: hashes, token counts, latency and status metadata. You can additionally keep a conversation log (each question and the answer the model gave) for auditing — it is visible in your dashboard and can be switched off by the platform. Schema metadata and result rows are never persisted.

It writes SQL, but it can never run it. Every generated statement must pass SqlGuard on your server — exactly one read-only SELECT on approved tables, no denied columns, no SELECT *, an enforced LIMIT — and it executes on your local read-only connection under a database-side statement timeout. The cloud never connects to your database.

Anything Laravel can talk to — MySQL, MariaDB, PostgreSQL and SQLite are first-class. During install you choose which tables the AI must NOT answer from; everything else is approved and auto-configured from the schema, and sensitive columns like passwords and tokens are always blocked.

Yes. Point the package at your own Ollama server with a SQL-capable model instead of the DBMind cloud, and writing, guarding, executing and answering all run on hardware you control.

Tokens are shown once at creation and stored only as SHA-256 hashes. You can disable a token temporarily or revoke it permanently from the dashboard, and every call it makes is metered and logged.

The gateway returns a clear quota-exceeded response and your application keeps working — it simply tells users it cannot answer until the quota resets or you upgrade the plan.

Plans are billed monthly through Qi Card. The free tier needs no card at all, and you can cancel anytime from the billing page.