Decision model

✓ functional

An optional fast cloud model that answers tiny pick-one questions so several features can offer a smart first guess.

Several SQLly features come down to very small questions: which of these charts? do these two tables join on customer_id? does this column look like emails? The decision model is an optional cloud model dedicated to exactly that - small typed questions (pick one of several, rate on a scale, yes or no) answered in a fraction of a second, so AI features can make quick in-loop judgments.

It cannot write. The model only answers the small structured questions SQLly asks; it never produces text or SQL of its own.
Off by default. Nothing is enabled and nothing is sent until you turn it on, and every feature falls back to your regular AI backend when it is off.
What gets sent. Enabling it means your questions, schema names, and SQL text go to TypeSafe AI; query result data is sent only if you separately allow it.
What it powers. Suggested first charts, inferred joins in the Query Builder, personal-data detection, risk grading in Schema Compare, and attention ranking in the Activity Monitor.

Turning it on

Open Settings › AI › Decision model, turn on Enable decision model, paste your API key and choose Save key, then press Test to confirm the service answers. Leave “Allow result data to be sent” off unless you want result rows included in these checks.

Fast, narrow, optional. A fraction of a second per question, a closed set of answers, and a switch you can flip off at any time - every feature works without it.