AI features in your app

Vorx apps can use AI at runtime through a built-in connector. You describe the feature in a prompt; Vorx wires up the model, the backend, and the UI.

This is AI inside the app you're building — separate from the Vorx agent that builds it for you.

What you can add

• Chat assistants — a helper that answers questions about your app's data.
• Summaries — condense long text into a few lines.
• Classification and tagging — sort incoming records automatically.
• Semantic search — find by meaning, not just keywords.
• Image generation — create images from a prompt.

Just ask, for example:

Add a button on each support ticket that summarises the conversation and suggests a reply.

No keys to manage

The AI connector is included, so there are no provider accounts or API keys to set up for the common cases.

Bring your own model

Prefer a specific provider or model? Add its key under Settings → Cloud → Secrets and tell Vorx to use it. The call runs server-side in a backend function, so your key never reaches the browser. See Secrets.

Keep it grounded

AI features work best against real data. Point the feature at the records it should read — "answer only from this customer's orders" — so responses stay accurate and scoped. An assistant with access to everything is both less useful and harder to reason about.

Scope it to the signed-in user's data too, so an assistant can't summarise records the person asking isn't allowed to see. See Authentication.

Watch what it does

Model calls from your app run server-side, so you cannot watch them from the app yet (logs and monitoring) — paste the error into chat when an AI feature returns something unexpected.

Next steps

• Backend functions — where AI calls run
• Secrets — bringing your own key
• Plans and credit usage — how usage is metered