Built-in AI: smart features, no API keys
Goal. Add real AI features — summaries, chat, classification, even image generation — just by describing them, with no API keys or model accounts.
Add an AI button that turns a rough, messy note into a clean task: a clear title, a suggested due date, and 2–3 tags.
Turn on the backend first (Lesson 2). Lovable manages the model key for you (LOVABLE_API_KEY) and runs the call server-side in an edge function, so no key is ever exposed in the browser. (Same move powers a support-reply drafter, a content summariser, or a CSV classifier — whatever your app needs.)
- 1With Lovable Cloud on, ask for an AI feature like the prompt above. Lovable wires the serverless function and the model automatically — you never create a provider account or paste a key.
- 2The default model is Gemini 3 Flash (preview); ask Lovable to switch to a deeper model (a Gemini Pro or a GPT model) for harder reasoning, or to an image model to generate pictures.
- 3Watch your AI balance: every workspace currently includes a small free AI allowance ($1/month, billed by actual use). Track it in Settings → Cloud & AI balance.
You'll see. An AI-powered feature — messy note in, clean tagged task out — running inside your app with no provider setup.
Cost. Built-in AI runs from a free monthly AI balance (currently $1 per workspace, usage-based, all plans); building the feature also spends chat credits. No external API keys.
Takeaway. Lovable bundles model access — describe the AI feature and it's wired server-side. Just keep an eye on the small AI balance.