Intelligent Debugging

When something breaks in your generated app, you have three reliable shortcuts — no need to hunt through files yourself. The AI reads your code, so it can find and fix the cause from a short description.

Three ways to report a bug

1. Paste the error. Drop the stack trace or console message into the chat; the AI opens the file at the right line and proposes a fix.
2. Point at a file. "There's a bug in orders.controller.js when status is null."
3. Describe the symptom. "The cart total is wrong when discount codes are stacked." The AI traces the data flow and patches the right places.

Give it what it needs

• The exact error text, copy-pasted — not a screenshot.
• What you did right before it broke.
• Expected vs. actual — "should show 3 items, shows 0".
• Any relevant IDs or inputs that trigger it.

Converge in a couple of rounds

• If the first fix isn't quite right, reply with what you now observe — "still crashes, but with a different error: …" — not just "didn't work".
• The AI keeps the previous turn plus the latest file state, so you usually land it in two or three rounds.
• If it drifts, roll back to before the fix and try a more specific description.

Preventing bugs in the first place

• Ask a Testing agent to add tests for critical flows so regressions surface early.
• Request input validation up front — "validate the checkout form and show clear errors".
• Keep changes small; a bug in a tight edit is far easier to trace than in a giant one.

Common mistakes to avoid

• Describing the fix instead of the problem. Tell the AI what's wrong and let it choose the fix.
• Pasting a screenshot of an error. Paste the text — the AI can't copy from an image reliably.
• Piling new features onto a broken build. Fix first, then continue.

FAQ