9 min readUpdated September 22, 2026
How to debug an AI-built app
A repeatable method: reproduce it, read the real error, find the layer, change one thing, verify. Written for apps whose code you did not type.
Troubleshooting
Troubleshooting is tracing a visible failure to its cause. Start with a failed request, empty list or deployment, then verify the fix.
Where should I start with troubleshooting?
Start with the guide that matches your question or the error you can see. Each guide below names the problem, gives checks you can run, and links to related reading.
Which symptom should you investigate first?
Choose one action you can repeat: opening a saved record, submitting a form or loading the deployed page. Record what you expected, what appeared instead and the first failing request. A blank screen, empty list and rejected request can look similar but need different fixes.
| Observation | Next check | Avoid |
|---|---|---|
| 401 response | Was a valid credential sent to the correct environment? | Changing ownership rules before checking identity |
| Successful response with an empty list | Do the filter and access policy match this account's rows? | Disabling access rules to make data appear |
| Preview works, live fails | Compare deployed revision and environment configuration | Assuming the latest preview is already published |
| 429 response | Read provider limits and retry instructions | Repeating the request in a tight loop |
How do you know the fix addressed the cause?
Illustrative case: a record saves but disappears after refresh. First check whether the write succeeded; then check whether the next read uses the same account and filter. Do not treat an optimistic screen update as proof that the record reached storage.
- 1Repeat the original failing action with the same input after one targeted change.
- 2Refresh and read the result again. Check a second account only in a safe test environment.
- 3Run the denial case too: a signed-out visitor or different owner should not gain access as a side effect of the fix.
- 4Keep a regression check for the behavior, not merely a screenshot of the repaired page.
Read the troubleshooting guides
8 min readUpdated September 22, 2026
How to troubleshoot an AI-built application
A symptom-to-cause index. Start from what you can see — a blank page, an empty list, a failed login, a dead deploy — and follow it to the layer responsible.
7 min readUpdated September 22, 2026
Why am I getting a 401 error?
A 401 means the server could not identify you. The causes, how to tell which one you have, and how to confirm the fix.
8 min readUpdated September 22, 2026
Why does my app work locally but not after deployment?
The differences between your machine and a deployment — configuration, case sensitivity, build strictness, real users — and how to find which one bit you.
8 min readUpdated October 21, 2026
Why can't my frontend access my database?
Empty lists, permission errors and silent failures between the browser and a hosted database — and how to tell which of the three you have.
6 min readUpdated September 22, 2026
Why isn't my environment variable working?
The variable is set and the code still reads undefined. The six reasons that happens, in the order worth checking them.
7 min readUpdated September 22, 2026
Why is my API key exposed?
How keys end up visible in a bundle or a repository, how to confirm it, what to do in the first ten minutes, and how to stop it recurring.
6 min readUpdated September 22, 2026
Why isn't my GitHub deployment updating?
Your change is committed and the live site is unchanged. Branch, build, cache or domain — how to find out which, in four checks.
7 min readUpdated September 22, 2026
Why is my AI API returning 429?
Rate limits, token limits, concurrency limits and accidental loops — how to tell which one you hit and how to stop hitting it.
Looking for something else? Browse every topic.