Generation Failed
AI generation runs as a cloud agent. It can fail because the agent cannot reach the target site, your workspace is out of credits, a referenced secret is missing, or the feature instructions do not give the agent enough context.
First Checks
- Open the failed generation in TestVibe.
- Read the visible failure message and any completed sections.
- Confirm the target site URL is correct and reachable.
- Confirm your workspace has enough credits in Settings → Billing.
- Confirm any
{{secret:NAME}}or{{var:NAME}}tokens in the feature match entries in Settings → Variables & Secrets. - For private apps, confirm the project's tunnel device is online.
Common Causes
| Cause | What to check |
|---|---|
| Insufficient credits | Check your balance in Settings → Billing and top up if needed. |
| Site unreachable | Confirm the URL; for private apps, check the tunnel device. |
| Secret token not found | Add the named entry in Settings → Variables & Secrets. |
| Login or test data missing | Add secrets and prerequisite steps. |
| Vague feature instructions | Improve the Gherkin scenario with visible actions and expected outcomes. |
| Agent or model misconfigured | Review the agent, model, and reasoning effort in Settings → AI Model. |
When Output Is Partial
Sometimes generation fails after producing useful notes, a feature draft, or a partial test file. Review the output before retrying. It may tell you exactly which selector, page, credential, or requirement was unclear.
Retry Or Edit First
| Situation | Best next action |
|---|---|
| Temporary network or environment issue | Retry after the environment is healthy. |
| Missing secret or credit | Fix setup, then regenerate. |
| Feature is too broad | Split it into smaller scenarios, then regenerate. |
| Test intent is unclear | Edit the Gherkin, then regenerate. |
| Generated code is close but wrong | Edit the test or refine instructions before regenerating. |