Skip to main content

​
Build does not start

Check the project state, required user inputs, available AI credits, any estimate or confirmation step, source availability, and required capability or integration configuration. If the project is already processing another build, wait for its visible state before starting another one.

​
Build is waiting for input

Waiting is a distinct state, not a failure. Open the run timeline and look for a pending user decision, capability approval, missing configuration value, or connector prerequisite. Complete the request and resume the build when the dashboard provides that action. Do not create duplicate builds while the current run is waiting for input.

​
Build failed

Use the visible run timeline and error summary to classify the problem:
  • Generation problem: the requested change could not be produced.
  • Configuration problem: a project value, capability, or integration is missing.
  • Source problem: the selected source or revision is unavailable.
  • Validation problem: the result did not satisfy project checks.
  • Application problem: the generated application builds but has a runtime or behavior issue.
Retry after correcting the cause. If the error is unchanged, collect the project ID, run ID, revision, time, and visible error before contacting support.

​
Build completed but the change is missing

Confirm the latest revision and build summary, then verify that the requested change was included in that run. Check whether the preview is using that revision and whether the request required a capability that is not admitted for the project.

​
Build appears stuck

Use the visible state rather than elapsed time alone. The implementation distinguishes queued/pending, running, completed, completed with limitations, blocked, failed, canceled, and continuation-required outcomes. A run that is blocked or waiting for input needs the requested action; a failed run needs diagnosis before retrying. Related: P2HS AI, Codegen lifecycle, Billing and AI credits.