DocumentationNextHow a test run works
Builds
Automatic repair
What happens when a build fails.
When a build fails because of the code, such as a compile error, a dependency problem or a missing or invalid descriptor, the agent reads the diagnostics, changes the files and builds again. It tries up to 3 automatic fixes per run; each fix is saved as its own snapshot.
Timeouts and problems on our side are not repaired by the agent. If the project still fails after the last attempt, you see the failing step and the last error, and you can continue in chat with Fix mode.
