Cloud Noce AI debugger
The AI debugger turns a failed build’s logs into an understandable explanation and suggested actions. It is available from the failed build page and the build explanation API. If the AI service is unavailable, the raw logs remain your primary troubleshooting source.
What does it analyze?
The debugger examines the error recorded in that build: the failed stage, package manager/compiler message, command, detected framework and other clues in the logs. It helps you identify the cause sooner; it does not automatically modify the repository.
Open the failed build
In Deployments/Builds, select the latest build with failed status. First inspect the end of its logs and the first actual error.
Request an AI explanation
Select Explain. The explanation includes a likely cause and several suggested checks or fixes.
Compare the suggestion with your source
Check the runtime version, lockfile, root directory, Dockerfile, build command and build-time variables in the repository. The AI suggestion may not fully account for your application architecture.
Make a small, reversible change
Commit and deploy the fix, then compare the new build with the previous one. Do not print secrets in prompts, commits or logs.
Common errors
- Incompatible dependencies or runtime versions;
- An outdated lockfile or private package without credentials;
- An incorrect root directory or build context in a monorepo;
- A required variable defined only for runtime when the build needs it;
- An incorrect build output or start command;
- A Dockerfile attempting to copy a required file outside its context.
When the build succeeds but runtime fails
The AI debugger handles build errors. For runtime crashes, check Logs, process status, ports, variables, add-on connections and memory, and use the troubleshooting guide.