Skip to content

Troubleshooting

403 error when updating the PR description or title

This means the GitHub Action's token isn't allowed to write to the repository. Grant the necessary permissions to the GITHUB_TOKEN secret at <your_repo_url>/settings/actions, under Workflow permissions — select Read and write permissions.

Labels aren't being applied even though enable_labels: true

Labeling needs issues: write in addition to pull-requests: write (labels live on the issue object in the GitHub API, even for pull requests). If your workflow sets an explicit permissions: block rather than relying on the repository-wide default, add it there:

permissions:
  pull-requests: write
  issues: write

The Action fails with a 400 error from OpenAI mentioning temperature or max_tokens

This should not happen with the models this Action's openai_model default targets, but if you've pointed openai_model at a newer "reasoning" model family not yet in the REASONING_MODEL_PREFIXES list in pr_description/llm.py (currently gpt-5*, gpt-6*, o1*, o3*, o4*), the Action may still send an unsupported temperature/max_tokens combination for it. Open an issue (or a PR extending that list) with the model name.

The generated description/title/labels seem cut off or missing entirely

  • Check max_tokens — this bounds how much the model is allowed to generate for its response. A very small value can truncate the output.
  • Check max_diff_tokens — a large pull request may have its diff trimmed before it ever reaches the model. Raising this, or narrowing exclude_patterns, gives the model more (or more relevant) context.

Structured features (title/labels/breaking-change) silently don't come through

If the OpenAI/Azure deployment doesn't support structured outputs (an older Azure API version is the most common cause), the Action falls back to a plain-text description for that run and logs a warning rather than failing — check the workflow run's logs for "Structured PR content request was rejected by the API".