Skip to main content

​
Cloud

  • Retry with a smaller source and a clear, focused prompt.
  • Reduce the requested slide count when a large job repeatedly times out.
  • Confirm uploaded files finish processing before generation begins.
  • For API tasks, preserve the request identifier and final response when contacting support.

​
Self-hosted

  • Verify the selected text provider URL, model name, and credentials.
  • Confirm the model supports the structured output required by the release.
  • Check container logs for provider, schema, timeout, or memory errors.
  • Simplify the prompt and reduce slide count to separate model problems from capacity problems.
  • Confirm /app_data is writable and has sufficient space.
If generation completes but the result is weak, improve the source, audience, goal, and outline before changing infrastructure.