microsoft
Add pre-flight API connection checks
Do you need to file an issue?
- [X] I have searched the existing issues and this feature is not already filed.
- [X] My model is hosted on OpenAI or Azure. If not, please look at the "model providers" issue and don't file a new one here.
- [X] I believe this is a legitimate feature request, not just a question. If this is a question, please use the Discussions area.
Is your feature request related to a problem? Please describe.
We frequently receive issue filing from users with various failures during the pipeline that when traced back to an original error are the result of some connectivity issue such as an invalid key or other permissions problem. Unfortunately, this may manifest late in the pipeline and be revealed by seemingly unrelated errors, which are difficult to diagnose. A fail-fast approach would send a test ping up front and exit pipeline execution immediately in the event of a failure.
Describe the solution you'd like
No response
Additional context
No response
This should be done for any models/endpoints used in case there are config issues (completion, embeddings, custom step config, etc.). Could be part of a startup pipeline validation step.
Note that this feature will need to run up-front and synchronously. Part of our difficulty in catching exceptions and failing is that many pipeline steps run async and we do not have a global manager that can tell running steps to abort if a parallel step fails.