Closed SinKasula closed 11 months ago
I'm not sure I understand how restarting the application will help when the endpoint returns an error code. Is the application crashing in this situation?
In case where openai endpoint returns unexpected errors, supervisor could restart the request
I'm afraid that's not how supervisors work. They're meant to restart the application when it crashes. That isn't happening here.
AFAIK, the only time it might make sense to automatically retry a request is if you've hit a rate limit and need to wait. In other situations, you need to check why there was an error and modify your request. What kind of error are you seeing, that you want an automatic retry?
As of now, you need to write the code to check for this, and retry if necessary.
@SinKasula I noticed you upvoted https://github.com/restlessronin/openai_ex/issues/70, which is related to this. Please respond in detail about your retry use case on that issue. I will close this issue for now.
You are right, It only helps in cases of limited error cases. Nvm, I'll setup retry logic for cases necessary
Describe the feature or improvement you're requesting
Could you please update the application.ex to include an option to retry when open ai endpoint responds with an error code.
A simple fix can be on this line https://github.com/restlessronin/openai_ex/blob/56be67e48a436e2950003e49eca439e1864f3d36/lib/openai_ex/application.ex#L13
opts = [ strategy: :one_for_one, restart: :transient, max_retry: 3 ]
Supervisor.start_link(children, opts)
Additional context
No response