But I do not receive any information but instead: data: {“warning”:“This model version is deprecated. Migrate before January 4, 2024 to avoid disruption of service. Learn more https://platform.openai.com docs/deprecations”,“id”:“*************object”:“text_completion”,“created”:1691094783,“choices”:[{“text”:" (",“index”:0,“logprobs”:null,“finish_reason”:null}],“model”:“text-davinci-003”}
Problem is that I cannot change the model right now, because I have to use davinci completions for the next 2 months. Problem: I have disruption of service with this error message. Does anyone have an idea how to work around this problem?
So the model is still running until January 4, but how do I get the response from the server?
You should look at the HTTP code, where here it would be “200 success”. Also, feed a few errors like bad API keys, bad models, or huge max_tokens, and you will see the messages for those don’t start with “warning”.
I completely understand your frustration. I, too, recently faced a similar issue with the text-davinci stream API in my application which has a large user base. The constant deprecation warning was causing significant disruptions (It stopped working).
A solution to this entailed moving all of these services to the Chat APIs, which was quite a significant task. Unfortunately, this seems to be the only workaround as OpenAI is planning to decommission the use of the current model version.
The sudden implementation of the warning message resulted in many similar instances of confusion and trouble with functionality.
Hope this insight helps and you’re able to work through this disruption swiftly.
Formerly, I merely used string-handling from the console log, but now that I have switched to parse, I was able to shut the problems off. The warning now does not effect me, but it is still showing up, even in instances with “200 success”.
Yeah it will show up as 200. Not sure why it was suggested, 200 just means the server understood your request and was able to respond. A warning is a proper response, even error messages where the dev has not specified a response type will be considered 200.
As for the data, if its not affecting your code, I’d just ignore it and log all warnings on the backend. If it is affecting it and for some reason you need the whole JSON, you can just delete the “warning” key before using it.