Run failed: OpenAI API hits BadRequestError: Error code: 400 - {'error': {'code': 'TooManyHeaders', 'message': 'Your request contains too many headers. Please remove some headers and try again.'}} [Error reference: https://platform.openai.com/docs/guides/

Question

I get the following error when I try to run a GPT-4 model deployment run in MS Promptflow.

Run failed: OpenAI API hits BadRequestError: Error code: 400 - {'error': {'code': 'TooManyHeaders', 'message': 'Your request contains too many headers. Please remove some headers and try again.'}} [Error reference: https://platform.openai.com/docs/guides/error-codes/api-errors]

Everything worked fine yesterday and I didn't change any settings. Does anyone know what could be causing this?

Answer 1

Hey all,

So here is my conclusion from my overall research. I may be wrong as well :)

Recently (maybe a few days back), Microsoft pushed a new feature to validate all the openai api request headers.

• If the request header is more than 8 (>8) it will start throwing this error. So the threshold is 8. If you provide even one more (i.e 9 ) it will be the 400 bad requests.

{ "error"``: {`` "code"``: ``"TooManyHeaders"``,`` "message"``: ``"Your request contains too many headers. Please remove some headers and try again." } }

But this push is not released to all models and regions. So this behaviour can be experienced by differently on different endpoints

So far I can easily reproduce this issue in the below API and region

gpt-4 version 0613, eastus2

Feel free to correct me or provide your own feedback and responses

-Jay

Answer 2

Hi Ryan Edgerley Thanks for the confirmation. Arjun G Jurs, Craig @Luc Elsby Mingyang Xu - Network Christian Wunderlich, The fix has now been made available in production. Please check and let me know if you still see any issues. Thanks Saurabh

Answer 3

This issue now seems to have magically disappeared this morning. I can successfully run prompt flows using GPT-4 across a number of different Azure accounts.