Can we access request headers (OpenAI) ?

[kantandane]

Hi, im wondering if pydantic-ai allows to access request response headers ?

OpenAI for example provides informations in headers like so:

Field	Sample Value	Description
x-ratelimit-limit-requests	60	The maximum number of requests that are permitted before exhausting the rate limit.
x-ratelimit-limit-tokens	150000	The maximum number of tokens that are permitted before exhausting the rate limit.
x-ratelimit-remaining-requests	59	The remaining number of requests that are permitted before exhausting the rate limit.
x-ratelimit-remaining-tokens	149984	The remaining number of tokens that are permitted before exhausting the rate limit.
x-ratelimit-reset-requests	1s	The time until the rate limit (based on requests) resets to its initial state.
x-ratelimit-reset-tokens	6m0s	The time until the rate limit (based on tokens) resets to its initial state.

Which I absolutely need in my app to be able to temporarily disable the Agent if we get close to the limits.

Is there a way to get this information ?

Thank you !

[samuelcolvin]

Sorry, are you asking to be able to set request headers or read response headers, or both?

[YanSte]

I think you can have a look with OpenAIModel from Pydantic Ai.

That will solve your issue.

[kantandane]

@samuelcolvin I want to be able to READ response headers.
For each request, OpenaAI sends back in the response headers my account limit status, this is what I want to be able to access when using Pydantic-Ai.

@YanSte Thank you, in OpenAIModel documentation I see we can provide our own existing http_client, so I guess this is the way to do it !

[samuelcolvin]

If you want to observe the headers (and not use them in your application code), you can use the httpx instrumentation in logfire - https://logfire.pydantic.dev/docs/integrations/http-clients/httpx/

If you want to use the values at runtime, pass your own httpx client as @YanSte suggests.

[kantandane]

Thank you to both of you.

I succeeded by following your advice: I created a custom AsyncClient class that I use as the http_client of my OpenAIModel.
I now have all I need to do what I want to.