Custom headers/cookies in InferenceClient

[Wauplin]

Related to huggingface/api-inference-community#287 (and especially huggingface/api-inference-community#287 (comment)) cc @StephenHodgson @Narsil

With this PR, user can set custom headers and cookies to the InferenceClient session. This can prove useful when using custom endpoints. If set, custom headers overwrite the default user-agent and authorization headers.

(+fix documentation => save audio as .flac instead of .wav)

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[codecov]

Codecov Report

Patch coverage: 100.00% and project coverage change: +0.01 🎉

Comparison is base (b4e01b0) 82.18% compared to head (b3db890) 82.19%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1507      +/-   ##
==========================================
+ Coverage   82.18%   82.19%   +0.01%     
==========================================
  Files          57       57              
  Lines        6089     6093       +4     
==========================================
+ Hits         5004     5008       +4     
  Misses       1085     1085              

Impacted Files	Coverage Δ
src/huggingface_hub/_inference.py	77.04% <100.00%> (+0.51%)	⬆️

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[StephenHodgson]

LGTM

[LysandreJik]

Looks good to me, would like to know @Narsil's thoughts on it given that it's working with the inference API

[Narsil]

Yup LGTM.

I am guessing this comment relies on the undelying lib, but cookies is a header, I'm kind of curious of what happens if you set both to different values.

[Wauplin]

@Narsil to satisfy your curiosity:

import curl
import requests

curl.parse(requests.get("http://www.google.com", headers={"Cookie": "foo=bar"}))
curl.parse(requests.get("http://www.google.com", cookies={"foo": "something", "abc": "def"}))
curl.parse(requests.get("http://www.google.com", headers={"Cookie": "foo=bar"}, cookies={"foo": "something", "abc": "def"}))

which outputs

curl -X GET -H 'Accept: */*' -H 'Accept-Encoding: gzip, deflate' -H 'Connection: keep-alive' -H 'Cookie: foo=bar' -H 'User-Agent: python-requests/2.28.1' http://www.google.com:80/
curl -X GET -H 'Accept: */*' -H 'Accept-Encoding: gzip, deflate' -H 'Connection: keep-alive' -H 'Cookie: abc=def; foo=something' -H 'User-Agent: python-requests/2.28.1' http://www.google.com:80/
curl -X GET -H 'Accept: */*' -H 'Accept-Encoding: gzip, deflate' -H 'Connection: keep-alive' -H 'Cookie: foo=bar' -H 'User-Agent: python-requests/2.28.1' http://www.google.com:80/

Conclusion:

1. Setting in headers has priority.
2. Setting in both headers and cookies just ignores the cookies value.

And thanks all 3 for the reviews! :)

[Narsil]

I love having my curiosity satisfied. Must be has caused a lot of headaches :)