What token to use for Github

[kidmose]

Briefly describe what you are trying to do

Set up gptel to use Github Copilot

1. Installed gptel with package-install

2. Created token from https://github.com/settings/tokens (Tried both "classic" and "fine-grained")

3. Set up gptel in emacs, init.el:

    (use-package gptel
      :ensure t
      :custom
      (gptel-model 'gpt-4o)
      (gptel-log-level 'debug)
      (gptel-backend
       (gptel-make-openai "Github Models"
                          :host "models.inference.ai.azure.com"
                          :endpoint "/chat/completions"
                          :stream t
                          :key "<REDACTED>"
                          :models '(gpt-4o))))
   

4. Started gptel buffer: M-x gptel

5. Selected default (I.e. "Github Models" from above)

6. Entered prompt and press C-c RET

Expected: Some response from the LLM, entered into the buffer.

Received:

• Nothing in gptel buffer (*Github Models*)
• Error in *gptel-log* (full content below): 401 Your account type is not currently supported

Is there anyone able to provide some guidance on what I am missing here? (Apparently some permission/role from my GH org. but which?)
Would be greatly appreciated :)

Additional context
Emacs version: GNU Emacs 28.2
Operating system: Debian on 12 on WSL

** gptel.log **

{
  "gptel": "request headers",
  "timestamp": "2024-12-10 15:56:10"
}
{
  "Content-Type": "application/json",
  "Authorization": "Bearer <REDACTED>"
}
{
  "gptel": "request body",
  "timestamp": "2024-12-10 15:56:10"
}
{
  "model": "gpt-4o",
  "messages": [
    {
      "role": "system",
      "content": "You are a large language model living in Emacs and a helpful assistant. Respond concisely."
    },
    {
      "role": "user",
      "content": "hi"
    }
  ],
  "stream": true,
  "temperature": 1.0
}
{"gptel": "request Curl command", "timestamp": "2024-12-10 15:56:10"}
curl \
--disable \
--location \
--silent \
--compressed \
-XPOST \
-y300 \
-Y1 \
-D- \
-w\(<REDACTED>\ .\ \%\{size_header\}\) \
-d\{\"model\"\:\"gpt-4o\"\,\"messages\"\:\[\{\"role\"\:\"system\"\,\"content\"\:\"You\ are\ a\ large\ language\ model\ living\ in\ Emacs\ and\ a\ helpful\ assistant.\ Respond\ concisely.\"\}\,\{\"role\"\:\"user\"\,\"content\"\:\"hi\"\}\]\,\"stream\"\:true\,\"temperature\"\:1.0\} \
-HContent-Type\:\ application/json \
-HAuthorization\:\ Bearer\ <REDACTED> \
https\://models.inference.ai.azure.com/chat/completions
{
  "gptel": "response headers",
  "timestamp": "2024-12-10 15:56:11"
}
"HTTP/2 401 \r\ndate: Tue, 10 Dec 2024 14:56:11 GMT\r\ncontent-type: application/json; charset=utf-8\r\nx-ms-error-code: unauthorized\r\nrequest-context: appId=\r\nstrict-transport-security: max-age=31536000; includeSubDomains; preload\r\nx-content-type-options: nosniff\r\nx-aml-cluster: vienna-eastus2-02\r\nx-request-time: 0.090\r\n\r"
{
  "gptel": "response body",
  "timestamp": "2024-12-10 15:56:11"
}
{
  "error": {
    "code": "unauthorized",
    "message": "Your account type is not currently supported",
    "details": "Your account type is not currently supported"
  }
}

[kidmose]

I believe these are the relevant settings: https://github.com/settings/copilot (screenshot below).
My guess is that it is "Copilot in github.com" that the above config refers to (models.inference.ai.azure.com/chat/completions) so that needs to be enabled,
... or I need to figure out how to use e.g. the "Copilot Chat in the IDE" API instead.

[kidmose]

Getting "Copilot on github.com" enabled, I was able to progress to a different error: 401: Bad credentials
This seems to strongly suggest that it is the "Copilot on github.com" feature that needs to be enabled for using https://models.inference.ai.azure.com/chat/completions .

I tried to create a classic token, with "copilot" permission; gives 401 Bad credentials
and i tried a fine-grained token, with "Read access to Copilot Chat" and "Read access to Copilot Editor Context"´; also 401 Bad credentials

Any inputs on what permission are necessary (or any hints, what so ever) is much appreciated :)

[gs-101]

Hello, I'm the one who implemented the GitHub support for gptel (see #386).

Sorry if the current description is misleading, but it only supports GitHub Models (here is a blog post about it), which basically exposes APIs for multiple LLMs, so GitHub users can interact with them. I use it with that and it works prefectly.

For Copilot, there's copilot.el + copilot-chat.el

For use with gptel, maybe someone with access to it could extend the GitHub support (but for it to be good I think completion support would have to be implemented).

EDIT 1: As for token permissions, the token works without any of them ticked
EDIT 2: Nevermind the lack of access to Copilot, it's free now!