Add the ability to customize available models for OpenAI-compatible services

[amtoaer]

Closes #11984, closes #11075.

Release Notes:

• Added the ability to customize available models for OpenAI-compatible services (#11984) (#11075).

[christ-offer]

Very nice! :D

[rtfeldman]

Thanks for the contribution - this is looking good overall! 😃

I have one requested design change, and then some questions about code structure.

[amtoaer]

Thank you. I've updated it. Please let me know if there are any other issues.

[rtfeldman]

Nice, thank you @amtoaer! 🎉

[sa1]

Some central documentation for new features would be nice.

[pramjana]

The GIF in this thread changes the displayed image too fast, making it hard to follow. Also, the longer image above is missing some information. Zed Docs mentions "language models" as the key, and so does gliptic in 16576.

It's confusing!

Please can someone provide a simple, correct method to set up OpenRouter models in Zed (one model example is enough, thank you!).

[notpeter]

Please can someone provide a simple, correct method to set up OpenRouter models in Zed (one model example is enough, thank you!).

I believe they are using OpenAI API compatibility, so based on the Asssistant Configuration: OpenAI docs:

  "language_models": {
    "openai": {
      "api_url": "https://api.deepseek.com/v1",
      "available_models": [
        {
          "provider": "openai",
          "name": "deepseek_chat",
          "max_tokens": 32000
        },
        {
          "provider": "openai",
          "name": "deepseek_code",
          "max_tokens": 32000
        }
      ]
    }
  }

Untested, but you will need to enter your API key under the OpenAI section (press enter to confirm) of the Assistant configuration ("assistant: show configuration" from the command palette).

[pramjana]

Thank you for sharing @notpeter! I got the following working in Zed for the openrouter based api.

{
  "assistant": {
    "enabled": true,
    "default_model": {
      "provider": "openai",
      "model": "qwen/qwen-2.5-72b-instruct"
    },
    "version": "2"
  },
  "language_models": {
    "openai": {
      "version": "1",
      "api_url": "https://openrouter.ai/api/v1",
      "available_models": [
        {
          "name": "qwen/qwen-2.5-72b-instruct",
          "display_name": null,
          "max_tokens": 32000,
          "max_output_tokens": 8192,
          "max_completion_tokens": null
        }
      ]
    }
  },
}