Configuring custom language models.

[ushankax]

Hello! I'm trying to configure custom language model. In VSCode's Continue extension I do it like this:

  "models": [
    {
      "title": "<some title>",
      "provider": "<some provider>",
      "model": "<some model>",
      "apiKey": "<apikey>",
      "apiBase": "http://custom-api",
      "requestOptions": {
        "headers": {
          "X-ASGK-TOKEN": "<uuid>"
        }
      }
    }
  ],

apiBase is like api_url in your config, but what about requestOptions? It's required to my model. https://docs.continue.dev/customize/json-reference#models - more about Continue's properties.

requestOptions: Model-specific HTTP request options, same format as top-level requestOptions, which they override.

How to configure like this in Zed?

[turboladen]

I'm trying to figure this out too. FWIW, here's the backing struct for the JSON Object there:

zed/crates/settings_content/src/language_model.rs

Line 255 in 61d9696

pub struct OpenAiCompatibleAvailableModel {

. Full populated, an example would look something like:

    "openai_compatible": {
      "my_provider": {
        "api_url": "https://my_provider.com",
        "available_models": [
          {
            "name": "cool-model",
            "display_name": "Cool Model",
            "max_tokens": 128000,
            "max_output_tokens": 64000,
            "max_completion_tokens": 64000,
            "capabilities": {
              "tools": true,
              "images": true,
              "parallel_tool_calls": true,
              "prompt_cache_key": true,
              "chat_completions": true,
            },
          },
        ],
      },
    },

Doesn't look like there's room for request options--not at that level of the config at least.