Custom Provider (Bring Your Own Models)

A Raycast extension that allows you to manage custom AI provider settings through a convenient interface. The extension works with the providers.yaml file used by Raycast AI to configure external providers (see documentation here).

Features

• View a list of all configured providers
• View detailed information about each provider (models, API keys, additional parameters)
• Edit existing or add new providers and models
• Synchronize models from any OpenAI-compatible provider's /models endpoint

Use the Sync Models action on a provider to replace its configured model list with the provider's live OpenAI-compatible catalog. The request uses the provider's configured API key when present. Existing models remain unchanged if fetching or validating the catalog fails.

The extension automatically works with the Raycast AI configuration file located at:

~/.config/raycast/ai/providers.yaml

For security purposes, a backup copy of the providers.yaml file is created before making changes.

providers.yaml Structure

The providers.yaml file has the following structure:

providers:
  - id: provider_id              # Unique provider identifier (required)
    name: Provider Name          # Provider name (required)
    base_url: https://api.example.com  # Base API URL (required)

    # API keys (optional)
    # If authentication is required, specify at least one key
    # If individual models require different keys, specify a separate `key` for each model
    api_keys:
      provider_key: ENV_VAR_NAME

    # Additional parameters for `/chat/completions` endpoint (optional)
    additional_parameters:
      param1: value1
      param2: value2

    # List of provider models (required, minimum 1 model)
    models:
      - id: model_id             # Model identifier used by the provider (required)
        name: Model Name         # Model name in Raycast (required)
        provider: provider_key    # Mapping to a specific API key (optional)
        description: Model description  # Model description (optional)
        context: 128000          # Context window size (required)

        # Model abilities (optional)
        # All properties within abilities are also optional
        abilities:
          temperature:
            supported: true
          vision:
            supported: true
          system_message:
            supported: true
          tools:
            supported: false
          reasoning_effort:
            supported: false

Configuration Examples

Simple Provider with One Model

providers:
  - id: my_provider
    name: My Provider
    base_url: https://api.example.com
    api_keys:
      default: MY_API_KEY
    models:
      - id: gpt-4
        name: GPT-4
        context: 128000
        abilities:
          temperature:
            supported: true
          vision:
            supported: true

Provider with Multiple Models and Different API Keys

providers:
  - id: multi_provider
    name: Multi Provider
    base_url: https://api.example.com
    api_keys:
      openai: OPENAI_KEY
      anthropic: ANTHROPIC_KEY
    models:
      - id: gpt-4o
        name: GPT-4o
        provider: openai
        context: 200000
      - id: claude-sonnet
        name: Claude Sonnet
        provider: anthropic
        context: 200000

Provider Without API Keys

providers:
  - id: local_provider
    name: Local Provider
    base_url: http://localhost:4000
    models:
      - id: local-model
        name: Local Model
        context: 128000

Important Notes

1. API Compatibility: Since the OpenAI API is not a standard, not all providers may work correctly with Raycast AI. It's recommended to check your provider's documentation.

2. Model Abilities: If abilities are specified incorrectly, the model may not work correctly in Raycast AI. Always refer to the provider's documentation.

3. YAML Format: The extension DOES NOT preserve comments and formatting where possible, but when manually editing the file, ensure the YAML syntax is correct.

License

MIT