oai-compatible-copilot

The owned_by field in a model configuration (also known as provider or provide) is used to group provider-specific API keys. The extension stores these keys using the pattern oaicopilot.apiKey.<providerId_lowercase>.

To configure keys for multiple providers:

1. Open VS Code settings and define your models with different owned_by values.
2. Open the Command Palette (Ctrl+Shift+P) and run OAICopilot: Set OAI Compatible Multi-Provider API Key to enter the keys for each provider.

"oaicopilot.models": [
    {
        "id": "Qwen/Qwen3-Coder-480B-A35B-Instruct",
        "owned_by": "modelscope",
        "context_length": 256000,
        "max_tokens": 8192,
        "temperature": 0,
        "top_p": 1
    },
    {
        "id": "qwen3-coder",
        "owned_by": "iflow",
        "baseUrl": "https://apis.iflow.cn/v1",
        "context_length": 256000,
        "max_tokens": 8192,
        "temperature": 0,
        "top_p": 1
    }
]

The extra field allows you to inject arbitrary parameters directly into the API request body. This is useful for accessing vendor-specific features or experimental parameters not covered by the standard configuration fields.

Key behaviors:

• Parameters in extra are merged into the request body.
• They are added after standard parameters. If a conflict occurs, the value in extra takes precedence.
• Supported across all apiMode values (openai, openai-responses, ollama, anthropic, gemini).
• Values can be any valid JSON type (string, number, boolean, object, array).

Common use cases:

• OpenAI specific: seed, logprobs, top_logprobs, suffix, presence_penalty.
• Vendor specific: Custom sampling methods, debugging flags, or Beta features.
• Enabling Thinking/Reasoning: Using vendor-specific nested objects to enable chain-of-thought displays.

// Example: Using extra for OpenAI, Ollama, and Anthropic specific settings
"oaicopilot.models": [
    {
        "id": "custom-model",
        "owned_by": "openai",
        "extra": {
            "seed": 42,
            "logprobs": true,
            "top_logprobs": 5,
            "suffix": "###",
            "presence_penalty": 0.1
        }
    },
    {
        "id": "local-model",
        "owned_by": "ollama",
        "baseUrl": "http://localhost:11434",
        "apiMode": "ollama",
        "extra": {
            "keep_alive": "5m",
            "raw": true
        }
    },
    {
        "id": "claude-model",
        "owned_by": "anthropic",
        "baseUrl": "https://api.anthropic.com",
        "apiMode": "anthropic",
        "extra": {
            "service_tier": "standard_only"
        }
    }
]

You can enable model reasoning/thinking displays in Copilot by using provider-specific parameters within the extra field or specific reasoning fields.

For OpenAI Responses: Set apiMode to openai-responses and use reasoning_effort along with a reasoning.summary object in extra.

For Gemini: Set apiMode to gemini and use generationConfig.thinkingConfig.includeThoughts inside the extra field.

// OpenAI Responses with detailed reasoning
{
  "id": "gpt-4o-mini",
  "owned_by": "openai",
  "baseUrl": "https://api.openai.com/v1",
  "apiMode": "openai-responses",
  "reasoning_effort": "high",
  "extra": {
    "reasoning": {
      "summary": "detailed"
    }
  }
}

// Gemini with thought summaries
{
  "id": "gemini-3-flash-preview",
  "owned_by": "gemini",
  "baseUrl": "https://generativelanguage.googleapis.com",
  "apiMode": "gemini",
  "extra": {
    "generationConfig": {
      "thinkingConfig": {
        "includeThoughts": true
      }
    }
  }
}

To display the Thinking module in Copilot, you must use provider-specific parameters via the extra field or specific reasoning fields, depending on the apiMode and provider.

OpenAI Responses

Set apiMode: "openai-responses" and use the extra field for reasoning summary configuration:

{
  "id": "gpt-4o-mini",
  "owned_by": "openai",
  "baseUrl": "https://api.openai.com/v1",
  "apiMode": "openai-responses",
  "reasoning_effort": "high",
  "extra": {
    "reasoning": {
      "summary": "detailed"
    }
  }
}

Gemini

Set apiMode: "gemini" and use the extra field to enable thinkingConfig within generationConfig:

{
  "id": "gemini-3-flash-preview",
  "owned_by": "gemini",
  "baseUrl": "https://generativelanguage.googleapis.com",
  "apiMode": "gemini",
  "extra": {
    "generationConfig": {
      "thinkingConfig": {
        "includeThoughts": true
      }
    }
  }
}

// Gemini Thinking Example
{
  "id": "gemini-3-flash-preview",
  "owned_by": "gemini",
  "baseUrl": "https://generativelanguage.googleapis.com",
  "apiMode": "gemini",
  "extra": {
    "generationConfig": {
      "thinkingConfig": {
        "includeThoughts": true
      }
    }
  }
}

To use any OpenAI, Ollama, Anthropic, or Gemini compatible API in VS Code's GitHub Copilot Chat, follow these steps:

1. Install: Install the OAI Compatible Provider for Copilot extension from the VS Code Marketplace.
2. Configure Base URL and Models: Open VS Code settings and configure oaicopilot.baseUrl and oaicopilot.models.
3. Access Model Selector: Open the GitHub Copilot Chat interface, click the model selector, and choose "Manage Models...".
4. Select Provider: Choose the "OAI Compatible" provider.
5. Set API Key: Enter your API key (it will be saved locally).
6. Add Models: Select the models you want to appear in the model selector.

"oaicopilot.baseUrl": "https://api-inference.modelscope.cn/v1",
"oaicopilot.models": [
    {
        "id": "Qwen/Qwen3-Coder-480B-A35B-Instruct",
        "owned_by": "modelscope",
        "context_length": 256000,
        "max_tokens": 8192,
        "temperature": 0,
        "top_p": 1
    }
]

The provider supports several distinct API modes, which determine how chat messages are converted and sent to the backend:

• openai (Default): Uses the standard OpenAI /chat/completions endpoint. It includes stream_options: { include_usage: true } in the request.
• ollama: Uses the Ollama native /api/chat endpoint. Messages are converted to the Ollama format.
• anthropic: Uses the Anthropic /messages endpoint. It handles message conversion and optional cache control.
• gemini: Uses the Google Gemini generateContent endpoint. It converts messages into Gemini's contents and systemInstruction format.
• openai-responses: Uses the OpenAI Responses API. This mode supports stateful conversations by using a previous_response_id to avoid sending the full chat history in every request. It uses a special stateful marker (application/vnd.oaicopilot.stateful-marker) to track conversation state.