GitHub Copilot

GitHub Copilot chat and the Copilot CLI can call the Gonka API. Inline completions stay on GitHub.

Base URL for the CLI stops at /v1: https://api.proxy.gonka.gg/v1

VS Code chat

1. Command Palette → Chat: Manage Language Models
2. Add Models → Custom Endpoint
3. API type: Chat Completions
4. Save the snippet VS Code opens in chatLanguageModels.json. Keep the key in an input variable.

[
  {
    "name": "Gonka Proxy",
    "vendor": "customendpoint",
    "apiKey": "${input:gonkaProxyKey}",
    "apiType": "chat-completions",
    "models": [
      {
        "id": "MiniMaxAI/MiniMax-M2.7",
        "name": "MiniMax M2.7",
        "url": "https://api.proxy.gonka.gg/v1/chat/completions",
        "toolCalling": true,
        "maxInputTokens": 160000,
        "maxOutputTokens": 8192
      }
    ]
  }
]

This url is the full chat path. Every other client stops at /v1. If the model does not show up in the picker, restart VS Code. Set the token limits from the context window on Available models. DeepSeek tool calls need reasoning compatibility.

Copilot CLI

The model must stream and call tools.

export COPILOT_PROVIDER_TYPE=openai
export COPILOT_PROVIDER_BASE_URL=https://api.proxy.gonka.gg/v1
export COPILOT_PROVIDER_API_KEY=sk-...
export COPILOT_MODEL=MiniMaxAI/MiniMax-M2.7
copilot

Shared notes on timeouts, MCP, and fallback are on Agents and IDEs.