OAI2LMApi - OpenAI to Language Model API Bridge

A VSCode extension that connects OpenAI-compatible APIs to VSCode's Language Model API, enabling integration with GitHub Copilot Chat and other AI-powered features in VSCode.

Features

• ✅ Full OpenAI API Compatibility: Works with any OpenAI-compatible API endpoint
• ✅ Google Gemini Support: Native support for Google Gemini API with dedicated channel
• ✅ VSCode Language Model API Integration: Seamlessly integrates with VSCode's built-in language model features
• ✅ Streaming Support: Real-time streaming responses for better user experience
• ✅ Thinking/Reasoning Support: Stream reasoning content from models that support it (e.g., o1, Claude with thinking)
• ✅ Automatic Model Loading: Fetches available models from the API endpoint on startup
• ✅ Model Caching: Loaded models are cached locally for faster startup times
• ✅ Secure API Key Storage: API keys are stored securely using VSCode's SecretStorage
• ✅ Tool Calling Support: Full support for tool/function calling with improved reliability
• ✅ XML-based Tool Calling: Fallback prompt-based tool calling for models without native support
• ✅ Tool Calling Filter: Optionally filter models by tool/function calling support
• ✅ Per-Model Configuration: Customize token limits, temperature, and capabilities per model
• ✅ Easy Configuration: Simple setup through VSCode settings and commands

Requirements

• VSCode version 1.107.0 or higher
• An OpenAI-compatible API endpoint
• API key for authentication

Installation

From VS Code Marketplace (Recommended)

1. Open VSCode
2. Go to Extensions view (Ctrl+Shift+X / Cmd+Shift+X)
3. Search for "OAI2LMApi"
4. Click Install

Or install directly via VS Code Marketplace.

From VSIX

1. Download the .vsix file from the Releases page
2. In VSCode, open the Command Palette (Ctrl+Shift+P / Cmd+Shift+P)
3. Run "Extensions: Install from VSIX..." and select the downloaded file

From Source

1. Clone this repository
2. Run pnpm install to install dependencies
3. Run pnpm run compile to build the extension
4. Press F5 in VSCode to run the extension in debug mode

Quick Start

1. Install the extension
2. Open Command Palette (Ctrl+Shift+P / Cmd+Shift+P)
3. Run OAI2LMApi: Set API Key and enter your API key
4. (Optional) Configure the API endpoint in settings if not using OpenAI
5. Models will be automatically loaded and available in GitHub Copilot Chat

Commands

Configuration

Configure the extension through VSCode settings (Ctrl+, or Cmd+,):

Basic Settings

Gemini Settings

Claude Settings

Model Overrides

The oai2lmapi.modelOverrides setting allows per-model configuration. Keys are model name patterns (supports wildcards like gemini-*), values are configuration objects.

The oai2lmapi.channelModelOverrides setting allows per-channel configuration. Keys are channel names (e.g. openai, gemini, claude) and values are objects mapping model patterns to override objects. Matching overrides are merged in order: global modelOverrides then channel overrides.

Example Configuration

{
  "oai2lmapi.apiEndpoint": "https://api.openai.com/v1",
  "oai2lmapi.autoLoadModels": true,
  "oai2lmapi.showModelsWithoutToolCalling": false,
  "oai2lmapi.enableGeminiChannel": true,
  "oai2lmapi.geminiApiEndpoint": "https://generativelanguage.googleapis.com",
  "oai2lmapi.enableClaudeChannel": true,
  "oai2lmapi.claudeApiEndpoint": "https://api.anthropic.com/v1",
  "oai2lmapi.modelOverrides": {
    "gemini-2.0-flash-thinking-exp": {
      "thinkingLevel": "auto",
      "usePromptBasedToolCalling": true
    },
    "claude-*": {
      "maxOutputTokens": 8192
    }
  },
  "oai2lmapi.channelModelOverrides": {
    "openai": {
      "gpt-4o": {
        "temperature": 0.7
      }
    },
    "claude": {
      "claude-3.7*": {
        "thinkingLevel": "high"
      }
    }
  }
}

API Key Storage

The API key is stored securely using VSCode's built-in SecretStorage. Use the OAI2LMApi: Set API Key command to set your key.

Note: If you previously stored an API key in settings (the deprecated oai2lmapi.apiKey setting), it will be automatically migrated to secure storage.

Supported APIs

This extension works with any API that implements the OpenAI chat completions format:

• OpenAI API
• Azure OpenAI
• Google Gemini API (native support via Gemini channel)
• Anthropic Claude (native Claude channel via @anthropic-ai/sdk)
• LocalAI
• Ollama (with OpenAI compatibility layer)
• LM Studio
• vLLM
• liteLM (transfer from other APIs)
• OpenRouter
• Any custom OpenAI-compatible implementation

Troubleshooting

Extension doesn't activate

• Ensure you're using VSCode 1.107.0 or higher
• Verify the extension is enabled in the Extensions view

Models not loading

• Run OAI2LMApi: Set API Key to ensure your API key is configured
• For Gemini: Enable oai2lmapi.enableGeminiChannel and run OAI2LMApi: Set Gemini API Key
• For Claude: Enable oai2lmapi.enableClaudeChannel and run OAI2LMApi: Set Claude API Key
• Verify your API endpoint is correct and accessible
• Check the OAI2LMApi Output Channel (View > Output, select "OAI2LMApi") for detailed logs
• Try OAI2LMApi: Refresh Models to manually reload

API errors

• Ensure your API endpoint supports /v1/models and /v1/chat/completions
• For Gemini: The extension uses the native Gemini API format, not OpenAI compatibility
• Verify network connectivity to the API endpoint

Tool calling issues

• Some models don't support native function calling
• Use usePromptBasedToolCalling: true in model overrides for XML-based fallback
• Check model capabilities in the Output Channel logs

Development Notes

• Model metadata patterns are maintained in @oai2lmapi/model-metadata (packages/model-metadata/src/index.ts) and shared across all packages.

License

MIT - See LICENSE for details.

Contributing

Contributions are welcome! Please feel free to submit issues or pull requests on GitHub.