Generic Provider for Copilot

Use frontier open LLMs like Qwen3 Coder, Kimi K2, DeepSeek V3.1, GLM 4.5 and more in VS Code with GitHub Copilot Chat powered by any OpenAI-compatible provider 🔥

Thanks

Heavily inspired (and then extended) by https://github.com/JohnnyZ93/oai-compatible-copilot

✨ Features

• Configuration GUI: Intuitive webview-based interface for managing providers and models with validation and error handling. Access this with the quick picker entry "GenericCopilot: Open Configuration GUI"
• Provider-First Configuration: Define providers once with shared settings (baseUrl, headers, API keys) that are automatically inherited by models
• Multiple Provider Support: Manage API keys for unlimited providers with automatic per-provider key storage using vscode secret storage.
• Flexible Headers & parameters: Set custom parameters for any model.

Requirements

• VS Code: 1.105.0 or higher

• Dependency: GitHub Copilot Chat extension

• API Keys: OpenAI-compatible provider API keys

• Supported Vercel AI SDK Providers: This extension currently supports the following provider types: openai, openai-compatible, openrouter, and google.

⚡ Quick Start

Option A: Using the Configuration GUI (Recommended)

1. Use the GUI

1. Open Configuration GUI:

   • Press Ctrl+Shift+P (Windows/Linux) or Cmd+Shift+P (macOS)
   • Type "GenericCopilot: Open Configuration GUI"
   • Press Enter

2. Add Providers:

   • Click "+ Add Provider"
   • Enter provider id (e.g., "iflow") and base URL
   • Optionally configure default parameters

3. Add Models:

   • Click "+ Add Model"
   • Enter model ID and select a provider
   • Configure model-specific settings as needed

4. Save: Click "Save Configuration" button

2. Set API Keys

If an API key is not found for a provider, you will be prompted in the QuickPick box.

1. Open Command Palette (Ctrl+Shift+P or Cmd+Shift+P)
2. Run: "GenericCopilot: Set Generic Compatible Multi-Provider Apikey"
3. Select your provider (e.g., iflow)
4. Enter the API key for that provider

Repeat for each provider. Keys are stored securely in VS Code's secret storage as generic-copilot.apiKey.<provider-key>.

3. Use in Copilot Chat

1. Open GitHub Copilot Chat
2. Click the model picker
3. Select "Manage Models..."
4. Choose "Generic Compatible" provider
5. Select the models you want to enable
6. Start chatting!

📖 Configuration Guide

Configuration is managed in VS Code's settings.json file. You can either edit the JSON directly or use the Configuration GUI (GenericCopilot: Open Configuration GUI).

The configuration is split into two main parts: providers and models.

Provider Configuration (generic-copilot.providers)

Providers define the connection details for an API endpoint. Models reference a provider to inherit its settings.

Schema:

Model Configuration (generic-copilot.models)

Models define the specific LLMs you want to use. Each model must be associated with a provider.

Schema:

model_properties Schema

model_parameters Schema

⚙️ Configuration Example

Here is a complete example for your settings.json file, demonstrating how to configure multiple providers and models.

{
  "generic-copilot.providers": [
    {
      "id": "openrouter-connection",
      "vercelType": "openrouter",
      "displayName": "OpenRouter",
    },
    {
      "id": "zai",
      "vercelType": "openai-compatible",
      "displayName": "Zai",
      "baseUrl": "https://open.zaidata.com/v1",
      "headers": {
        "X-Source": "vscode-extension"
      }
    }
  ],
  "generic-copilot.models": [
    {
      "_comment": "A simple model configuration inheriting from OpenRouter.",
      "slug": "claude-sonnet-default",
      "id": "anthropic/claude-3.5-sonnet",
      "provider": "openrouter",
      "model_properties": {
        "context_length": 200000,
        "family": "claude"
      },
      "model_parameters": {
        "temperature": 0.7
      }
    },
    {
      "slug": "glm-4.6-fast",
      "id": "glm-4.6",
      "provider": "zai",
      "displayName": "GLM-4.6 (Fast)",
      "model_properties": {
        "context_length": 256000
      },
      "model_parameters": {
        "temperature": 0.1
      }
    },
    {
      "_comment": "A model with custom parameters passed via the 'extra' field.",
      "slug": "gemini-flash-custom",
      "id": "google/gemini-flash-1.5",
      "provider": "openrouter",
      "model_parameters": {
        "temperature": 0.5,
        "extra": {
          "top_p": 0.9,
          "stop": ["\n"]
        }
      }
    }
  ]
}

�🔑 API Key Management

Per-Provider Keys

Each provider has its own API key stored securely:

• Storage Key: generic-copilot.apiKey.<provider-key>
• Example: For provider key: "iflow", the storage key is generic-copilot.apiKey.iflow

🎛️ Advanced Configuration

Custom Headers

Headers can be set at the provider level and will be inherited by all models associated with that provider. See the Provider Configuration section for details.

API Request Format

When making requests to the model provider:

1. Model ID Mapping: The id from model_properties is sent as the model parameter in the API request
2. Parameters Only: Only model_parameters (temperature, max_tokens, etc.) are included in the request body
3. Excluded Metadata: model_properties like baseUrl, context_length, and family are NOT sent to the API - they're used internally by the extension.
4. Unknown Keys: Custom parameters can be added via model_parameters.extra and will be passed through to the API

💡 Tips & Best Practices

Use family and model names carefully. Copilot changes behavior based on these names:

Model Name variations

• gpt-5-codex | gpt-5-codex : uses Codex-style prompt branch
• gpt-5* | gpt-5 : can use apply_patch exclusively; agent prompts differ for gpt-5
• o4-mini | o4-mini : allowed apply_patch and prefers JSON notebook representation
• claude-3.5-sonnet | claude-3.5-sonnet : prefers instructions in user message and after history

Family Name variations

• GPT family | gpt (excl. gpt-4o) : supports apply_patch, prefers JSON notebook representation
• Claude / Anthropic | claude / Anthropic : supports multi_replace/replace_string, can use replace_string exclusively, MCP image_url disallowed
• Gemini | gemini : supports replace_string, healing/strong-replace hints required, cannot accept image_url in requests. Supports Gemini 3 thought signatures (requires google provider).
• Grok | grok-code : supports replace_string and can use replace_string exclusively

Naming Convention

Use lowercase provider keys that match the service name for consistency:

• ✅ "key": "openai"
• ✅ "key": "anthropic"
• ❌ "key": "OpenAI"

ConfigId for Variants

Use descriptive configId values:

• "thinking" / "no-thinking"
• "fast" / "accurate"

Headers for Custom Auth

If a provider uses non-standard authentication, set it in the headers object of the provider's configuration.

Gemini 3 & Thought Signatures

Gemini 3 models (e.g. Gemini 3 Pro) introduce a requirement for preserving "thought signatures" during multi-turn conversations involving function calls.

This extension implements automated handling of these signatures only when using the google provider type.

To properly support Gemini 3:

1. Use vercelType: "google" for your provider configuration.
2. Ensure your model family is set to "gemini".

🐛 Troubleshooting

Models Not Appearing

1. Check provider key matches exactly in both provider and model config
2. Verify baseUrl is correct and accessible
3. Look for errors in VS Code Developer Console (Help > Toggle Developer Tools)

Authentication Errors

1. Verify API key is set: Run "Set Multi-Provider Apikey" command
2. Check if provider requires custom headers in its provider configuration.
3. Ensure baseUrl includes correct path (usually /v1)

Provider Not Found

1. Confirm provider field in your model configuration matches a provider's id exactly (case-sensitive).
2. Check Developer Console for warnings about missing providers.
3. Verify JSON syntax is valid (no trailing commas, quotes closed).
4. Remember: Only baseUrl and headers are inherited from providers.

📄 License

• License: MIT License Copyright (c) 2025