Integrate LiteLLM proxies into GitHub Copilot Chat through the litellm-connector vendor that surfaces hundreds of LiteLLM-backed models directly inside VS Code.
Open the Copilot Chat interface and select "Manage Models..." from the model picker.
Choose the "LiteLLM" provider (vendor litellm-connector) and run the Manage LiteLLM Provider command (litellm-connector.manage) from the Command Palette.
Enter your LiteLLM base URL (e.g., http://localhost:4000 or your hosted proxy) and optionally your API key. Values are stored securely in VS Code secret storage.
Pick the LiteLLM models you want to enable. Each entry retains the cheapest and fastest selection options to favor cost or throughput.
Each model entry also offers cheapest and fastest mode for each model. fastest selects the provider with highest throughput and cheapest selects the provider with lowest price per output token.
✨ Features
Exposes hundreds of LiteLLM-backed models inside Copilot Chat through the litellm-connector vendor registration and litellm-connector.manage configuration command.
Handles streaming responses, built-in tool calling, and image-capable vision models when the LiteLLM endpoint advertises those capabilities.
Detects LiteLLM-openAI parameter limitations, strips incompatible arguments to avoid errors, and shows friendly warnings to reconfigure when authorization or connectivity issues occur.
Works with self-hosted LiteLLM proxies or hosted instances, so you can keep data inside your network or use managed APIs.
Requirements
VS Code 1.107.0 or higher (match the engines.vscode constraint).
GitHub Copilot Chat (github.copilot-chat) must be installed since we register a chat provider within that experience.
A LiteLLM endpoint (self-hosted or cloud) plus its base URL and optional API key.
🛠️ Development
git clone https://github.com/gethnet/litellm-vscode-chat
cd litellm-vscode-chat
pnpm install # downloads @vscode/dts (download-api) via postinstall
pnpm run compile
Press F5 to launch an Extension Development Host.
Common scripts:
Build: pnpm run compile
Download API metadata: pnpm run download-api (also triggered automatically after install and updates)
Watch: pnpm run watch
Lint and fix: pnpm run lint
Format: pnpm run format
Test: pnpm run test (runs compile + vscode-test)
Bump version: pnpm run bump-version (updates package metadata via scripts/bump-version.js)
