Copilot-Proxy

A proxy server that bridges GitHub Copilot Chat with GLM coding models by mimicking the Ollama API interface.

What it does

This proxy server intercepts requests from GitHub Copilot's Ollama provider and forwards them to a GLM coding plan backend. By implementing the Ollama API interface, it allows the GitHub Copilot VS Code extension to communicate with alternative language models seamlessly.

flowchart TD
   A[GitHub Copilot Chat] -- Ollama API (localhost:11434) --> B[Copilot-Proxy Server]
   B --> C[GLM Coding Plan Backend]

Loading

Quick Start

Prerequisites

1. Python 3.10+
2. UV for dependency management and packaging (install instructions)
3. Z.AI Coding Plan access with a valid API key

Install from PyPI (recommended)

# Ensure uv is installed first
uv pip install copilot-proxy

# Or run without installing globally
uvx copilot-proxy --help

Run the proxy locally

# Quick one-liner using uvx
uvx copilot-proxy --host 127.0.0.1 --port 11434

# Or inside a synced project environment
uv sync
uv run copilot-proxy

The server listens on http://localhost:11434 by default (same port Ollama uses). Make sure Ollama itself is stopped to avoid port conflicts.

Configure credentials

Provide your Z.AI API key before launching the proxy:

# PowerShell (current session only)
$env:ZAI_API_KEY = "your-zai-api-key"

# bash/zsh
export ZAI_API_KEY="your-zai-api-key"

You can optionally set a custom endpoint with ZAI_API_BASE_URL, though the default already targets the Coding Plan URL https://api.z.ai/api/coding/paas/v4.

Configure GitHub Copilot in VS Code

• Open the GitHub Copilot Chat panel in VS Code
• Click on the current model name to view available models
• Click 'Manage Models...'
• Select 'Ollama' from the list of providers
• Choose your preferred model from the available GLM models

Available Models

The proxy advertises the GLM Coding Plan lineup so Copilot (or any Ollama-compatible client) can switch between them seamlessly:

Model	Description	Use Case Highlights
GLM-4.6	Flagship coding model with top-tier reasoning	Complex refactors, multi-file tasks, tool use
GLM-4.5	Balanced performance for everyday coding	General coding, debugging, architecture input
GLM-4.5-Air	Lightweight, faster response variant	Quick iterations, drafting, lower-latency use

Tip: These identifiers match the GLM Coding Plan catalog, so any OpenAI-compatible tool can use them by pointing to https://api.z.ai/api/coding/paas/v4 with your Coding Plan API key.

How it Works

The proxy server implements the Ollama API specification, allowing GitHub Copilot's Ollama provider to communicate with it. When Copilot sends requests to localhost:11434, the proxy intercepts these requests and forwards them to the GLM coding plan backend, then returns the responses in Ollama-compatible format.

Troubleshooting

Common Issues:

1. Port conflict errors

   • Ensure Ollama is not running (both services use port 11434)
   • Check that no other service is using port 11434
   • On Windows, use: netstat -ano | findstr :11434
   • On Unix/Linux/Mac, use: lsof -i :11434

2. Ollama provider not responding in Copilot Chat

   • Verify the proxy server is running
   • Check the terminal for any error messages
   • Ensure the GLM backend is accessible

3. Models not appearing in VS Code

   • Restart VS Code after starting the proxy server
   • Make sure you've selected 'Ollama' as the provider in Copilot settings
   • Check that the proxy server is responding at http://localhost:11434

Developing locally

uv sync
uv run uvicorn copilot_proxy.app:app --reload --port 11434

Use uv run pytest (once tests are added) or uvx ruff check . for linting.

Releasing to PyPI with UV

1. Bump the version in pyproject.toml.

2. Build the distributions:

   uv build

3. Check the metadata:

   uvx twine check dist/*

4. Publish to TestPyPI (recommended before production):

   uv publish --repository testpypi

5. Publish to PyPI:

   uv publish

Both uv publish commands expect the relevant API token to be available in the UV_PUBLISH_TOKEN environment variable.

GitHub Actions trusted publisher

This repository includes .github/workflows/publish.yml, which builds and uploads releases automatically on GitHub tag releases. To enable it:

1. Create a PyPI trusted publisher (pending or project-specific) pointing at:
   • Project: copilot-proxy
   • Owner: modpotato
   • Repository: copilot-proxy
   • Workflow: publish.yml
   • Environment: release
2. In GitHub, create the matching repository environment (Settings → Environments → New environment → release).
3. Push a tag (e.g. v0.1.0) to GitHub (git push origin v0.1.0). The workflow will build with uv, publish to PyPI via OIDC, and create the GitHub release automatically.
4. For dry runs, use the Run workflow button; the manual dispatch builds and validates without publishing or creating a release.

License

This project is licensed under the MIT License - see the LICENSE file for details.