# copilet2api

Cloudflare Worker that proxies **GitHub Copilot** to **OpenAI** and **Anthropic** compatible HTTP APIs.

> **Warning:** This uses unofficial Copilot endpoints. GitHub may change or block them at any time. Excessive automated use may trigger abuse detection and limit your Copilot access. Use responsibly.

## Features

- OpenAI-compatible: `GET /v1/models`, `POST /v1/chat/completions`, `POST /v1/responses` (streaming supported)
- Models like `gpt-5.4-mini`, `gpt-5.3-codex` auto-route to Copilot **Responses API** when using `/v1/chat/completions`
- Anthropic-compatible: `POST /v1/messages` (streaming supported)
- Client auth via API key (default `sk-shumengya666`, configurable via `API_KEY`)
- GitHub token stored as Worker Secret (`GITHUB_TOKEN`)

## Quick start

### 1. Install

```bash
npm install
```

### 2. Configure GitHub token (local)

```bash
cp .dev.vars.example .dev.vars
# Edit .dev.vars and set GITHUB_TOKEN (ghu_, gho_, or github_pat_ with Copilot access)
```

### 3. Run locally

```bash
npm run dev
```

### 4. Deploy

```bash
npx wrangler secret put GITHUB_TOKEN
npm run deploy
```

Optionally override the client API key in production:

```bash
npx wrangler secret put API_KEY
```

## Authentication

Clients must send one of:

- `Authorization: Bearer <API_KEY>` (default: `sk-shumengya666`, not shown on `GET /`)
- `x-api-key: <API_KEY>`

Override with `wrangler secret put API_KEY` or `.dev.vars` for production.

The Worker uses `GITHUB_TOKEN` (Wrangler secret) to authenticate with Copilot:

| Token type | Prefix | How it works |
|------------|--------|--------------|
| OAuth (recommended) | `ghu_`, `gho_` | Exchanged via `GET /copilot_internal/v2/token` |
| Fine-grained PAT | `github_pat_` | Used directly; requires **Copilot Requests** permission; uses `Copilot-Integration-Id: copilot-developer-cli` |

Classic PAT (`ghp_`) is **not supported** by the Copilot API.

## Endpoints

| Method | Path | Description |
|--------|------|-------------|
| GET | `/` | Service info |
| GET | `/health` | Health check |
| GET | `/v1/models` | List Copilot models |
| POST | `/v1/chat/completions` | OpenAI chat completions |
| POST | `/v1/messages` | Anthropic messages |

## Usage examples

### gpt-5.4-mini / Codex 模型

这类模型**不能**走 `/chat/completions` 直连 Copilot，必须用 **Responses API**。本 Worker 已支持两种方式：

**方式 1（推荐）**：仍用 Chat Completions，Worker 会自动转发到 `/responses`：

```bash
curl http://localhost:8787/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4-mini",
    "max_tokens": 256,
    "messages": [{"role": "user", "content": "Hello"}]
  }'
```

**方式 2**：直接调用 OpenAI Responses 格式：

```bash
curl http://localhost:8787/v1/responses \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4-mini",
    "input": "Hello",
    "max_output_tokens": 256
  }'
```

注意：`max_output_tokens` 最小为 **16**。

### 流式 / 工具 / 视觉（gpt-5.4-mini 等）

`/v1/chat/completions` 与 `/v1/messages` 均已支持：

| 能力 | OpenAI Chat | Anthropic Messages |
|------|-------------|-------------------|
| 流式 | `stream: true` | `stream: true` |
| 工具 | `tools` + `tool_choice` | `tools` + `tool_choice` |
| 图片 | `image_url` 多模态 content | `image` + `source.base64` / `url` |

工具调用时 Chat 返回 `finish_reason: tool_calls`；Claude 返回 `stop_reason: tool_use` 与 `tool_use` 内容块。

### curl (OpenAI)

```bash
curl http://localhost:8787/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [{"role": "user", "content": "Hello"}]
  }'
```

### curl (Anthropic)

```bash
curl http://localhost:8787/v1/messages \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello"}]
  }'
```

### OpenAI Python SDK

```python
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["API_KEY"],
    base_url="https://<your-worker>.workers.dev/v1",
)
print(client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "hi"}],
))
```

### Anthropic Python SDK

```python
import anthropic

client = anthropic.Anthropic(
    api_key=os.environ["API_KEY"],
    base_url="https://<your-worker>.workers.dev",
)
print(client.messages.create(
    model="claude-sonnet-4",
    max_tokens=1024,
    messages=[{"role": "user", "content": "hi"}],
))
```

## Environment variables

| Variable | Source | Description |
|----------|--------|-------------|
| `API_KEY` | `wrangler.jsonc` vars or secret | Client API key (default `sk-shumengya666`, not exposed on `GET /`) |
| `GITHUB_TOKEN` | Secret / `.dev.vars` | GitHub OAuth or PAT with Copilot |
| `COPILOT_API_BASE` | vars (optional) | Default `https://api.githubcopilot.com`; set for Enterprise |

## Security notes

- Do **not** commit `.dev.vars` or real tokens to git.
- The default API key is public in `wrangler.jsonc`. For public deployments, set a strong `API_KEY` secret and consider Cloudflare Access or IP restrictions.
- Anyone with the API key can consume your Copilot quota.

## License

MIT