WindsurfPoolAPI

⚠️ Disclaimer / 声明

This project is for personal learning, research, and self-hosting only. Commercial use, resale, paid deployment, or repackaging as a service without written authorization is strictly prohibited.

本项目仅供个人学习、研究、自用。未经作者书面授权，禁止任何商业用途、付费代部署、中转转售或包装成服务对外销售。

✨ Features / 核心特性

| Feature | Description | | :--- | :--- | | Dual Protocol | /v1/chat/completions (OpenAI) + /v1/messages (Anthropic native) | | 113+ Models | Claude Opus 4.7 · GPT-5.4 · Gemini 3.1 · DeepSeek R1 · Grok 3 · Qwen 3 · Kimi K2.5 · GLM-5.1 and more | | Multi-Account Pool | Capacity-based load balancing, automatic failover, per-model rate-limit isolation | | Token & Credit Analytics | Per-API × per-model aggregation down to individual request level | | Admin Dashboard | Full-featured SPA: account management, proxy config, real-time logs, usage charts | | Batch Operations | Select multiple accounts, enable/disable in one click | | OAuth Login | Google / GitHub Firebase OAuth + manual token refresh | | Dynamic Stall Detection | Input-length-aware timeout (30s–90s) prevents false positives on large contexts | | Persistent State | All settings, account status, tokens survive restarts | | Image Upload | Multimodal support — send images via image_url blocks (base64 or URL) | | Tool Calling | <tool_call> protocol compatible — works with Cursor, Aider, and other AI coding tools | | Cursor Compatible | 80+ model name aliases including Cursor-friendly names without "claude" keyword | | Streaming SSE | OpenAI format with stream_options.include_usage support | | Zero Dependencies | Pure Node.js built-in modules, nothing to install |

• 双协议兼容 — OpenAI + Anthropic 原生端点，无需任何中间件
• 113+ 模型 — 启动时自动拉取 Windsurf 最新 catalog，实时更新
• 多账号池 — 按剩余容量均衡分配，自动故障转移，per-model 限速隔离
• Token + Credit 精细统计 — 按 API × 模型分层聚合，精确到单次请求
• Dashboard 管理后台 — 账号管理、代理配置、实时日志、使用图表、封禁侦测
• 批量操作 — 一键多选账号批量启用/停用
• OAuth 登录 — 支持 Google/GitHub Firebase OAuth 登录
• 动态超时检测 — 根据输入长度自适应超时阈值（30s~90s），大上下文不误判
• 全持久化 — 所有设置、账号状态、Token 均持久化存储，重启不丢失
• 零依赖 — 纯 Node.js 内置模块，开箱即用

🚀 Quick Start / 快速开始

Prerequisites / 前置条件

• Node.js ≥ 20
• Windsurf Language Server binary (language_server_linux_x64 or language_server_darwin_arm64)
• At least one Windsurf account (Free tier supports limited models)

Install & Run / 安装启动

git clone https://github.com/guanxiaol/WindsurfPoolAPI.git
cd WindsurfPoolAPI

# Place Language Server binary / 放置 Language Server 二进制
sudo mkdir -p /opt/windsurf
sudo cp /path/to/language_server_linux_x64 /opt/windsurf/
sudo chmod +x /opt/windsurf/language_server_linux_x64

# Optional: configure / 可选配置
cp .env.example .env    # Edit API_KEY, DASHBOARD_PASSWORD, etc.

# Start / 启动
node src/index.js

macOS — Run bash scripts/install-macos.sh for auto-start on login.

Windows — Run scripts\install-windows.bat for guided installation.

Dashboard: http://localhost:3003/dashboard

Docker

docker compose up -d --build

Mount the LS binary at /opt/windsurf/ on the host before starting.

🔑 Account Management / 账号管理

⚠️ Always use Token login! / 必须使用 Token 方式登录！

Windsurf has a known bug where email/password login may route requests to the wrong account.

Windsurf 官方存在 bug：邮箱/密码登录可能导致请求路由到错误账号。

Get your token / 获取 Token：https://windsurf.com/editor/show-auth-token?workflow=

# ✅ Add account via Token (recommended / 推荐)
curl -X POST http://localhost:3003/auth/login \
  -H "Content-Type: application/json" \
  -d '{"token": "your-windsurf-token"}'

# Batch add / 批量添加
curl -X POST http://localhost:3003/auth/login \
  -H "Content-Type: application/json" \
  -d '{"accounts": [{"token": "t1"}, {"token": "t2"}]}'

# List accounts / 列出账号
curl http://localhost:3003/auth/accounts

# Remove / 删除
curl -X DELETE http://localhost:3003/auth/accounts/{id}

📡 API Reference / 接口文档

OpenAI Compatible / OpenAI 兼容

curl http://localhost:3003/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-your-api-key" \
  -d '{
    "model": "gpt-4o-mini",
    "messages": [{"role": "user", "content": "Hello!"}],
    "stream": false
  }'

Anthropic Compatible / Anthropic 兼容

curl http://localhost:3003/v1/messages \
  -H "Content-Type: application/json" \
  -H "anthropic-version: 2023-06-01" \
  -H "x-api-key: sk-your-api-key" \
  -d '{
    "model": "claude-sonnet-4.6",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Environment Variables / 环境变量

| Variable | Default | Description | | :--- | :--- | :--- | | PORT | 3003 | HTTP server port | | API_KEY | (empty) | Auth key for /v1/* endpoints. Empty = open access | | DASHBOARD_PASSWORD | (empty) | Dashboard admin password | | DEFAULT_MODEL | claude-4.5-sonnet-thinking | Default model when none specified | | MAX_TOKENS | 8192 | Default max output tokens | | LOG_LEVEL | info | debug / info / warn / error | | LS_BINARY_PATH | /opt/windsurf/language_server_linux_x64 | Language Server path | | LS_PORT | 42100 | Language Server gRPC port |

Dashboard API

All endpoints require X-Dashboard-Password header.

| Method | Path | Description | | :--- | :--- | :--- | | GET | /dashboard/api/overview | System overview | | GET | /dashboard/api/accounts | List all accounts | | POST | /dashboard/api/accounts/batch-status | Batch enable/disable accounts | | POST | /dashboard/api/oauth-login | OAuth login (Google/GitHub) | | POST | /dashboard/api/accounts/:id/refresh-token | Refresh Firebase token | | POST | /dashboard/api/accounts/:id/rate-limit | Check account capacity | | GET | /dashboard/api/usage | Full usage statistics | | GET | /dashboard/api/usage/export | Export stats as JSON | | POST | /dashboard/api/usage/import | Import stats (auto-dedup) | | GET | /dashboard/api/logs/stream | Real-time SSE log stream |

🖥 Dashboard / 管理后台

Access at http://localhost:3003/dashboard

| Panel | Description | | :--- | :--- | | Overview | Runtime stats, account pool health, success rate | | Login | Windsurf token/email login, OAuth | | Accounts | Add/remove, batch enable/disable, per-account proxy, quota display | | Models | Global allow/blocklist, per-account model restrictions | | Proxy | Global + per-account HTTP/SOCKS5 proxy | | Logs | Real-time SSE log stream with level filtering | | Analytics | Token/Credit charts, 14-day trends, 24h distribution, request details | | Detection | Error pattern monitoring, account health | | Experimental | Cascade session reuse, model identity masking, preflight rate-limit |

Screenshots / 界面预览

🤖 Supported Models / 支持的模型

claude-3.5-sonnet · claude-3.7-sonnet[-thinking] · claude-4-sonnet[-thinking] · claude-4-opus[-thinking] · claude-4.1-opus[-thinking] · claude-4.5-sonnet[-thinking] · claude-4.5-haiku · claude-4.5-opus[-thinking] · claude-sonnet-4.6[-thinking][-1m] · claude-opus-4.6[-thinking] · claude-opus-4.7-{low,medium,high,xhigh,max}

gpt-4o · gpt-4o-mini · gpt-4.1[-mini/nano] · gpt-5[-mini] · gpt-5.2[-low/medium/high] · gpt-5.4[-low/medium/high/xhigh] · gpt-5.3-codex · o3[-mini/high/pro] · o4-mini

gemini-2.5-pro · gemini-2.5-flash · gemini-3.0-pro · gemini-3.0-flash · gemini-3.1-pro[-low/high]

deepseek-v3 · deepseek-r1 · grok-3[-mini] · grok-code-fast-1 · qwen-3 · qwen-3-coder · kimi-k2 · kimi-k2.5 · swe-1.5[-thinking] · swe-1.6-fast · arena-fast · arena-smart

Model catalog is auto-synced from Windsurf cloud on startup. Free accounts: gpt-4o-mini and gemini-2.5-flash only.

启动时自动从 Windsurf 云端拉取最新模型列表。免费账号仅可用 `gpt