tokhub

English | 中文

---

English

An AI API aggregation gateway that provides a unified OpenAI-compatible interface, supporting intelligent routing and token consumption statistics.

Part of the tokflux organization: https://gitlink.org.cn/tokflux

Features

• User registration/login with JWT authentication
• OpenAI-compatible API interface
• Multi-provider support (Zhipu GLM, Alibaba Qwen, Paratera)
• Users import their own API keys
• Intelligent routing (price-first / latency-first / round-robin)
• Health checks and automatic failover
• Streaming response support (SSE)
• Token usage statistics (daily / by model)

Security Features

• Password hashing with bcrypt
• API key encryption with AES-GCM
• JWT token authentication
• Rate limiting protection
• CORS middleware

Quick Start

Requirements

• Go 1.19+
• Node.js 18+
• SQLite3

Local Development

git clone https://gitlink.org.cn/tokflux/tokhub.git
cd tokhub
go build -o server ./cmd/server
./server

# Frontend development
cd web && npm install && npm run dev

Docker Deployment

docker-compose up -d

Configuration

providers:
  zhipu:
    enabled: true
    base_url: "https://open.bigmodel.cn/api/paas/v4"
    models:
      - name: "glm-4"
        pricing: {input: 0.1, output: 0.1}
  paratera:
    enabled: true
    base_url: "https://llmapi.paratera.com"
    models:
      - name: "GLM-5"
        pricing: {input: 0.08, output: 0.08}

auth:
  admin_key: "your-admin-password"
  secret_key: "your-secret-key"

API Usage

# Register
curl -X POST http://localhost:8080/auth/register \
  -d '{"username": "user", "password": "pass"}'

# Login
curl -X POST http://localhost:8080/auth/login \
  -d '{"username": "user", "password": "pass"}'

# Import Key
curl -X POST http://localhost:8080/user/keys \
  -H "Authorization: Bearer <token>" \
  -d '{"provider": "paratera", "api_key": "xxx", "model": "GLM-5"}'

# Call API
curl -X POST http://localhost:8080/v1/chat/completions \
  -H "Authorization: Bearer <token>" \
  -d '{"model": "GLM-5", "messages": [{"role": "user", "content": "hi"}]}'

Supported Providers

Provider	Models	Streaming
Paratera	GLM-5, glm-4, qwen-plus	Yes
Zhipu	glm-4, glm-3-turbo	Yes
Alibaba	qwen-plus, qwen-turbo	Yes

Documentation

• User Guide - Detailed usage instructions
• API Reference - API specification
• Architecture - System design
• Deployment - Deployment guide
• Contributing - How to contribute
• Security - Security policy
• Changelog - Version history

Architecture

         fluxcore
            ↑
       ┌────┴────┐
       │         │
  tokrouter   tokhub

fluxcore provides:

• Routing strategies (quota_first, price_first, latency_first, round_robin, manual)
• Format conversion (OpenAI ↔ Anthropic)
• Token estimation (Chinese-aware algorithm)
• Pricing calculation
• Input validation
• TraceID generation

Related Projects

• fluxcore - Shared core library (tokflux/fluxcore)
• tokrouter - Personal client tool (tokflux/tokrouter)
• tokmeter - Token metering tool (SDK)
  • Go SDK: tokflux/tokmeter-go
  • TypeScript SDK: @tokflux/meter

License

MIT © 2026 tokflux

---

中文

聚合多个云平台的大模型API，提供统一OpenAI兼容接口，支持智能路由和Token消耗统计。

tokflux 组织项目：https://gitlink.org.cn/tokflux

功能特性

• 用户名+密码注册登录（JWT认证）
• OpenAI兼容API接口
• 多Provider支持（智谱GLM、阿里Qwen、并行科技）
• 用户导入自己的API Key
• 智能路由（价格优先/延迟优先/轮询）
• 健康检查与自动故障转移
• 流式响应支持（SSE）
• Token消耗统计（按天/按模型）

安全特性

• 密码加盐哈希存储（bcrypt）
• API Key加密存储（AES-GCM）
• JWT Token认证
• 请求限流保护
• CORS 中间件

快速开始

环境要求

• Go 1.19+
• Node.js 18+
• SQLite3

本地运行

git clone https://gitlink.org.cn/tokflux/tokhub.git
cd tokhub
go build -o server ./cmd/server
./server

# 前端开发
cd web && npm install && npm run dev

Docker部署

docker-compose up -d

配置说明

providers:
  zhipu:
    enabled: true
    base_url: "https://open.bigmodel.cn/api/paas/v4"
    models:
      - name: "glm-4"
        pricing: {input: 0.1, output: 0.1}
  paratera:
    enabled: true
    base_url: "https://llmapi.paratera.com"
    models:
      - name: "GLM-5"
        pricing: {input: 0.08, output: 0.08}

auth:
  admin_key: "your-admin-password"
  secret_key: "your-secret-key"

API使用

# 注册
curl -X POST http://localhost:8080/auth/register \
  -d '{"username": "user", "password": "pass"}'

# 登录
curl -X POST http://localhost:8080/auth/login \
  -d '{"username": "user", "password": "pass"}'

# 导入Key
curl -X POST http://localhost:8080/user/keys \
  -H "Authorization: Bearer <token>" \
  -d '{"provider": "paratera", "api_key": "xxx", "model": "GLM-5"}'

# 调用API
curl -X POST http://localhost:8080/v1/chat/completions \
  -H "Authorization: Bearer <token>" \
  -d '{"model": "GLM-5", "messages": [{"role": "user", "content": "hi"}]}'

支持的Provider

Provider	模型	流式
并行科技	GLM-5, glm-4, qwen-plus	✅
智谱	glm-4, glm-3-turbo	✅
阿里云	qwen-plus, qwen-turbo	✅

文档

• 用户手册 - 详细使用说明
• API规范 - API接口文档
• 系统架构 - 系统设计
• 部署指南 - 部署说明
• 贡献指南 - 如何参与贡献
• 安全策略 - 安全说明
• 更新日志 - 版本历史

系统架构

         fluxcore
            ↑
       ┌────┴────┐
       │         │
  tokrouter   tokhub

fluxcore 提供：

• 路由策略（quota_first, price_first, latency_first, round_robin, manual）
• 格式转换（OpenAI ↔ Anthropic）
• Token 估算（中文感知算法）
• 价格计算
• 输入验证
• TraceID 生成

相关项目

• fluxcore - 共享核心库 (tokflux/fluxcore)
• tokrouter - 个人客户端工具 (tokflux/tokrouter)
• tokmeter - Token计量工具（SDK）
  • Go SDK: tokflux/tokmeter-go
  • JS SDK: @tokflux/meter

许可证

MIT © 2026 tokflux