diff --git a/.gitignore b/.gitignore index 4d6155f..bbc72e3 100644 --- a/.gitignore +++ b/.gitignore @@ -22,5 +22,11 @@ HA/config/ HA/db_data/ +# Router specific ignores +router/__pycache__/ +router/venv/ +router/*.pyc +router/*.log + # Keep structure !.gitkeep \ No newline at end of file diff --git a/docker-compose.yml b/docker-compose.yml index b3af26a..75c94ed 100644 --- a/docker-compose.yml +++ b/docker-compose.yml @@ -140,7 +140,7 @@ services: container_name: sillytavern restart: unless-stopped ports: - - "8000:8000" + - "8002:8000" # Changed port to avoid conflict with claude-router environment: - TZ=America/Chicago volumes: @@ -148,6 +148,26 @@ services: networks: - web_network + # Claude Router - AI API智能路由器 + claude-router: + build: ./router + container_name: claude-router + restart: unless-stopped + ports: + - "8000:8000" + environment: + - CLAUDE_API_KEY=${CLAUDE_API_KEY} + volumes: + - ./tokens.txt:/home/will/docker/tokens.txt:ro + networks: + - web_network + healthcheck: + test: ["CMD", "curl", "-f", "http://localhost:8000/health"] + interval: 30s + timeout: 10s + retries: 3 + start_period: 30s + volumes: discord_config: discord_logs: diff --git a/router/Dockerfile b/router/Dockerfile new file mode 100644 index 0000000..a679451 --- /dev/null +++ b/router/Dockerfile @@ -0,0 +1,32 @@ +FROM python:3.11-slim + +# Set working directory +WORKDIR /app + +# Install system dependencies +RUN apt-get update && apt-get install -y \ + curl \ + && rm -rf /var/lib/apt/lists/* + +# Copy requirements first for better caching +COPY requirements.txt . + +# Install Python dependencies +RUN pip install --no-cache-dir -r requirements.txt + +# Copy application code +COPY . . + +# Create non-root user +RUN useradd -m -u 1000 router && chown -R router:router /app +USER router + +# Expose port +EXPOSE 8000 + +# Health check +HEALTHCHECK --interval=30s --timeout=10s --start-period=30s --retries=3 \ + CMD curl -f http://localhost:8000/health || exit 1 + +# Run the application +CMD ["python", "app.py"] \ No newline at end of file diff --git a/router/README.md b/router/README.md new file mode 100644 index 0000000..096384f --- /dev/null +++ b/router/README.md @@ -0,0 +1,238 @@ +# Claude Router + +一个智能的Claude API路由器，支持Claude Pro和Claude API之间的自动故障转移。当Claude Pro达到使用限制时，自动切换到Claude API，确保服务的连续性。 + +## 功能特性 + +- **自动故障转移**: 检测到速率限制或使用限制时自动切换provider +- **定时健康检查**: 每小时前5分钟自动检测Claude Pro限额恢复 +- **智能恢复**: 自动切换回Claude Pro，优先使用高级功能 +- **手动切换**: 支持手动切换到指定provider +- **兼容Claude Code CLI**: 完全兼容Anthropic API格式 +- **Docker化部署**: 一键部署，开箱即用 + +## 快速开始 + +### 1. 使用Docker Compose部署 + +```bash +# 克隆或进入项目目录 +cd /home/will/docker/router + +# 构建并启动服务 +docker-compose up -d + +# 查看服务状态 +docker-compose ps +``` + +### 2. 验证服务运行 + +```bash +# 健康检查 +curl http://localhost:8000/health + +# 查看当前状态 +curl http://localhost:8000/v1/status +``` + +### 3. 配置Claude Code CLI + +设置环境变量将Claude Code CLI指向路由器： + +```bash +# 设置API endpoint为路由器地址 +export ANTHROPIC_API_URL="http://localhost:8000" + +# 添加到bashrc使其永久生效 +echo 'export ANTHROPIC_API_URL="http://localhost:8000"' >> ~/.bashrc + +# 测试配置 +echo "Hello Claude Router" | claude --print +``` + +**注意**: 无需修改ANTHROPIC_API_KEY，路由器会自动处理API密钥。 + +## API端点 + +### 主要端点 + +- `POST /v1/messages` - Claude API消息创建（兼容Anthropic API） +- `GET /health` - 健康检查 +- `GET /v1/status` - 获取路由器状态 +- `POST /v1/switch-provider` - 手动切换provider +- `POST /v1/health-check` - 手动触发Claude Pro健康检查 + +### 健康检查响应示例 + +```json +{ + "status": "healthy", + "current_provider": "claude_pro", + "failover_count": 0, + "last_failover": null, + "last_health_check": "2025-07-14T19:00:00.000Z", + "health_check_failures": 0, + "providers": { + "claude_pro": {"active": true}, + "claude_api": {"active": true} + } +} +``` + +## 配置说明 + +### 环境变量 + +- `CLAUDE_API_KEY`: Claude API密钥 +- `ROUTER_HOST`: 服务监听地址（默认: 0.0.0.0） +- `ROUTER_PORT`: 服务监听端口（默认: 8000） +- `MAX_RETRIES`: 最大重试次数（默认: 3） +- `RETRY_DELAY`: 重试延迟（默认: 1.0秒） + +### 健康检查配置 + +- `health_check_enabled`: 是否启用定时健康检查（默认: true） +- `health_check_cron`: 检查时间表达式（默认: "0-4 * * * *" - 每小时前5分钟） +- `health_check_message`: 测试消息内容（默认: "ping"） +- `health_check_model`: 使用的模型（默认: claude-3-haiku-20240307） + +### Token文件 + +路由器会自动从 `/home/will/docker/tokens.txt` 读取API密钥，无需手动配置环境变量。 + +## 故障转移机制 + +当检测到以下错误时，路由器会自动切换到下一个可用的provider： + +- 429 (Too Many Requests) +- 速率限制错误 +- 使用限制达到 +- "usage limit reached"相关错误 + +**优先级顺序**: Claude Pro → Claude API + +## 使用示例 + +### 基本API调用 + +```bash +curl -X POST http://localhost:8000/v1/messages \ + -H "Content-Type: application/json" \ + -H "Authorization: Bearer your_api_key" \ + -d '{ + "model": "claude-3-sonnet-20240229", + "max_tokens": 1024, + "messages": [ + {"role": "user", "content": "Hello, Claude!"} + ] + }' +``` + +### 手动切换provider + +```bash +curl -X POST http://localhost:8000/v1/switch-provider \ + -H "Content-Type: application/json" \ + -d '"claude_api"' +``` + +### 手动健康检查 + +```bash +# 立即检测Claude Pro是否可用 +curl -X POST http://localhost:8000/v1/health-check + +# 查看详细状态 +curl http://localhost:8000/v1/status +``` + +## 开发和调试 + +### 本地开发 + +```bash +# 创建虚拟环境 +python3 -m venv venv +source venv/bin/activate + +# 安装依赖 +pip install -r requirements.txt + +# 运行应用 +python app.py +``` + +### 查看日志 + +```bash +# Docker容器日志 +docker-compose logs -f claude-router + +# 实时日志 +docker logs -f claude-router +``` + +## 故障排除 + +### 常见问题 + +1. **服务无法启动** + - 检查tokens.txt文件是否存在且格式正确 + - 确认端口8000未被占用 + +2. **API调用失败** + - 验证API密钥是否有效 + - 检查网络连接到api.anthropic.com + +3. **自动切换不工作** + - 查看日志确认错误检测逻辑 + - 确认backup provider配置正确 + +### 监控 + +- 健康检查: `http://localhost:8000/health` +- 状态监控: `http://localhost:8000/v1/status` +- Docker健康检查: `docker inspect claude-router` + +## 技术架构 + +- **框架**: FastAPI + Uvicorn +- **HTTP客户端**: httpx +- **AI库**: anthropic +- **容器化**: Docker + Docker Compose +- **配置管理**: pydantic + python-dotenv + +## 版本信息 + +- 版本: 1.0.0 (MVP) +- Python: 3.11+ +- 支持: Claude-3 系列模型 + +## 更新日志 + +### v1.1.0 (2025-07-14) +- ✅ 添加定时健康检查功能 +- ✅ 每小时前5分钟自动检测Claude Pro限额恢复 +- ✅ 智能自动切换回Claude Pro +- ✅ 新增手动健康检查API +- ✅ 完善日志记录和状态监控 + +### v1.0.0 (2025-07-14) +- ✅ 基础路由器功能 +- ✅ Claude Pro到Claude API自动故障转移 +- ✅ Docker容器化部署 +- ✅ Claude Code CLI兼容性 + +## 后续开发计划 + +- [ ] 添加DeepSeek API支持 +- [ ] 添加ChatGPT API支持 +- [ ] 实现请求统计和监控面板 +- [ ] 添加请求缓存功能 +- [ ] 支持负载均衡 +- [ ] 集成Kimi v2 API + +## 许可证 + +MIT License \ No newline at end of file diff --git a/router/app.py b/router/app.py new file mode 100644 index 0000000..e453bd9 --- /dev/null +++ b/router/app.py @@ -0,0 +1,338 @@ +import asyncio +import json +import logging +from datetime import datetime +from typing import Dict, Any, Optional +from contextlib import asynccontextmanager + +import httpx +from fastapi import FastAPI, Request, HTTPException +from fastapi.responses import StreamingResponse, JSONResponse +from anthropic import Anthropic +from apscheduler.schedulers.asyncio import AsyncIOScheduler +from apscheduler.triggers.cron import CronTrigger + +from config import config + +# Configure logging +logging.basicConfig(level=logging.INFO) +logger = logging.getLogger(__name__) + +class ClaudeRouter: + def __init__(self): + self.current_provider = "claude_pro" + self.failover_count = 0 + self.last_failover = None + self.last_health_check = None + self.health_check_failures = 0 + self.scheduler = None + self.providers = { + "claude_pro": { + "api_key": config.claude_pro_api_key, + "base_url": config.claude_pro_base_url, + "active": True + }, + "claude_api": { + "api_key": config.claude_api_key, + "base_url": config.claude_api_base_url, + "active": True + } + } + + async def get_anthropic_client(self, provider: str) -> Anthropic: + """Get Anthropic client for the specified provider""" + if provider not in self.providers: + raise ValueError(f"Unknown provider: {provider}") + + provider_config = self.providers[provider] + return Anthropic( + api_key=provider_config["api_key"], + base_url=provider_config["base_url"] + ) + + async def should_failover(self, error: Exception) -> bool: + """Determine if we should failover based on the error""" + error_str = str(error).lower() + + # Check for rate limiting or usage limit errors + failover_indicators = [ + "rate_limit", + "usage limit", + "quota exceeded", + "429", + "too many requests", + "limit reached" + ] + + return any(indicator in error_str for indicator in failover_indicators) + + async def failover_to_next_provider(self): + """Switch to the next available provider""" + providers_list = list(self.providers.keys()) + current_index = providers_list.index(self.current_provider) + + # Try next provider + for i in range(1, len(providers_list)): + next_index = (current_index + i) % len(providers_list) + next_provider = providers_list[next_index] + + if self.providers[next_provider]["active"]: + logger.info(f"Failing over from {self.current_provider} to {next_provider}") + self.current_provider = next_provider + self.failover_count += 1 + self.last_failover = datetime.now() + return True + + logger.error("No active providers available for failover") + return False + + async def make_request(self, request_data: Dict[str, Any]) -> Dict[str, Any]: + """Make request with automatic failover""" + max_attempts = len(self.providers) + + for attempt in range(max_attempts): + try: + client = await self.get_anthropic_client(self.current_provider) + + # Extract parameters from request + messages = request_data.get("messages", []) + model = request_data.get("model", "claude-3-sonnet-20240229") + max_tokens = request_data.get("max_tokens", 4096) + stream = request_data.get("stream", False) + + logger.info(f"Making request with provider: {self.current_provider}") + + # Make the API call + if hasattr(client, 'messages'): + response = await asyncio.to_thread( + client.messages.create, + model=model, + max_tokens=max_tokens, + messages=messages, + stream=stream + ) + else: + # For older anthropic versions + response = await asyncio.to_thread( + client.completions.create, + model=model, + max_tokens_to_sample=max_tokens, + prompt=f"Human: {messages[0]['content']}\n\nAssistant:", + stream=stream + ) + + return response + + except Exception as e: + logger.error(f"Request failed with {self.current_provider}: {str(e)}") + + if await self.should_failover(e) and attempt < max_attempts - 1: + if await self.failover_to_next_provider(): + continue + + # If this is the last attempt or failover failed, raise the error + if attempt == max_attempts - 1: + raise HTTPException(status_code=500, detail=f"All providers failed. Last error: {str(e)}") + + raise HTTPException(status_code=500, detail="No providers available") + + async def health_check_claude_pro(self): + """Check if Claude Pro is available again""" + # Only check if we're not currently using Claude Pro + if self.current_provider == "claude_pro": + logger.debug("Skipping health check - already using Claude Pro") + return + + logger.info("Running Claude Pro health check...") + self.last_health_check = datetime.now() + + try: + client = Anthropic( + api_key=config.claude_pro_api_key, + base_url=config.claude_pro_base_url + ) + + # Send a minimal test message + if hasattr(client, 'messages'): + response = await asyncio.to_thread( + client.messages.create, + model=config.health_check_model, + max_tokens=10, + messages=[{"role": "user", "content": config.health_check_message}] + ) + else: + # For older anthropic versions + response = await asyncio.to_thread( + client.completions.create, + model=config.health_check_model, + max_tokens_to_sample=10, + prompt=f"Human: {config.health_check_message}\n\nAssistant:" + ) + + # If successful, switch back to Claude Pro + old_provider = self.current_provider + self.current_provider = "claude_pro" + self.health_check_failures = 0 + + logger.info(f"Claude Pro health check successful! Switched from {old_provider} to claude_pro") + + except Exception as e: + self.health_check_failures += 1 + error_str = str(e).lower() + + if any(indicator in error_str for indicator in ["rate_limit", "usage limit", "quota exceeded", "429", "too many requests", "limit reached"]): + logger.info(f"Claude Pro still rate limited: {str(e)}") + else: + logger.warning(f"Claude Pro health check failed (attempt {self.health_check_failures}): {str(e)}") + + def start_scheduler(self): + """Start the health check scheduler""" + if not config.health_check_enabled: + logger.info("Health check disabled in config") + return + + self.scheduler = AsyncIOScheduler() + + # Schedule health check using cron expression + self.scheduler.add_job( + self.health_check_claude_pro, + trigger=CronTrigger.from_crontab(config.health_check_cron), + id="claude_pro_health_check", + name="Claude Pro Health Check", + misfire_grace_time=60 + ) + + self.scheduler.start() + logger.info(f"Health check scheduler started with cron: {config.health_check_cron}") + + def stop_scheduler(self): + """Stop the health check scheduler""" + if self.scheduler: + self.scheduler.shutdown() + logger.info("Health check scheduler stopped") + +# Initialize router +router = ClaudeRouter() + +@asynccontextmanager +async def lifespan(app: FastAPI): + logger.info("Claude Router starting up...") + logger.info(f"Current provider: {router.current_provider}") + + # Start health check scheduler + router.start_scheduler() + + yield + + # Stop scheduler on shutdown + router.stop_scheduler() + logger.info("Claude Router shutting down...") + +app = FastAPI( + title="Claude Router", + description="Smart router for Claude API with automatic failover", + version="1.0.0", + lifespan=lifespan +) + +@app.get("/health") +async def health_check(): + """Health check endpoint""" + return { + "status": "healthy", + "current_provider": router.current_provider, + "failover_count": router.failover_count, + "last_failover": router.last_failover.isoformat() if router.last_failover else None, + "providers": { + name: {"active": provider_config["active"]} + for name, provider_config in router.providers.items() + }, + "last_health_check": router.last_health_check.isoformat() if router.last_health_check else None, + "health_check_failures": router.health_check_failures + } + +@app.post("/v1/messages") +async def create_message(request: Request): + """Handle Claude API message creation with failover""" + try: + request_data = await request.json() + stream = request_data.get("stream", False) + + if stream: + # Handle streaming response + async def generate_stream(): + try: + response = await router.make_request(request_data) + for chunk in response: + yield f"data: {json.dumps(chunk.model_dump())}\n\n" + yield "data: [DONE]\n\n" + except Exception as e: + error_data = {"error": str(e)} + yield f"data: {json.dumps(error_data)}\n\n" + + return StreamingResponse( + generate_stream(), + media_type="text/event-stream", + headers={ + "Cache-Control": "no-cache", + "Connection": "keep-alive" + } + ) + else: + # Handle non-streaming response + response = await router.make_request(request_data) + return response.model_dump() + + except Exception as e: + logger.error(f"Request processing failed: {str(e)}") + raise HTTPException(status_code=500, detail=str(e)) + +@app.post("/v1/switch-provider") +async def switch_provider(request: Request): + """Manually switch to a specific provider""" + provider = await request.json() + + if provider not in router.providers: + raise HTTPException(status_code=400, detail=f"Unknown provider: {provider}") + + if not router.providers[provider]["active"]: + raise HTTPException(status_code=400, detail=f"Provider {provider} is not active") + + old_provider = router.current_provider + router.current_provider = provider + + logger.info(f"Manually switched from {old_provider} to {provider}") + + return { + "message": f"Switched from {old_provider} to {provider}", + "current_provider": router.current_provider + } + +@app.get("/v1/status") +async def get_status(): + """Get current router status""" + return { + "current_provider": router.current_provider, + "failover_count": router.failover_count, + "last_failover": router.last_failover.isoformat() if router.last_failover else None, + "last_health_check": router.last_health_check.isoformat() if router.last_health_check else None, + "health_check_failures": router.health_check_failures, + "providers": router.providers + } + +@app.post("/v1/health-check") +async def manual_health_check(): + """Manually trigger Claude Pro health check""" + try: + await router.health_check_claude_pro() + return { + "message": "Health check completed", + "current_provider": router.current_provider, + "last_health_check": router.last_health_check.isoformat() if router.last_health_check else None + } + except Exception as e: + raise HTTPException(status_code=500, detail=f"Health check failed: {str(e)}") + +if __name__ == "__main__": + import uvicorn + uvicorn.run(app, host=config.host, port=config.port) \ No newline at end of file diff --git a/router/config.py b/router/config.py new file mode 100644 index 0000000..3d2797c --- /dev/null +++ b/router/config.py @@ -0,0 +1,54 @@ +import os +from typing import Optional +from pydantic import BaseModel + +class Config(BaseModel): + # Claude API configurations + claude_pro_api_key: str = "" + claude_api_key: str = "" + + # Router settings + port: int = 8000 + host: str = "0.0.0.0" + + # Retry settings + max_retries: int = 3 + retry_delay: float = 1.0 + + # API endpoints + claude_pro_base_url: str = "https://api.anthropic.com" + claude_api_base_url: str = "https://api.anthropic.com" + + # Health check settings + health_check_enabled: bool = True + health_check_cron: str = "0-4 * * * *" # Every hour, first 5 minutes + health_check_message: str = "ping" + health_check_model: str = "claude-3-haiku-20240307" # Use cheapest model for checks + + def __init__(self, **kwargs): + super().__init__(**kwargs) + # Load from environment or token file + self.load_from_env() + + def load_from_env(self): + """Load configuration from environment variables or token file""" + # Try environment variables first + self.claude_api_key = os.getenv("CLAUDE_API_KEY", "") + + # Load from tokens.txt if not found in env + if not self.claude_api_key: + try: + with open("/home/will/docker/tokens.txt", "r") as f: + for line in f: + if line.startswith("claude_API="): + self.claude_api_key = line.split("=", 1)[1].strip() + break + except FileNotFoundError: + pass + + # For MVP, we'll use the same API key for both pro and regular + # In practice, Claude Pro might use a different endpoint or key + self.claude_pro_api_key = self.claude_api_key + +# Global config instance +config = Config() \ No newline at end of file diff --git a/router/docker-compose.yml b/router/docker-compose.yml new file mode 100644 index 0000000..cd212b3 --- /dev/null +++ b/router/docker-compose.yml @@ -0,0 +1,25 @@ +version: '3.8' + +services: + claude-router: + build: . + container_name: claude-router + ports: + - "8000:8000" + environment: + - CLAUDE_API_KEY=${CLAUDE_API_KEY} + volumes: + - /home/will/docker/tokens.txt:/home/will/docker/tokens.txt:ro + restart: unless-stopped + networks: + - router-network + healthcheck: + test: ["CMD", "curl", "-f", "http://localhost:8000/health"] + interval: 30s + timeout: 10s + retries: 3 + start_period: 30s + +networks: + router-network: + driver: bridge \ No newline at end of file diff --git a/router/requirements.txt b/router/requirements.txt new file mode 100644 index 0000000..26138b8 --- /dev/null +++ b/router/requirements.txt @@ -0,0 +1,7 @@ +fastapi==0.104.1 +uvicorn==0.24.0 +httpx==0.25.2 +pydantic==2.5.0 +anthropic==0.7.8 +python-dotenv==1.0.0 +apscheduler==3.10.4 \ No newline at end of file