knowledge-agent/start_service.sh

#!/bin/bash

# Agent 服务管理脚本

# 服务配置
SERVICE_NAME="knowledge-agent"
PID_FILE="/tmp/${SERVICE_NAME}.pid"
LOG_FILE="logs/agent.log"
PORT=8080

# 创建日志目录
mkdir -p logs

# 显示帮助信息
show_help() {
    echo "Usage: $0 {start|stop|restart|status|logs}"
    echo ""
    echo "Commands:"
    echo "  start   - 启动服务"
    echo "  stop    - 停止服务"
    echo "  restart - 重启服务"
    echo "  status  - 查看服务状态"
    echo "  logs    - 查看实时日志"
    echo ""
    echo "Examples:"
    echo "  $0 start    # 启动服务"
    echo "  $0 stop     # 停止服务"
    echo "  $0 logs     # 查看实时日志"
    echo "  tail -f logs/agent.log  # 也可以直接使用 tail 命令"
}

# 检查服务是否运行
is_running() {
    if [ -f "$PID_FILE" ]; then
        pid=$(cat "$PID_FILE")
        if ps -p "$pid" > /dev/null 2>&1; then
            return 0
        else
            rm -f "$PID_FILE"
        fi
    fi
    return 1
}

# 启动服务
start_service() {
    echo "🚀 启动 ${SERVICE_NAME} 服务..."
    
    # 检查Python环境
    if ! command -v python3 &> /dev/null; then
        echo "❌ 错误: 未找到 python3 命令"
        exit 1
    fi
    
    # 检查依赖
    echo "📦 检查依赖..."
    python3 -c "import fastapi, uvicorn" 2>/dev/null
    if [ $? -ne 0 ]; then
        echo "❌ 错误: 缺少必要的依赖包"
        echo "请运行: pip install -r requirements.txt"
        exit 1
    fi
    
    # 检查 httpx
    echo "🔍 检查 httpx..."
    python3 -c "import httpx" 2>/dev/null
    if [ $? -ne 0 ]; then
        echo "❌ 错误: 缺少 httpx 依赖包"
        echo "请运行: pip install httpx"
        exit 1
    fi
    
    # 检查 LangGraph
    echo "🔍 检查 LangGraph..."
    python3 -c "import langgraph" 2>/dev/null
    if [ $? -ne 0 ]; then
        echo "⚠️  警告: LangGraph 未安装，将使用传统模式"
        echo "如需启用 LangGraph，请运行: pip install langgraph"
        echo ""
    fi
    
    # 检查服务是否已经运行
    if is_running; then
        echo "⚠️  警告: 服务已经在运行中 (PID: $(cat "$PID_FILE"))"
        echo "如需重启，请使用: $0 restart"
        exit 1
    fi
    
    # 启动服务
    echo "🌟 启动服务..."
    echo "📍 服务地址: http://localhost:${PORT}"
    echo "📚 API文档: http://localhost:${PORT}/docs"
    echo "🔍 健康检查: http://localhost:${PORT}/health"
    echo "📝 日志文件: ${LOG_FILE}"
    echo ""
    
    # 后台启动服务，记录PID
    # 生产环境禁用热重载以减少日志噪音
    export RELOAD_ENABLED=false
    export LOG_LEVEL=info

    export LANGCHAIN_TRACING_V2=true
    export LANGCHAIN_API_KEY=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
    export LANGCHAIN_PROJECT=knowledge-agent
    


    nohup python3 agent.py > "$LOG_FILE" 2>&1 &
    echo $! > "$PID_FILE"
    
    # 等待服务启动
    sleep 2
    
    if is_running; then
        echo "✅ 服务启动成功! (PID: $(cat "$PID_FILE"))"
        echo "📝 查看日志: $0 logs 或 tail -f ${LOG_FILE}"
        echo "🛑 停止服务: $0 stop"
    else
        echo "❌ 服务启动失败，请检查日志: ${LOG_FILE}"
        rm -f "$PID_FILE"
        exit 1
    fi
}

# 停止服务
stop_service() {
    echo "🛑 停止 ${SERVICE_NAME} 服务..."
    
    if is_running; then
        pid=$(cat "$PID_FILE")
        echo "正在停止进程 PID: $pid"
        
        # 尝试优雅停止
        kill "$pid" 2>/dev/null
        
        # 等待进程结束
        for i in {1..15}; do
            if ! ps -p "$pid" > /dev/null 2>&1; then
                break
            fi
            echo "等待进程结束... ($i/15)"
            sleep 1
        done
        
        # 如果进程仍然存在，强制杀死
        if ps -p "$pid" > /dev/null 2>&1; then
            echo "强制停止进程..."
            kill -9 "$pid" 2>/dev/null
        fi
        
        rm -f "$PID_FILE"
        echo "✅ 主进程已停止"
    else
        echo "ℹ️  主进程未运行"
    fi
    
    # 额外清理：查找并终止所有相关进程
    echo "🔍 检查是否有残留进程..."
    REMAINING_PROCESSES=$(ps aux | grep -E "(python.*agent|knowledge-agent)" | grep -v grep | awk '{print $2}')
    
    if [ ! -z "$REMAINING_PROCESSES" ]; then
        echo "⚠️  发现残留进程，正在清理..."
        for pid in $REMAINING_PROCESSES; do
            echo "终止残留进程 $pid..."
            kill -TERM $pid 2>/dev/null
        done
        
        # 等待2秒
        sleep 2
        
        # 检查是否还有残留
        STILL_REMAINING=$(ps aux | grep -E "(python.*agent|knowledge-agent)" | grep -v grep | awk '{print $2}')
        if [ ! -z "$STILL_REMAINING" ]; then
            echo "强制终止顽固进程..."
            for pid in $STILL_REMAINING; do
                echo "强制终止进程 $pid..."
                kill -9 $pid 2>/dev/null
            done
        fi
        
        echo "✅ 残留进程已清理"
    else
        echo "✅ 没有发现残留进程"
    fi
}

# 重启服务
restart_service() {
    echo "🔄 重启 ${SERVICE_NAME} 服务..."
    stop_service
    sleep 2
    start_service
}

# 查看服务状态
show_status() {
    echo "🔍 ${SERVICE_NAME} 服务状态:"
    
    if is_running; then
        pid=$(cat "$PID_FILE")
        echo "✅ 状态: 运行中"
        echo "📊 PID: $pid"
        echo "📍 端口: $PORT"
        echo "📝 日志: $LOG_FILE"
        echo "🔗 地址: http://localhost:${PORT}"
        
        # 显示进程信息
        echo ""
        echo "📋 进程信息:"
        ps -p "$pid" -o pid,ppid,cmd,etime
    else
        echo "❌ 状态: 未运行"
        echo "💡 启动服务: $0 start"
    fi
}

# 查看实时日志
show_logs() {
    echo "📝 查看实时日志 (按 Ctrl+C 退出)..."
    echo "日志文件: ${LOG_FILE}"
    echo ""
    
    if [ -f "$LOG_FILE" ]; then
        tail -f "$LOG_FILE"
    else
        echo "❌ 日志文件不存在: ${LOG_FILE}"
        echo "请先启动服务: $0 start"
        exit 1
    fi
}

# 主逻辑
case "$1" in
    start)
        start_service
        ;;
    stop)
        stop_service
        ;;
    restart)
        restart_service
        ;;
    status)
        show_status
        ;;
    logs)
        show_logs
        ;;
    *)
        show_help
        exit 1
        ;;
esac