#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
智能轮询派单系统
- 自动轮询派单队列
- 负载均衡：轮流分配 + 工作量感知
- 自动重试机制
- 详细日志记录
"""

import requests
import time
import json
from datetime import datetime
from collections import deque

# 配置
API_BASE = "http://1.12.50.92:8005"
API_KEY = "tuhui_dispatch_key_2026"
HEADERS = {"X-API-Key": API_KEY}

# 轮询间隔（秒）
POLL_INTERVAL = 10
MAX_RETRIES = 3
RETRY_DELAY = 3

# 负载均衡：记录最近分配的设计师（轮询分配）
recent_assignments = deque(maxlen=10)

def log(message: str, level: str = "INFO"):
    """打印日志"""
    timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
    print(f"[{timestamp}] [{level}] {message}")

def get_online_designers() -> list:
    """获取在线设计师"""
    try:
        response = requests.get(f"{API_BASE}/online/designers", headers=HEADERS, timeout=5)
        if response.status_code == 200:
            data = response.json()
            return data["online_users"]
    except Exception as e:
        log(f"获取在线设计师失败：{e}", "ERROR")
    return []

def get_pending_tasks(limit: int = 10) -> list:
    """获取待分配任务"""
    try:
        response = requests.get(f"{API_BASE}/tasks/pending", headers=HEADERS, timeout=5, params={"limit": limit})
        if response.status_code == 200:
            data = response.json()
            return data["tasks"]
    except Exception as e:
        log(f"获取待分配任务失败：{e}", "ERROR")
    return []

def get_designer_workload() -> dict:
    """获取设计师工作量"""
    try:
        response = requests.get(f"{API_BASE}/designers/workload", headers=HEADERS, timeout=5)
        if response.status_code == 200:
            data = response.json()
            # 转换为字典：设计师 -> 待处理任务数
            return {d["designer"]: d["pending_tasks"] for d in data["designers"]}
    except Exception as e:
        log(f"获取工作量失败：{e}", "ERROR")
    return {}

def select_designer_smart(online_designers: list) -> str:
    """
    智能选择设计师（负载均衡）
    
    策略：
    1. 优先选最近没被分配的设计师（轮询）
    2. 如果都有任务，选工作量最少的
    3. 如果都没数据，随机选
    """
    if not online_designers:
        return None
    
    if len(online_designers) == 1:
        return online_designers[0]
    
    # 策略 1：找最近没被分配的
    for designer in online_designers:
        if designer not in recent_assignments:
            log(f"选择设计师（轮询）：{designer}")
            return designer
    
    # 策略 2：找工作量最少的
    workload = get_designer_workload()
    min_workload = float('inf')
    selected = None
    
    for designer in online_designers:
        pending = workload.get(designer, 0)
        if pending < min_workload:
            min_workload = pending
            selected = designer
    
    if selected:
        log(f"选择设计师（工作量最少：{min_workload}）：{selected}")
        return selected
    
    # 策略 3：随机选第一个
    log(f"选择设计师（默认）：{online_designers[0]}")
    return online_designers[0]

def assign_task_with_retry(task_id: str, designer: str, max_retries: int = MAX_RETRIES) -> bool:
    """带重试的分配任务"""
    for attempt in range(max_retries):
        try:
            response = requests.post(
                f"{API_BASE}/tasks/{task_id}/assign",
                headers=HEADERS,
                json={
                    "designer_name": designer,
                    "notes": "系统自动派单"
                },
                timeout=5
            )
            
            if response.status_code == 200:
                log(f"✅ 任务 {task_id} 已分配给 {designer}", "SUCCESS")
                # 记录分配历史
                recent_assignments.append(designer)
                return True
            else:
                log(f"派单失败 (尝试 {attempt+1}/{max_retries}): {response.text}", "ERROR")
                
        except Exception as e:
            log(f"网络错误 (尝试 {attempt+1}/{max_retries}): {e}", "ERROR")
        
        if attempt < max_retries - 1:
            time.sleep(RETRY_DELAY)
    
    return False

def poll_and_dispatch():
    """主轮询函数"""
    log("🚀 智能轮询派单系统启动", "START")
    log(f"轮询间隔：{POLL_INTERVAL}秒")
    log(f"API 地址：{API_BASE}")
    
    consecutive_errors = 0
    max_consecutive_errors = 10
    
    while True:
        try:
            # 1. 获取待分配任务
            tasks = get_pending_tasks(limit=20)
            
            if not tasks:
                log("📭 暂无待分配任务")
                consecutive_errors = 0
                time.sleep(POLL_INTERVAL)
                continue
            
            log(f"📋 待分配任务：{len(tasks)} 个")
            
            # 2. 获取在线设计师
            online_designers = get_online_designers()
            
            if not online_designers:
                log("😴 暂无设计师在线")
                consecutive_errors = 0
                time.sleep(POLL_INTERVAL)
                continue
            
            log(f"👥 在线设计师：{len(online_designers)} 人 - {', '.join(online_designers)}")
            
            # 3. 按优先级排序任务
            tasks.sort(key=lambda x: x["priority"], reverse=True)
            
            # 4. 智能分配
            assigned_count = 0
            for task in tasks:
                designer = select_designer_smart(online_designers)
                
                if designer:
                    success = assign_task_with_retry(task["task_id"], designer)
                    if success:
                        assigned_count += 1
                        # 短暂等待，避免并发冲突
                        time.sleep(0.5)
                    else:
                        log(f"❌ 任务 {task['task_id']} 分配失败", "ERROR")
                        break  # 停止分配，下次轮询再试
                else:
                    log("⚠️ 无法选择设计师", "ERROR")
                    break
            
            log(f"📊 本轮分配：{assigned_count}/{len(tasks)} 个任务", "SUCCESS")
            consecutive_errors = 0
            
        except Exception as e:
            log(f"❌ 轮询异常：{e}", "ERROR")
            consecutive_errors += 1
            
            if consecutive_errors >= max_consecutive_errors:
                log(f"⚠️ 连续 {consecutive_errors} 次错误，等待 30 秒...", "WARNING")
                time.sleep(30)
                consecutive_errors = 0
            else:
                time.sleep(POLL_INTERVAL)
        
        # 正常轮询间隔
        time.sleep(POLL_INTERVAL)

if __name__ == "__main__":
    try:
        poll_and_dispatch()
    except KeyboardInterrupt:
        log("👋 轮询派单系统已停止", "INFO")