Files
xianyu-auto-reply/ai_reply_engine.py
T
jinghui 198ccd3976 v1.1.0: AI回复max_tokens调优 + 商品编辑(标题/价格/描述) + API密钥掩码
- ai_reply_engine.py: 意图识别max_tokens=100, 回复生成max_tokens=2048
  (deepseek-v4-flash推理模型思考吃光token导致回复为空)
- 商品管理弹窗新增标题/价格/描述编辑, 写入本地item_info
  (闲鱼网页版不支持改真实商品, 界面已注明)
- GET ai-reply-settings密钥掩码 sk-4****27QK, PUT含****保留原值
- 新增自定义Dockerfile固化max_tokens补丁
- 新增.gitignore排除.env/data/logs/db, 仓库不含密钥
2026-10-22 12:07:09 +08:00

371 lines
15 KiB
Python

"""
AI回复引擎模块
集成XianyuAutoAgent的AI回复功能到现有项目中
"""
import os
import json
import time
import hashlib
import sqlite3
from typing import List, Dict, Optional
from loguru import logger
from openai import OpenAI
from db_manager import db_manager
class AIReplyEngine:
"""AI回复引擎"""
def __init__(self):
self.clients = {} # 存储不同账号的OpenAI客户端
self.agents = {} # 存储不同账号的Agent实例
self._cache = {} # AI回复缓存 {cache_key: (reply, expiry)}
self._intent_cache = {} # 意图缓存 {cache_key: (intent, expiry)}
self._cache_ttl = 3600 # 缓存有效期1小时
self._init_default_prompts()
def _make_cache_key(self, cookie_id: str, chat_id: str, message: str) -> str:
"""生成缓存键"""
raw = f"{cookie_id}|{chat_id}|{message}"
return hashlib.md5(raw.encode()).hexdigest()
def _get_cached_reply(self, cookie_id: str, chat_id: str, message: str) -> Optional[str]:
"""获取缓存的AI回复"""
key = self._make_cache_key(cookie_id, chat_id, message)
cached = self._cache.get(key)
if cached and time.time() < cached[1]:
logger.info(f"命中AI回复缓存 (账号: {cookie_id})")
return cached[0]
if cached:
del self._cache[key]
return None
def _set_cached_reply(self, cookie_id: str, chat_id: str, message: str, reply: str):
"""设置AI回复缓存"""
key = self._make_cache_key(cookie_id, chat_id, message)
self._cache[key] = (reply, time.time() + self._cache_ttl)
def _get_cached_intent(self, cookie_id: str, message: str) -> Optional[str]:
"""获取缓存的意图"""
key = self._make_cache_key(cookie_id, "", message)
cached = self._intent_cache.get(key)
if cached and time.time() < cached[1]:
return cached[0]
if cached:
del self._intent_cache[key]
return None
def _set_cached_intent(self, cookie_id: str, message: str, intent: str):
"""设置意图缓存"""
key = self._make_cache_key(cookie_id, "", message)
self._intent_cache[key] = (intent, time.time() + self._cache_ttl)
def _init_default_prompts(self):
"""初始化默认提示词"""
self.default_prompts = {
'classify': '''你是一个意图分类专家,需要判断用户消息的意图类型。
请根据用户消息内容,返回以下意图之一:
- price: 价格相关(议价、优惠、降价等)
- tech: 技术相关(产品参数、使用方法、故障等)
- refund: 退款相关(申请退款、退货、投诉等)
- default: 其他一般咨询
只返回意图类型,不要其他内容。''',
'price': '''你是一位经验丰富的销售专家,擅长议价。
语言要求:简短直接,每句≤10字,总字数≤40字。
议价策略:
1. 根据议价次数递减优惠:第1次小幅优惠,第2次中等优惠,第3次最大优惠
2. 接近最大议价轮数时要坚持底线,强调商品价值
3. 优惠不能超过设定的最大百分比和金额
4. 语气要友好但坚定,突出商品优势
注意:结合商品信息、对话历史和议价设置,给出合适的回复。''',
'tech': '''你是一位技术专家,专业解答产品相关问题。
语言要求:简短专业,每句≤10字,总字数≤40字。
回答重点:产品功能、使用方法、注意事项。
注意:基于商品信息回答,避免过度承诺。''',
'refund': '''你是一位售后客服,处理用户的退款退货请求。
语言要求:礼貌耐心,语气温和,每句≤15字,总字数≤60字。
注意:请严格遵循【退款策略】指示。''',
'default': '''你是一位资深电商卖家,提供优质客服。
语言要求:简短友好,每句≤10字,总字数≤40字。
回答重点:商品介绍、物流、售后等常见问题。
注意:结合商品信息,给出实用建议。'''
}
def get_client(self, cookie_id: str) -> Optional[OpenAI]:
"""获取指定账号的OpenAI客户端"""
if cookie_id not in self.clients:
settings = db_manager.get_ai_reply_settings(cookie_id)
if not settings['ai_enabled']:
logger.warning(f"AI回复未启用 (账号: {cookie_id})")
return None
if not settings['api_key']:
logger.warning(f"AI回复的 API Key 未配置 (账号: {cookie_id})")
return None
try:
self.clients[cookie_id] = OpenAI(
api_key=settings['api_key'],
base_url=settings['base_url'],
timeout=60
)
logger.info(f"为账号 {cookie_id} 创建OpenAI客户端")
except Exception as e:
logger.error(f"创建OpenAI客户端失败 {cookie_id}: {e}")
return None
return self.clients[cookie_id]
def is_ai_enabled(self, cookie_id: str) -> bool:
"""检查指定账号是否启用AI回复"""
settings = db_manager.get_ai_reply_settings(cookie_id)
return settings['ai_enabled']
def detect_intent(self, message: str, cookie_id: str) -> str:
"""检测用户消息意图(带缓存)"""
cached = self._get_cached_intent(cookie_id, message)
if cached:
return cached
client = self.get_client(cookie_id)
if not client:
return 'default'
try:
settings = db_manager.get_ai_reply_settings(cookie_id)
custom_prompts = json.loads(settings['custom_prompts']) if settings['custom_prompts'] else {}
classify_prompt = custom_prompts.get('classify', self.default_prompts['classify'])
response = client.chat.completions.create(
model=settings['model_name'],
messages=[
{"role": "system", "content": classify_prompt},
{"role": "user", "content": message}
],
max_tokens=100,
temperature=0.1
)
raw_content = response.choices[0].message.content
intent = raw_content.strip().lower() if raw_content else 'default'
if intent not in ['price', 'tech', 'refund', 'default']:
intent = 'default'
self._set_cached_intent(cookie_id, message, intent)
return intent
except Exception as e:
logger.error(f"意图检测失败 {cookie_id}: {e}")
return 'default'
def generate_reply(self, message: str, item_info: dict, chat_id: str,
cookie_id: str, user_id: str, item_id: str) -> Optional[str]:
"""生成AI回复(带缓存,带重试)"""
if not self.is_ai_enabled(cookie_id):
return None
# 检查缓存
cached_reply = self._get_cached_reply(cookie_id, chat_id, message)
if cached_reply:
return cached_reply
client = self.get_client(cookie_id)
if not client:
logger.warning(f"生成AI回复失败: 无法获取 OpenAI 客户端 (账号: {cookie_id})")
return None
# 获取设置(重试参数也放在 settings 里)
try:
settings = db_manager.get_ai_reply_settings(cookie_id)
except Exception as e:
logger.error(f"获取AI回复设置失败 {cookie_id}: {e}")
return None
# 读取重试配置
max_retry = 3
retry_interval = 5
# 检测意图
intent = self.detect_intent(message, cookie_id)
logger.info(f"检测到意图: {intent} (账号: {cookie_id})")
# 获取对话历史和议价信息(在重试循环外,只做一次)
context = self.get_conversation_context(chat_id, cookie_id)
bargain_count = self.get_bargain_count(chat_id, cookie_id)
refund_policy = settings.get('refund_policy', 'allow')
# 检查议价轮数限制
if intent == "price":
max_bargain_rounds = settings.get('max_bargain_rounds', 3)
if bargain_count >= max_bargain_rounds:
logger.info(f"议价次数已达上限 ({bargain_count}/{max_bargain_rounds}),拒绝继续议价")
refuse_reply = "抱歉,这个价格已经是最优惠的了,不能再便宜了哦!"
self.save_conversation(chat_id, cookie_id, user_id, item_id, "user", message, intent)
self.save_conversation(chat_id, cookie_id, user_id, item_id, "assistant", refuse_reply, intent)
return refuse_reply
# 构建提示词相关(在重试循环外,只做一次)
custom_prompts = json.loads(settings['custom_prompts']) if settings.get('custom_prompts') else {}
system_prompt = custom_prompts.get(intent, self.default_prompts[intent])
item_desc = f"商品标题: {item_info.get('title', '未知')}\n"
item_desc += f"商品价格: {item_info.get('price', '未知')}元\n"
item_desc += f"商品描述: {item_info.get('desc', '无')}"
context_str = "\n".join([f"{msg['role']}: {msg['content']}" for msg in context[-10:]])
max_bargain_rounds = settings.get('max_bargain_rounds', 3)
max_discount_percent = settings.get('max_discount_percent', 10)
max_discount_amount = settings.get('max_discount_amount', 100)
refund_instruction = "" if intent != "refund" else (
"【退款策略】该商品支持退款,请引导用户通过平台申请退款,告知流程和预计时间。"
if refund_policy == "allow"
else "【退款策略】该商品不支持退款,请礼貌说明原因(虚拟商品/已发货/特殊商品等),并表示歉意。"
)
user_prompt = f"""商品信息:
{item_desc}
对话历史:
{context_str}
议价设置:
- 当前议价次数:{bargain_count}
- 最大议价轮数:{max_bargain_rounds}
- 最大优惠百分比:{max_discount_percent}%
- 最大优惠金额:{max_discount_amount}元
{refund_instruction}
用户消息:{message}
请根据以上信息生成回复:"""
# === 带重试的 API 调用 ===
last_error = None
for attempt in range(max_retry):
try:
response = client.chat.completions.create(
model=settings['model_name'],
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": user_prompt}
],
max_tokens=2048,
temperature=0.7
)
raw_reply = response.choices[0].message.content
if not raw_reply:
raise ValueError("AI 返回内容为空")
reply = raw_reply.strip()
# 保存对话记录
self.save_conversation(chat_id, cookie_id, user_id, item_id, "user", message, intent)
self.save_conversation(chat_id, cookie_id, user_id, item_id, "assistant", reply, intent)
# 更新议价次数
if intent == "price":
self.increment_bargain_count(chat_id, cookie_id)
# 写入缓存
self._set_cached_reply(cookie_id, chat_id, message, reply)
logger.info(f"AI回复生成成功 (账号: {cookie_id}): {reply}")
return reply
except Exception as e:
last_error = e
if attempt < max_retry - 1:
logger.warning(f"AI回复生成失败 (第{attempt + 1}/{max_retry}次重试) {cookie_id}: {e}")
time.sleep(retry_interval)
else:
logger.error(f"AI回复生成失败 (已重试{max_retry}次) {cookie_id}: {e}")
return None
def get_conversation_context(self, chat_id: str, cookie_id: str, limit: int = 20) -> List[Dict]:
"""获取对话上下文"""
try:
with db_manager.lock:
cursor = db_manager.conn.cursor()
cursor.execute('''
SELECT role, content FROM ai_conversations
WHERE chat_id = ? AND cookie_id = ?
ORDER BY created_at DESC LIMIT ?
''', (chat_id, cookie_id, limit))
results = cursor.fetchall()
# 反转顺序,使其按时间正序
context = [{"role": row[0], "content": row[1]} for row in reversed(results)]
return context
except Exception as e:
logger.error(f"获取对话上下文失败: {e}")
return []
def save_conversation(self, chat_id: str, cookie_id: str, user_id: str,
item_id: str, role: str, content: str, intent: str = None):
"""保存对话记录"""
try:
with db_manager.lock:
cursor = db_manager.conn.cursor()
cursor.execute('''
INSERT INTO ai_conversations
(cookie_id, chat_id, user_id, item_id, role, content, intent)
VALUES (?, ?, ?, ?, ?, ?, ?)
''', (cookie_id, chat_id, user_id, item_id, role, content, intent))
db_manager.conn.commit()
except Exception as e:
logger.error(f"保存对话记录失败: {e}")
def get_bargain_count(self, chat_id: str, cookie_id: str) -> int:
"""获取议价次数"""
try:
with db_manager.lock:
cursor = db_manager.conn.cursor()
cursor.execute('''
SELECT COUNT(*) FROM ai_conversations
WHERE chat_id = ? AND cookie_id = ? AND intent = 'price' AND role = 'user'
''', (chat_id, cookie_id))
result = cursor.fetchone()
return result[0] if result else 0
except Exception as e:
logger.error(f"获取议价次数失败: {e}")
return 0
def increment_bargain_count(self, chat_id: str, cookie_id: str):
"""增加议价次数(通过保存记录自动增加)"""
# 议价次数通过查询price意图的用户消息数量来计算,无需单独操作
pass
def clear_cache(self, cookie_id: str = None):
"""清理AI回复和意图缓存"""
if cookie_id:
prefix = hashlib.md5(f"{cookie_id}|".encode()).hexdigest()[:8]
self._cache = {k: v for k, v in self._cache.items() if not k.startswith(prefix)}
self._intent_cache = {k: v for k, v in self._intent_cache.items() if not k.startswith(prefix)}
logger.info(f"清理账号 {cookie_id} 的AI回复缓存")
else:
self._cache.clear()
self._intent_cache.clear()
logger.info("清理所有AI回复缓存")
def clear_client_cache(self, cookie_id: str = None):
"""清理客户端缓存"""
if cookie_id:
self.clients.pop(cookie_id, None)
logger.info(f"清理账号 {cookie_id} 的客户端缓存")
else:
self.clients.clear()
logger.info("清理所有客户端缓存")
# 全局AI回复引擎实例
ai_reply_engine = AIReplyEngine()