trade-assistant/backend/app/services/silent_pattern.py

from typing import Dict, Any, Optional, List
from datetime import datetime, timedelta
from sqlalchemy.ext.asyncio import AsyncSession
from sqlalchemy import select, func, and_
from app.models.customer import Customer, Conversation, Message
import logging

logger = logging.getLogger(__name__)


class SilentPatternService:
    def __init__(self, db: AsyncSession):
        self.db = db

    async def analyze_silent_risk(self, user_id: str) -> List[Dict[str, Any]]:
        cutoff_3d = datetime.utcnow() - timedelta(days=3)
        cutoff_7d = datetime.utcnow() - timedelta(days=7)

        result = await self.db.execute(
            select(Customer).where(
                and_(
                    Customer.user_id == user_id,
                    Customer.status.in_(["lead", "negotiating"]),
                )
            )
        )
        customers = result.scalars().all()

        risk_scores = []
        for c in customers:
            score, reasons = await self._calculate_risk_score(c, cutoff_3d, cutoff_7d)
            if score > 0:
                risk_scores.append({
                    "customer_id": str(c.id),
                    "name": c.name,
                    "company": c.company,
                    "country": c.country,
                    "status": c.status,
                    "estimated_value": c.estimated_value,
                    "last_contact_at": c.last_contact_at.isoformat() if c.last_contact_at else None,
                    "silence_days": (datetime.utcnow() - c.last_contact_at).days if c.last_contact_at else 0,
                    "risk_score": score,
                    "risk_level": self._risk_level(score),
                    "reasons": reasons,
                })

        risk_scores.sort(key=lambda x: x["risk_score"], reverse=True)
        return risk_scores

    async def _calculate_risk_score(
        self, customer: Customer, cutoff_3d: datetime, cutoff_7d: datetime
    ) -> tuple:
        score = 0
        reasons = []

        if not customer.last_contact_at:
            return (0, [])

        silence_days = (datetime.utcnow() - customer.last_contact_at).days

        if silence_days >= 7:
            score += 40
            reasons.append(f"沉默超过7天")
        elif silence_days >= 3:
            score += 20
            reasons.append(f"沉默超过3天")

        conv_query = await self.db.execute(
            select(Conversation).where(
                and_(
                    Conversation.customer_id == customer.id,
                    Conversation.user_id == customer.user_id,
                )
            ).order_by(Conversation.created_at.desc()).limit(1)
        )
        conv = conv_query.scalar_one_or_none()
        if not conv:
            return (score, reasons)

        msg_count_query = await self.db.execute(
            select(func.count(Message.id)).where(
                and_(
                    Message.conversation_id == conv.id,
                    Message.direction == "inbound",
                )
            )
        )
        inbound_count = msg_count_query.scalar() or 0

        if inbound_count >= 5 and silence_days >= 3:
            score += 20
            reasons.append(f"前期沟通频繁({inbound_count}条)后突然沉默")

        if customer.status == "lead" and silence_days >= 3:
            score += 15
            reasons.append("潜在客户阶段需及时跟进")

        if customer.status == "negotiating" and silence_days >= 2:
            score += 25
            reasons.append("谈判阶段客户需保持热度")

        recent_query = await self.db.execute(
            select(Message).where(
                and_(
                    Message.conversation_id == conv.id,
                    Message.created_at >= cutoff_7d,
                )
            ).order_by(Message.created_at.desc()).limit(3)
        )
        recent_msgs = recent_query.scalars().all()

        if recent_msgs:
            last_inbound = None
            for m in recent_msgs:
                if m.direction == "inbound":
                    last_inbound = m
                    break
            if last_inbound and silence_days >= 1:
                content_lower = last_inbound.content.lower()
                closing_signals = ["i'll think", "let me check", "too expensive", "high price", "not now", "maybe later", "considering"]
                for signal in closing_signals:
                    if signal in content_lower:
                        score += 15
                        reasons.append(f"客户回复含消极信号: \"{signal}\"")
                        break

        return (min(score, 100), reasons)

    def _risk_level(self, score: int) -> str:
        if score >= 70:
            return "high"
        elif score >= 40:
            return "medium"
        elif score >= 20:
            return "low"
        return "minimal"

    async def get_suggestions(self, user_id: str, customer_id: str) -> List[str]:
        score_result = await self.analyze_silent_risk(user_id)
        customer_scores = [s for s in score_result if s["customer_id"] == customer_id]
        if not customer_scores:
            return []

        score = customer_scores[0]
        suggestions = []
        silence_days = score["silence_days"]

        if silence_days >= 7:
            suggestions.extend([
                f"客户{score['name']}已沉默{silence_days}天，建议发送产品更新或行业资讯重新激活",
                "考虑提供限时优惠或样品折扣打动客户",
            ])
        elif silence_days >= 3:
            suggestions.extend([
                f"客户{score['name']}沉默{silence_days}天，建议发送跟进消息询问是否有进一步需求",
                "可分享相关案例或成功故事保持客户兴趣",
            ])

        if "negotiating" in score.get("status", ""):
            suggestions.append("谈判阶段客户，建议主动提供更多产品细节或定制方案")

        if "消极信号" in str(score.get("reasons", [])):
            suggestions.append("客户曾表达价格顾虑，建议重新审视报价或提供增值服务")

        if not suggestions:
            suggestions.append("客户状态良好，建议保持定期跟进节奏")

        return suggestions