互联网医院AI问诊系统架构设计：从智能分诊到在线诊疗的完整链路

用户端(H5/小程序/App)
        ↓
API Gateway
        ↓
业务中台
 ├─ 用户中心
 ├─ 订单中心
 ├─ 医生排班服务
 ├─ 电子病历服务
 ├─ 处方服务
 ├─ 支付服务
 └─ AI问诊服务（核心）
        ↓
AI能力层
 ├─ NLP模型
 ├─ 症状知识图谱
 ├─ 分诊规则引擎
 └─ 大模型推理服务

用户描述症状 → NLP解析 → 症状标签提取 → 匹配知识图谱 → 推荐科室

from fastapi import FastAPI
import jieba

app = FastAPI()

SYMPTOMS = ["发烧", "咳嗽", "头痛", "腹泻", "胸闷"]

def extract_symptoms(text: str):
    words = jieba.lcut(text)
    return [w for w in words if w in SYMPTOMS]

@app.post("/triage")
def triage(data: dict):
    symptoms = extract_symptoms(data["content"])

    if "发烧" in symptoms or "咳嗽" in symptoms:
        dept = "呼吸内科"
    elif "腹泻" in symptoms:
        dept = "消化内科"
    else:
        dept = "全科"

    return {"symptoms": symptoms, "department": dept}

输入：我发烧咳嗽三天
输出：呼吸内科

INIT → 主诉 → 既往史 → 用药史 → 过敏史 → 总结 → 转医生

public class InquiryFlow {

    private Map<String, String> nextStep = Map.of(
        "INIT", "CHIEF",
        "CHIEF", "HISTORY",
        "HISTORY", "DRUG",
        "DRUG", "ALLERGY",
        "ALLERGY", "SUMMARY"
    );

    public String next(String current){
        return nextStep.getOrDefault(current, "END");
    }
}

请描述您的主要症状？
是否有既往疾病？
近期是否服药？

症状 → 疾病 → 科室 → 医生 → 推荐方案

function riskLevel(symptoms){
  if(symptoms.includes("胸痛") && symptoms.includes("呼吸困难")){
    return "high";
  }
  return "normal";
}

请立即线下就医或拨打急救电话

{
  "chief": "发烧三天",
  "history": "无慢性病",
  "drug": "未服药",
  "suggestDept": "呼吸内科"
}

def generate_record(data):
    return f"""
主诉：{data['chief']}
既往史：{data['history']}
用药史：{data['drug']}
建议科室：{data['dept']}
"""

import redis
r = redis.Redis()

def cache_answer(key, value):
    r.setex(key, 300, value)