File size: 5,749 Bytes
c7d192d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
# agent.py
import os
from typing import Dict, List, Any, Optional
from langgraph.graph import Graph
from langchain.schema import BaseMessage, HumanMessage, AIMessage
from langchain_community.llms import HuggingFacePipeline
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
import torch
from sentence_transformers import SentenceTransformer
import faiss
import numpy as np
import json


class MedicalAgent:
    def __init__(self):
        self.llm = self._load_huatuogpt()
        self.rag = MedicalRAG()
        self.conversation_history = []
        self.question_count = 0
        self.max_questions = 3
        self.max_words_per_question = 5

    def _load_huatuogpt(self):
        """Load HuatuoGPT model from HuggingFace"""
        model_name = "HuatuoGPT/HuatuoGPT-7B"
        tokenizer = AutoTokenizer.from_pretrained(model_name)
        model = AutoModelForCausalLM.from_pretrained(
            model_name,
            torch_dtype=torch.float16,
            device_map="auto"
        )

        pipe = pipeline(
            "text-generation",
            model=model,
            tokenizer=tokenizer,
            max_new_tokens=100,
            temperature=0.7,
            do_sample=True
        )

        return HuggingFacePipeline(pipeline=pipe)

    def process_patient_input(self, patient_text: str) -> Dict[str, Any]:
        """Process patient input and generate response"""
        self.conversation_history.append(f"Patient: {patient_text}")

        # Check if we've reached question limit
        if self.question_count >= self.max_questions:
            return self._generate_summary()

        # Analyze symptoms and decide next action
        analysis = self._analyze_symptoms(patient_text)

        if analysis["needs_follow_up"]:
            follow_up_question = self._generate_follow_up_question(analysis)
            self.question_count += 1
            return {
                "type": "question",
                "content": follow_up_question,
                "question_count": self.question_count
            }
        else:
            return self._generate_summary()

    def _analyze_symptoms(self, patient_text: str) -> Dict[str, Any]:
        """Analyze symptoms using RAG and LLM"""
        # Search medical knowledge base
        relevant_info = self.rag.search(patient_text, k=3)

        prompt = f"""

        Patient complaint: {patient_text}

        Relevant medical information: {relevant_info}



        Analyze the symptoms and determine:

        1. If we need follow-up questions (True/False)

        2. What key information is missing

        3. Suggested follow-up questions (max 5 words each)



        Respond in JSON format:

        {{

            "needs_follow_up": boolean,

            "missing_info": list,

            "possible_questions": list

        }}

        """

        response = self.llm(prompt)
        try:
            analysis = json.loads(response)
        except:
            analysis = {
                "needs_follow_up": True,
                "missing_info": ["symptom details"],
                "possible_questions": ["How long have headache?", "Any other symptoms?"]
            }

        return analysis

    def _generate_follow_up_question(self, analysis: Dict) -> str:
        """Generate concise follow-up question"""
        possible_questions = analysis.get("possible_questions", [])
        if possible_questions:
            question = possible_questions[0]
            # Ensure question is within word limit
            words = question.split()[:self.max_words_per_question]
            return " ".join(words)
        else:
            return "Any other symptoms?"

    def _generate_summary(self) -> Dict[str, Any]:
        """Generate summary for doctor"""
        conversation_text = "\n".join(self.conversation_history)

        prompt = f"""

        Patient conversation:

        {conversation_text}



        Generate a concise medical summary for the doctor including:

        - Main symptoms

        - Key findings

        - Suggested preliminary diagnosis

        - Recommended tests if any



        Keep it under 150 words.

        """

        summary = self.llm(prompt)
        return {
            "type": "summary",
            "content": summary,
            "question_count": self.question_count
        }

    def process_doctor_question(self, doctor_text: str) -> str:
        """Process doctor's follow-up questions"""
        prompt = f"""

        Doctor's question: {doctor_text}



        Rephrase this question to be clear and simple for the patient.

        Keep it under 5 words and make it easy to understand.

        """

        simplified_question = self.llm(prompt)
        return simplified_question


class MedicalRAG:
    def __init__(self):
        self.encoder = SentenceTransformer('all-MiniLM-L6-v2')
        self.index = faiss.IndexFlatL2(384)
        self.knowledge_base = []

    def add_medical_knowledge(self, documents: List[str]):
        """Add medical knowledge documents to RAG"""
        self.knowledge_base.extend(documents)
        embeddings = self.encoder.encode(documents)
        self.index.add(np.array(embeddings))

    def search(self, query: str, k: int = 3) -> List[str]:
        """Search medical knowledge base"""
        query_embedding = self.encoder.encode([query])
        distances, indices = self.index.search(query_embedding, k)

        results = []
        for idx in indices[0]:
            if idx < len(self.knowledge_base):
                results.append(self.knowledge_base[idx])

        return results