File size: 6,587 Bytes
788074d
 
 
 
fc636ce
 
788074d
 
 
 
fc636ce
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
788074d
fc636ce
 
 
 
 
 
 
 
 
 
 
 
788074d
fc636ce
 
 
 
 
788074d
fc636ce
 
 
 
 
 
 
 
788074d
fc636ce
 
 
 
 
 
 
788074d
fc636ce
 
 
 
 
 
788074d
fc636ce
 
 
 
 
 
 
 
 
788074d
fc636ce
 
 
 
 
 
 
 
 
 
 
 
788074d
 
 
fc636ce
 
 
 
 
788074d
fc636ce
788074d
 
 
 
 
 
 
 
 
fc636ce
 
788074d
 
 
fc636ce
 
788074d
fc636ce
788074d
fc636ce
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
788074d
 
 
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
172
173
174
175
176
177
178
179
180
181
# app.py
import streamlit as st
import sqlite3
import plotly.express as px
from uuid import uuid4
from langgraph.graph import START, StateGraph, END
from langchain_groq import ChatGroq
from langgraph.checkpoint.sqlite import SqliteSaver
from langchain_community.tools.tavily_search import TavilySearchResults
from langchain_core.messages import HumanMessage, SystemMessage, ToolMessage
from typing import TypedDict, Annotated, List

# Configuration
class MedicalConfig:
    SYSTEM_PROMPT = """You are an AI clinical assistant. Follow these rules:
1. Analyze patient data using latest medical guidelines
2. Always check for drug interactions
3. Suggest tests only when necessary
4. Use structured actions:
   - order_lab_test: {test_name, reason}
   - prescribe_medication: {name, dosage, frequency}
5. Research using tavily_search when uncertain"""
    
    RED_FLAGS = {
        'symptoms': ['chest pain', 'shortness of breath', 'severe headache'],
        'vitals': {'temp': (38.5, '°C'), 'hr': (120, 'bpm'), 'bp': ('180/120', 'mmHg')}
    }

# State Management
class AgentState(TypedDict):
    messages: Annotated[List[dict], lambda l, r: l + r]
    patient_data: dict
    approvals: Annotated[List[dict], lambda l, r: l + r]

class MedicalAgent:
    def __init__(self):
        self.model = ChatGroq(temperature=0.1, model="Llama-3.3-70b-Specdec")
        self.tools = {
            "medical_actions": {
                "order_lab_test": self.order_lab_test,
                "prescribe_medication": self.prescribe_medication
            },
            "research": TavilySearchResults(max_results=3)
        }
        self.checkpointer = SqliteSaver(sqlite3.connect("medical.db"))
        self._build_graph()

    def _build_graph(self):
        graph = StateGraph(AgentState)
        graph.add_node("analyze", self.analyze_patient)
        graph.add_node("execute", self.execute_actions)
        graph.add_node("safety_check", self.safety_checks)
        
        graph.add_edge(START, "analyze")
        graph.add_conditional_edges(
            "analyze",
            self.route_actions,
            {"medical": "safety_check", "research": "execute"}
        )
        graph.add_edge("safety_check", "execute")
        graph.add_edge("execute", "analyze")
        
        graph.set_entry_point("analyze")
        self.graph = graph.compile(checkpointer=self.checkpointer)

    def analyze_patient(self, state: AgentState):
        response = self.model.bind_tools(list(self.tools['medical_actions'].keys())).invoke([
            SystemMessage(content=MedicalConfig.SYSTEM_PROMPT),
            HumanMessage(content=f"Patient Data: {state['patient_data']}")
        ])
        return {"messages": [response]}
    
    def safety_checks(self, state: AgentState):
        current_action = state['messages'][-1].tool_calls[0]
        risk = self.check_contraindications(state['patient_data'], current_action)
        return {"approvals": [{"action": current_action, "risk": risk}]}

    def execute_actions(self, state: AgentState):
        results = []
        for action in state.get('approvals', []):
            if action['risk']:
                results.append(f"Blocked dangerous action: {action['action']['name']}")
            else:
                tool = self.tools['medical_actions'][action['action']['name']]
                results.append(tool(action['action']['args']))
        return {"messages": [{"content": "\n".join(results)}]}

    def check_contraindications(self, patient_data, action):
        # Implement actual medical safety checks
        if 'prescribe' in action['name']:
            return any(drug in patient_data['medications'] 
                      for drug in ['warfarin', 'insulin'])
        return False

    def order_lab_test(self, params):
        return f"Lab ordered: {params['test_name']} ({params['reason']})"

    def prescribe_medication(self, params):
        return f"Prescribed: {params['name']} {params['dosage']} {params['frequency']}"

# Streamlit UI
def main():
    st.set_page_config(page_title="AI Clinic", layout="wide")
    
    # Initialize agent
    if 'agent' not in st.session_state:
        st.session_state.agent = MedicalAgent()
    
    # Patient intake
    with st.sidebar:
        st.header("Patient Intake")
        patient_data = {
            "symptoms": st.multiselect("Symptoms", ["Fever", "Cough", "Chest Pain"]),
            "history": {
                "conditions": st.text_input("Medical History"),
                "medications": st.text_input("Current Medications")
            },
            "vitals": {
                "temp": st.number_input("Temp (°C)", 35.0, 42.0, 37.0),
                "bp": st.text_input("BP (mmHg)", "120/80")
            }
        }
    
    # Main interface
    st.title("AI-Powered Clinical Support System")
    
    col1, col2 = st.columns([2, 1])
    
    with col1:
        st.subheader("Patient Overview")
        self.render_patient_dashboard(patient_data)
        
        if st.button("Start Analysis"):
            thread_id = f"case_{uuid4().hex[:8]}"
            with st.spinner("Analyzing..."):
                st.session_state.agent.graph.invoke(
                    {"patient_data": patient_data},
                    {"configurable": {"thread_id": thread_id}}
                )
    
    with col2:
        st.subheader("Action Center")
        self.render_approval_interface()

def render_patient_dashboard(self, data):
    tab1, tab2 = st.tabs(["Vitals", "Timeline"])
    
    with tab1:
        fig = px.line(
            x=["Temperature", "Blood Pressure"],
            y=[data['vitals']['temp'], 120],
            title="Vital Signs"
        )
        st.plotly_chart(fig)
    
    with tab2:
        st.vega_lite_chart({
            "mark": {"type": "line", "point": True},
            "encoding": {
                "x": {"field": "hour", "type": "ordinal"},
                "y": {"field": "temp", "type": "quantitative"}
            },
            "data": {"values": [{"hour": i, "temp": 36.5 + i/10} for i in range(24)]}
        })

def render_approval_interface(self):
    if 'current_action' in st.session_state:
        action = st.session_state.current_action
        st.warning("Action Requires Approval")
        st.write(f"**Type:** {action['name'].replace('_', ' ').title()}")
        st.json(action['args'])
        
        if st.button("Approve"):
            self.process_approval(True)
        if st.button("Reject"):
            self.process_approval(False)
    else:
        st.info("No pending actions")

if __name__ == "__main__":
    main()