Update app.py
Browse files
app.py
CHANGED
@@ -1,3 +1,4 @@
|
|
|
|
1 |
import os
|
2 |
import sys
|
3 |
import json
|
@@ -9,9 +10,11 @@ import base64
|
|
9 |
from datetime import datetime
|
10 |
from typing import List, Dict, Optional, Tuple
|
11 |
from enum import Enum
|
12 |
-
from fastapi import FastAPI, HTTPException, UploadFile, File, Query, Form
|
13 |
from fastapi.responses import StreamingResponse, JSONResponse
|
14 |
from fastapi.middleware.cors import CORSMiddleware
|
|
|
|
|
15 |
from pydantic import BaseModel
|
16 |
import asyncio
|
17 |
from bson import ObjectId
|
@@ -20,17 +23,19 @@ from gtts import gTTS
|
|
20 |
from pydub import AudioSegment
|
21 |
import PyPDF2
|
22 |
import mimetypes
|
|
|
|
|
23 |
from txagent.txagent import TxAgent
|
24 |
from db.mongo import get_mongo_client
|
25 |
-
|
26 |
-
from docx import Document
|
27 |
# Logging
|
28 |
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
|
29 |
logger = logging.getLogger("TxAgentAPI")
|
30 |
|
31 |
# App
|
32 |
-
app = FastAPI(title="TxAgent API", version="2.6.0")
|
33 |
|
|
|
34 |
app.add_middleware(
|
35 |
CORSMiddleware,
|
36 |
allow_origins=["*"],
|
@@ -39,6 +44,13 @@ app.add_middleware(
|
|
39 |
allow_headers=["*"]
|
40 |
)
|
41 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
42 |
# Pydantic Models
|
43 |
class ChatRequest(BaseModel):
|
44 |
message: str
|
@@ -55,7 +67,7 @@ class VoiceOutputRequest(BaseModel):
|
|
55 |
text: str
|
56 |
language: str = "en"
|
57 |
slow: bool = False
|
58 |
-
return_format: str = "mp3"
|
59 |
|
60 |
# Enums
|
61 |
class RiskLevel(str, Enum):
|
@@ -71,7 +83,26 @@ patients_collection = None
|
|
71 |
analysis_collection = None
|
72 |
alerts_collection = None
|
73 |
|
74 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
75 |
def clean_text_response(text: str) -> str:
|
76 |
text = re.sub(r'\n\s*\n', '\n\n', text)
|
77 |
text = re.sub(r'[ ]+', ' ', text)
|
@@ -87,7 +118,6 @@ def extract_section(text: str, heading: str) -> str:
|
|
87 |
return ""
|
88 |
|
89 |
def structure_medical_response(text: str) -> Dict:
|
90 |
-
"""Improved version that handles both markdown and plain text formats"""
|
91 |
def extract_improved(text: str, heading: str) -> str:
|
92 |
patterns = [
|
93 |
rf"{re.escape(heading)}:\s*\n(.*?)(?=\n\s*\n|\Z)",
|
@@ -95,7 +125,6 @@ def structure_medical_response(text: str) -> Dict:
|
|
95 |
rf"{re.escape(heading)}[\s\-]+(.*?)(?=\n\s*\n|\Z)",
|
96 |
rf"\n{re.escape(heading)}\s*\n(.*?)(?=\n\s*\n|\Z)"
|
97 |
]
|
98 |
-
|
99 |
for pattern in patterns:
|
100 |
match = re.search(pattern, text, re.DOTALL | re.IGNORECASE)
|
101 |
if match:
|
@@ -103,9 +132,8 @@ def structure_medical_response(text: str) -> Dict:
|
|
103 |
content = re.sub(r'^\s*[\-\*]\s*', '', content, flags=re.MULTILINE)
|
104 |
return content
|
105 |
return ""
|
106 |
-
|
107 |
-
text = text.replace('**', '').replace('__', '')
|
108 |
|
|
|
109 |
return {
|
110 |
"summary": extract_improved(text, "Summary of Patient's Medical History") or
|
111 |
extract_improved(text, "Summarize the patient's medical history"),
|
@@ -118,15 +146,12 @@ def structure_medical_response(text: str) -> Dict:
|
|
118 |
}
|
119 |
|
120 |
def detect_suicide_risk(text: str) -> Tuple[RiskLevel, float, List[str]]:
|
121 |
-
"""Analyze text for suicide risk factors and return assessment"""
|
122 |
suicide_keywords = [
|
123 |
'suicide', 'suicidal', 'kill myself', 'end my life',
|
124 |
'want to die', 'self-harm', 'self harm', 'hopeless',
|
125 |
'no reason to live', 'plan to die'
|
126 |
]
|
127 |
-
|
128 |
explicit_mentions = [kw for kw in suicide_keywords if kw in text.lower()]
|
129 |
-
|
130 |
if not explicit_mentions:
|
131 |
return RiskLevel.NONE, 0.0, []
|
132 |
|
@@ -145,7 +170,6 @@ def detect_suicide_risk(text: str) -> Tuple[RiskLevel, float, List[str]]:
|
|
145 |
temperature=0.2,
|
146 |
max_new_tokens=256
|
147 |
)
|
148 |
-
|
149 |
json_match = re.search(r'\{.*\}', response, re.DOTALL)
|
150 |
if json_match:
|
151 |
assessment = json.loads(json_match.group())
|
@@ -165,7 +189,6 @@ def detect_suicide_risk(text: str) -> Tuple[RiskLevel, float, List[str]]:
|
|
165 |
return RiskLevel.LOW, risk_score, explicit_mentions
|
166 |
|
167 |
async def create_alert(patient_id: str, risk_data: dict):
|
168 |
-
"""Create an alert document in the database"""
|
169 |
alert_doc = {
|
170 |
"patient_id": patient_id,
|
171 |
"type": "suicide_risk",
|
@@ -185,16 +208,13 @@ def serialize_patient(patient: dict) -> dict:
|
|
185 |
return patient_copy
|
186 |
|
187 |
def compute_patient_data_hash(data: dict) -> str:
|
188 |
-
"""Compute SHA-256 hash of patient data or report."""
|
189 |
serialized = json.dumps(data, sort_keys=True)
|
190 |
return hashlib.sha256(serialized.encode()).hexdigest()
|
191 |
|
192 |
def compute_file_content_hash(file_content: bytes) -> str:
|
193 |
-
"""Compute SHA-256 hash of file content."""
|
194 |
return hashlib.sha256(file_content).hexdigest()
|
195 |
|
196 |
def extract_text_from_pdf(pdf_data: bytes) -> str:
|
197 |
-
"""Extract text from a PDF file."""
|
198 |
try:
|
199 |
pdf_reader = PyPDF2.PdfReader(io.BytesIO(pdf_data))
|
200 |
text = ""
|
@@ -206,85 +226,70 @@ def extract_text_from_pdf(pdf_data: bytes) -> str:
|
|
206 |
raise HTTPException(status_code=400, detail="Failed to extract text from PDF")
|
207 |
|
208 |
async def analyze_patient_report(patient_id: Optional[str], report_content: str, file_type: str, file_content: bytes):
|
209 |
-
|
210 |
-
|
211 |
-
|
212 |
-
|
213 |
-
|
214 |
-
|
215 |
-
|
216 |
-
|
217 |
-
|
218 |
-
|
219 |
-
|
220 |
-
|
221 |
-
|
222 |
-
|
223 |
-
|
224 |
-
|
225 |
-
|
226 |
-
|
227 |
-
"2. Identify risks or red flags (including mental health and suicide risk).\n"
|
228 |
-
"3. Highlight missed diagnoses or treatments.\n"
|
229 |
-
"4. Suggest next clinical steps.\n"
|
230 |
-
f"\nPatient Report ({file_type}):\n{'-'*40}\n{report_content[:10000]}"
|
231 |
-
)
|
232 |
-
|
233 |
-
# Perform analysis
|
234 |
-
raw_response = agent.chat(
|
235 |
-
message=prompt,
|
236 |
-
history=[],
|
237 |
-
temperature=0.7,
|
238 |
-
max_new_tokens=1024
|
239 |
-
)
|
240 |
-
structured_response = structure_medical_response(raw_response)
|
241 |
|
242 |
-
|
243 |
-
|
244 |
-
|
245 |
-
|
246 |
-
|
247 |
-
|
248 |
-
|
249 |
|
250 |
-
|
251 |
-
|
252 |
-
|
253 |
-
|
254 |
-
|
255 |
-
|
256 |
-
"suicide_risk": suicide_risk,
|
257 |
-
"raw": raw_response,
|
258 |
-
"report_hash": report_hash,
|
259 |
-
"file_type": file_type
|
260 |
-
}
|
261 |
|
262 |
-
|
263 |
-
|
264 |
-
|
265 |
-
|
266 |
-
|
|
|
|
|
|
|
|
|
|
|
267 |
|
268 |
-
|
269 |
-
|
270 |
-
|
|
|
|
|
271 |
|
272 |
-
|
273 |
-
|
274 |
|
275 |
-
|
276 |
-
|
277 |
-
raise HTTPException(status_code=500, detail="Failed to analyze patient report")
|
278 |
|
279 |
async def analyze_all_patients():
|
280 |
-
"""Analyze all patients in the database."""
|
281 |
patients = await patients_collection.find({}).to_list(length=None)
|
282 |
for patient in patients:
|
283 |
await analyze_patient(patient)
|
284 |
await asyncio.sleep(0.1)
|
285 |
|
286 |
async def analyze_patient(patient: dict):
|
287 |
-
"""Analyze patient data (existing logic for patient records)."""
|
288 |
try:
|
289 |
serialized = serialize_patient(patient)
|
290 |
patient_id = serialized.get("fhir_id")
|
@@ -342,9 +347,7 @@ async def analyze_patient(patient: dict):
|
|
342 |
logger.error(f"Error analyzing patient: {e}")
|
343 |
|
344 |
def recognize_speech(audio_data: bytes, language: str = "en-US") -> str:
|
345 |
-
"""Convert speech to text using Google's speech recognition."""
|
346 |
recognizer = sr.Recognizer()
|
347 |
-
|
348 |
try:
|
349 |
with io.BytesIO(audio_data) as audio_file:
|
350 |
with sr.AudioFile(audio_file) as source:
|
@@ -362,7 +365,6 @@ def recognize_speech(audio_data: bytes, language: str = "en-US") -> str:
|
|
362 |
raise HTTPException(status_code=500, detail="Error processing speech")
|
363 |
|
364 |
def text_to_speech(text: str, language: str = "en", slow: bool = False) -> bytes:
|
365 |
-
"""Convert text to speech using gTTS and return as MP3 bytes."""
|
366 |
try:
|
367 |
tts = gTTS(text=text, lang=language, slow=slow)
|
368 |
mp3_fp = io.BytesIO()
|
@@ -394,6 +396,8 @@ async def startup_event():
|
|
394 |
logger.info("✅ TxAgent initialized")
|
395 |
|
396 |
db = get_mongo_client()["cps_db"]
|
|
|
|
|
397 |
patients_collection = db["patients"]
|
398 |
analysis_collection = db["patient_analysis_results"]
|
399 |
alerts_collection = db["clinical_alerts"]
|
@@ -401,8 +405,10 @@ async def startup_event():
|
|
401 |
|
402 |
asyncio.create_task(analyze_all_patients())
|
403 |
|
|
|
404 |
@app.get("/status")
|
405 |
-
async def status():
|
|
|
406 |
return {
|
407 |
"status": "running",
|
408 |
"timestamp": datetime.utcnow().isoformat(),
|
@@ -411,7 +417,11 @@ async def status():
|
|
411 |
}
|
412 |
|
413 |
@app.get("/patients/analysis-results")
|
414 |
-
async def get_patient_analysis_results(
|
|
|
|
|
|
|
|
|
415 |
try:
|
416 |
query = {}
|
417 |
if name:
|
@@ -438,7 +448,11 @@ async def get_patient_analysis_results(name: Optional[str] = Query(None)):
|
|
438 |
raise HTTPException(status_code=500, detail="Failed to retrieve analysis results")
|
439 |
|
440 |
@app.post("/chat-stream")
|
441 |
-
async def chat_stream_endpoint(
|
|
|
|
|
|
|
|
|
442 |
async def token_stream():
|
443 |
try:
|
444 |
conversation = [{"role": "system", "content": agent.chat_prompt}]
|
@@ -472,9 +486,10 @@ async def chat_stream_endpoint(request: ChatRequest):
|
|
472 |
@app.post("/voice/transcribe")
|
473 |
async def transcribe_voice(
|
474 |
audio: UploadFile = File(...),
|
475 |
-
language: str = Query("en-US", description="Language code for speech recognition")
|
|
|
476 |
):
|
477 |
-
"
|
478 |
try:
|
479 |
audio_data = await audio.read()
|
480 |
if not audio.filename.lower().endswith(('.wav', '.mp3', '.ogg', '.flac')):
|
@@ -490,8 +505,11 @@ async def transcribe_voice(
|
|
490 |
raise HTTPException(status_code=500, detail="Error processing voice input")
|
491 |
|
492 |
@app.post("/voice/synthesize")
|
493 |
-
async def synthesize_voice(
|
494 |
-
|
|
|
|
|
|
|
495 |
try:
|
496 |
audio_data = text_to_speech(request.text, request.language, request.slow)
|
497 |
|
@@ -515,9 +533,10 @@ async def voice_chat_endpoint(
|
|
515 |
audio: UploadFile = File(...),
|
516 |
language: str = Query("en-US", description="Language code for speech recognition"),
|
517 |
temperature: float = Query(0.7, ge=0.1, le=1.0),
|
518 |
-
max_new_tokens: int = Query(512, ge=50, le=1024)
|
|
|
519 |
):
|
520 |
-
"
|
521 |
try:
|
522 |
audio_data = await audio.read()
|
523 |
user_message = recognize_speech(audio_data, language)
|
@@ -548,18 +567,11 @@ async def analyze_clinical_report(
|
|
548 |
file: UploadFile = File(...),
|
549 |
patient_id: Optional[str] = Form(None),
|
550 |
temperature: float = Form(0.5),
|
551 |
-
max_new_tokens: int = Form(1024)
|
|
|
552 |
):
|
553 |
-
""
|
554 |
-
Analyze a clinical patient report from an uploaded file.
|
555 |
-
Parameters:
|
556 |
-
- file: Uploaded clinical report file (PDF, TXT, DOCX)
|
557 |
-
- patient_id: Optional patient ID to associate with this report
|
558 |
-
- temperature: Controls randomness of response (0.1-1.0)
|
559 |
-
- max_new_tokens: Maximum length of response
|
560 |
-
"""
|
561 |
try:
|
562 |
-
# Validate file type
|
563 |
content_type = file.content_type
|
564 |
allowed_types = [
|
565 |
'application/pdf',
|
@@ -573,10 +585,8 @@ async def analyze_clinical_report(
|
|
573 |
detail=f"Unsupported file type: {content_type}. Supported types: PDF, TXT, DOCX"
|
574 |
)
|
575 |
|
576 |
-
# Read file content
|
577 |
file_content = await file.read()
|
578 |
|
579 |
-
# Extract text from file
|
580 |
if content_type == 'application/pdf':
|
581 |
text = extract_text_from_pdf(file_content)
|
582 |
elif content_type == 'text/plain':
|
@@ -587,7 +597,6 @@ async def analyze_clinical_report(
|
|
587 |
else:
|
588 |
raise HTTPException(status_code=400, detail="Unsupported file type")
|
589 |
|
590 |
-
# Clean and validate text
|
591 |
text = clean_text_response(text)
|
592 |
if len(text.strip()) < 50:
|
593 |
raise HTTPException(
|
@@ -595,7 +604,6 @@ async def analyze_clinical_report(
|
|
595 |
detail="Extracted text is too short (minimum 50 characters required)"
|
596 |
)
|
597 |
|
598 |
-
# Analyze the report
|
599 |
analysis = await analyze_patient_report(
|
600 |
patient_id=patient_id,
|
601 |
report_content=text,
|
@@ -603,13 +611,11 @@ async def analyze_clinical_report(
|
|
603 |
file_content=file_content
|
604 |
)
|
605 |
|
606 |
-
# Manually convert ObjectId and timestamp if needed
|
607 |
if "_id" in analysis and isinstance(analysis["_id"], ObjectId):
|
608 |
analysis["_id"] = str(analysis["_id"])
|
609 |
if "timestamp" in analysis and isinstance(analysis["timestamp"], datetime):
|
610 |
analysis["timestamp"] = analysis["timestamp"].isoformat()
|
611 |
|
612 |
-
# Return response using jsonable_encoder
|
613 |
return JSONResponse(content=jsonable_encoder({
|
614 |
"status": "success",
|
615 |
"analysis": analysis,
|
@@ -627,7 +633,6 @@ async def analyze_clinical_report(
|
|
627 |
detail=f"Failed to analyze report: {str(e)}"
|
628 |
)
|
629 |
|
630 |
-
|
631 |
if __name__ == "__main__":
|
632 |
import uvicorn
|
633 |
uvicorn.run(app, host="0.0.0.0", port=8000)
|
|
|
1 |
+
# app.py (in TxAgent-API)
|
2 |
import os
|
3 |
import sys
|
4 |
import json
|
|
|
10 |
from datetime import datetime
|
11 |
from typing import List, Dict, Optional, Tuple
|
12 |
from enum import Enum
|
13 |
+
from fastapi import FastAPI, HTTPException, UploadFile, File, Query, Form, Depends
|
14 |
from fastapi.responses import StreamingResponse, JSONResponse
|
15 |
from fastapi.middleware.cors import CORSMiddleware
|
16 |
+
from fastapi.security import OAuth2PasswordBearer
|
17 |
+
from fastapi.encoders import jsonable_encoder
|
18 |
from pydantic import BaseModel
|
19 |
import asyncio
|
20 |
from bson import ObjectId
|
|
|
23 |
from pydub import AudioSegment
|
24 |
import PyPDF2
|
25 |
import mimetypes
|
26 |
+
from docx import Document
|
27 |
+
from jose import JWTError, jwt
|
28 |
from txagent.txagent import TxAgent
|
29 |
from db.mongo import get_mongo_client
|
30 |
+
|
|
|
31 |
# Logging
|
32 |
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
|
33 |
logger = logging.getLogger("TxAgentAPI")
|
34 |
|
35 |
# App
|
36 |
+
app = FastAPI(title="TxAgent API", version="2.6.0")
|
37 |
|
38 |
+
# CORS
|
39 |
app.add_middleware(
|
40 |
CORSMiddleware,
|
41 |
allow_origins=["*"],
|
|
|
44 |
allow_headers=["*"]
|
45 |
)
|
46 |
|
47 |
+
# JWT settings (must match CPS-API)
|
48 |
+
SECRET_KEY = os.getenv("SECRET_KEY", "your-secret-key") # Same as CPS-API
|
49 |
+
ALGORITHM = "HS256"
|
50 |
+
|
51 |
+
# OAuth2 scheme (point to CPS-API's login endpoint)
|
52 |
+
oauth2_scheme = OAuth2PasswordBearer(tokenUrl="https://rocketfarmstudios-cps-api.hf.space/auth/login")
|
53 |
+
|
54 |
# Pydantic Models
|
55 |
class ChatRequest(BaseModel):
|
56 |
message: str
|
|
|
67 |
text: str
|
68 |
language: str = "en"
|
69 |
slow: bool = False
|
70 |
+
return_format: str = "mp3"
|
71 |
|
72 |
# Enums
|
73 |
class RiskLevel(str, Enum):
|
|
|
83 |
analysis_collection = None
|
84 |
alerts_collection = None
|
85 |
|
86 |
+
# JWT validation
|
87 |
+
async def get_current_user(token: str = Depends(oauth2_scheme)):
|
88 |
+
credentials_exception = HTTPException(
|
89 |
+
status_code=401,
|
90 |
+
detail="Could not validate credentials",
|
91 |
+
headers={"WWW-Authenticate": "Bearer"},
|
92 |
+
)
|
93 |
+
try:
|
94 |
+
payload = jwt.decode(token, SECRET_KEY, algorithms=[ALGORITHM])
|
95 |
+
email: str = payload.get("sub")
|
96 |
+
if email is None:
|
97 |
+
raise credentials_exception
|
98 |
+
except JWTError:
|
99 |
+
raise credentials_exception
|
100 |
+
user = await users_collection.find_one({"email": email})
|
101 |
+
if user is None:
|
102 |
+
raise credentials_exception
|
103 |
+
return user
|
104 |
+
|
105 |
+
# Helper functions (unchanged from your original code)
|
106 |
def clean_text_response(text: str) -> str:
|
107 |
text = re.sub(r'\n\s*\n', '\n\n', text)
|
108 |
text = re.sub(r'[ ]+', ' ', text)
|
|
|
118 |
return ""
|
119 |
|
120 |
def structure_medical_response(text: str) -> Dict:
|
|
|
121 |
def extract_improved(text: str, heading: str) -> str:
|
122 |
patterns = [
|
123 |
rf"{re.escape(heading)}:\s*\n(.*?)(?=\n\s*\n|\Z)",
|
|
|
125 |
rf"{re.escape(heading)}[\s\-]+(.*?)(?=\n\s*\n|\Z)",
|
126 |
rf"\n{re.escape(heading)}\s*\n(.*?)(?=\n\s*\n|\Z)"
|
127 |
]
|
|
|
128 |
for pattern in patterns:
|
129 |
match = re.search(pattern, text, re.DOTALL | re.IGNORECASE)
|
130 |
if match:
|
|
|
132 |
content = re.sub(r'^\s*[\-\*]\s*', '', content, flags=re.MULTILINE)
|
133 |
return content
|
134 |
return ""
|
|
|
|
|
135 |
|
136 |
+
text = text.replace('**', '').replace('__', '')
|
137 |
return {
|
138 |
"summary": extract_improved(text, "Summary of Patient's Medical History") or
|
139 |
extract_improved(text, "Summarize the patient's medical history"),
|
|
|
146 |
}
|
147 |
|
148 |
def detect_suicide_risk(text: str) -> Tuple[RiskLevel, float, List[str]]:
|
|
|
149 |
suicide_keywords = [
|
150 |
'suicide', 'suicidal', 'kill myself', 'end my life',
|
151 |
'want to die', 'self-harm', 'self harm', 'hopeless',
|
152 |
'no reason to live', 'plan to die'
|
153 |
]
|
|
|
154 |
explicit_mentions = [kw for kw in suicide_keywords if kw in text.lower()]
|
|
|
155 |
if not explicit_mentions:
|
156 |
return RiskLevel.NONE, 0.0, []
|
157 |
|
|
|
170 |
temperature=0.2,
|
171 |
max_new_tokens=256
|
172 |
)
|
|
|
173 |
json_match = re.search(r'\{.*\}', response, re.DOTALL)
|
174 |
if json_match:
|
175 |
assessment = json.loads(json_match.group())
|
|
|
189 |
return RiskLevel.LOW, risk_score, explicit_mentions
|
190 |
|
191 |
async def create_alert(patient_id: str, risk_data: dict):
|
|
|
192 |
alert_doc = {
|
193 |
"patient_id": patient_id,
|
194 |
"type": "suicide_risk",
|
|
|
208 |
return patient_copy
|
209 |
|
210 |
def compute_patient_data_hash(data: dict) -> str:
|
|
|
211 |
serialized = json.dumps(data, sort_keys=True)
|
212 |
return hashlib.sha256(serialized.encode()).hexdigest()
|
213 |
|
214 |
def compute_file_content_hash(file_content: bytes) -> str:
|
|
|
215 |
return hashlib.sha256(file_content).hexdigest()
|
216 |
|
217 |
def extract_text_from_pdf(pdf_data: bytes) -> str:
|
|
|
218 |
try:
|
219 |
pdf_reader = PyPDF2.PdfReader(io.BytesIO(pdf_data))
|
220 |
text = ""
|
|
|
226 |
raise HTTPException(status_code=400, detail="Failed to extract text from PDF")
|
227 |
|
228 |
async def analyze_patient_report(patient_id: Optional[str], report_content: str, file_type: str, file_content: bytes):
|
229 |
+
identifier = patient_id if patient_id else compute_file_content_hash(file_content)
|
230 |
+
report_data = {"identifier": identifier, "content": report_content, "file_type": file_type}
|
231 |
+
report_hash = compute_patient_data_hash(report_data)
|
232 |
+
logger.info(f"🧾 Analyzing report for identifier: {identifier}")
|
233 |
+
|
234 |
+
existing_analysis = await analysis_collection.find_one({"identifier": identifier, "report_hash": report_hash})
|
235 |
+
if existing_analysis:
|
236 |
+
logger.info(f"✅ No changes in report data for {identifier}, skipping analysis")
|
237 |
+
return existing_analysis
|
238 |
+
|
239 |
+
prompt = (
|
240 |
+
"You are a clinical decision support AI. Analyze the following patient report:\n"
|
241 |
+
"1. Summarize the patient's medical history.\n"
|
242 |
+
"2. Identify risks or red flags (including mental health and suicide risk).\n"
|
243 |
+
"3. Highlight missed diagnoses or treatments.\n"
|
244 |
+
"4. Suggest next clinical steps.\n"
|
245 |
+
f"\nPatient Report ({file_type}):\n{'-'*40}\n{report_content[:10000]}"
|
246 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
247 |
|
248 |
+
raw_response = agent.chat(
|
249 |
+
message=prompt,
|
250 |
+
history=[],
|
251 |
+
temperature=0.7,
|
252 |
+
max_new_tokens=1024
|
253 |
+
)
|
254 |
+
structured_response = structure_medical_response(raw_response)
|
255 |
|
256 |
+
risk_level, risk_score, risk_factors = detect_suicide_risk(raw_response)
|
257 |
+
suicide_risk = {
|
258 |
+
"level": risk_level.value,
|
259 |
+
"score": risk_score,
|
260 |
+
"factors": risk_factors
|
261 |
+
}
|
|
|
|
|
|
|
|
|
|
|
262 |
|
263 |
+
analysis_doc = {
|
264 |
+
"identifier": identifier,
|
265 |
+
"patient_id": patient_id,
|
266 |
+
"timestamp": datetime.utcnow(),
|
267 |
+
"summary": structured_response,
|
268 |
+
"suicide_risk": suicide_risk,
|
269 |
+
"raw": raw_response,
|
270 |
+
"report_hash": report_hash,
|
271 |
+
"file_type": file_type
|
272 |
+
}
|
273 |
|
274 |
+
await analysis_collection.update_one(
|
275 |
+
{"identifier": identifier, "report_hash": report_hash},
|
276 |
+
{"$set": analysis_doc},
|
277 |
+
upsert=True
|
278 |
+
)
|
279 |
|
280 |
+
if patient_id and risk_level in [RiskLevel.MODERATE, RiskLevel.HIGH, RiskLevel.SEVERE]:
|
281 |
+
await create_alert(patient_id, suicide_risk)
|
282 |
|
283 |
+
logger.info(f"✅ Stored analysis for identifier {identifier}")
|
284 |
+
return analysis_doc
|
|
|
285 |
|
286 |
async def analyze_all_patients():
|
|
|
287 |
patients = await patients_collection.find({}).to_list(length=None)
|
288 |
for patient in patients:
|
289 |
await analyze_patient(patient)
|
290 |
await asyncio.sleep(0.1)
|
291 |
|
292 |
async def analyze_patient(patient: dict):
|
|
|
293 |
try:
|
294 |
serialized = serialize_patient(patient)
|
295 |
patient_id = serialized.get("fhir_id")
|
|
|
347 |
logger.error(f"Error analyzing patient: {e}")
|
348 |
|
349 |
def recognize_speech(audio_data: bytes, language: str = "en-US") -> str:
|
|
|
350 |
recognizer = sr.Recognizer()
|
|
|
351 |
try:
|
352 |
with io.BytesIO(audio_data) as audio_file:
|
353 |
with sr.AudioFile(audio_file) as source:
|
|
|
365 |
raise HTTPException(status_code=500, detail="Error processing speech")
|
366 |
|
367 |
def text_to_speech(text: str, language: str = "en", slow: bool = False) -> bytes:
|
|
|
368 |
try:
|
369 |
tts = gTTS(text=text, lang=language, slow=slow)
|
370 |
mp3_fp = io.BytesIO()
|
|
|
396 |
logger.info("✅ TxAgent initialized")
|
397 |
|
398 |
db = get_mongo_client()["cps_db"]
|
399 |
+
global users_collection # Add this to access users_collection for authentication
|
400 |
+
users_collection = db["users"]
|
401 |
patients_collection = db["patients"]
|
402 |
analysis_collection = db["patient_analysis_results"]
|
403 |
alerts_collection = db["clinical_alerts"]
|
|
|
405 |
|
406 |
asyncio.create_task(analyze_all_patients())
|
407 |
|
408 |
+
# Protected Endpoints (add Depends(get_current_user) to all endpoints)
|
409 |
@app.get("/status")
|
410 |
+
async def status(current_user: dict = Depends(get_current_user)):
|
411 |
+
logger.info(f"Status endpoint accessed by {current_user['email']}")
|
412 |
return {
|
413 |
"status": "running",
|
414 |
"timestamp": datetime.utcnow().isoformat(),
|
|
|
417 |
}
|
418 |
|
419 |
@app.get("/patients/analysis-results")
|
420 |
+
async def get_patient_analysis_results(
|
421 |
+
name: Optional[str] = Query(None),
|
422 |
+
current_user: dict = Depends(get_current_user)
|
423 |
+
):
|
424 |
+
logger.info(f"Fetching analysis results by {current_user['email']}")
|
425 |
try:
|
426 |
query = {}
|
427 |
if name:
|
|
|
448 |
raise HTTPException(status_code=500, detail="Failed to retrieve analysis results")
|
449 |
|
450 |
@app.post("/chat-stream")
|
451 |
+
async def chat_stream_endpoint(
|
452 |
+
request: ChatRequest,
|
453 |
+
current_user: dict = Depends(get_current_user)
|
454 |
+
):
|
455 |
+
logger.info(f"Chat stream initiated by {current_user['email']}")
|
456 |
async def token_stream():
|
457 |
try:
|
458 |
conversation = [{"role": "system", "content": agent.chat_prompt}]
|
|
|
486 |
@app.post("/voice/transcribe")
|
487 |
async def transcribe_voice(
|
488 |
audio: UploadFile = File(...),
|
489 |
+
language: str = Query("en-US", description="Language code for speech recognition"),
|
490 |
+
current_user: dict = Depends(get_current_user)
|
491 |
):
|
492 |
+
logger.info(f"Voice transcription initiated by {current_user['email']}")
|
493 |
try:
|
494 |
audio_data = await audio.read()
|
495 |
if not audio.filename.lower().endswith(('.wav', '.mp3', '.ogg', '.flac')):
|
|
|
505 |
raise HTTPException(status_code=500, detail="Error processing voice input")
|
506 |
|
507 |
@app.post("/voice/synthesize")
|
508 |
+
async def synthesize_voice(
|
509 |
+
request: VoiceOutputRequest,
|
510 |
+
current_user: dict = Depends(get_current_user)
|
511 |
+
):
|
512 |
+
logger.info(f"Voice synthesis initiated by {current_user['email']}")
|
513 |
try:
|
514 |
audio_data = text_to_speech(request.text, request.language, request.slow)
|
515 |
|
|
|
533 |
audio: UploadFile = File(...),
|
534 |
language: str = Query("en-US", description="Language code for speech recognition"),
|
535 |
temperature: float = Query(0.7, ge=0.1, le=1.0),
|
536 |
+
max_new_tokens: int = Query(512, ge=50, le=1024),
|
537 |
+
current_user: dict = Depends(get_current_user)
|
538 |
):
|
539 |
+
logger.info(f"Voice chat initiated by {current_user['email']}")
|
540 |
try:
|
541 |
audio_data = await audio.read()
|
542 |
user_message = recognize_speech(audio_data, language)
|
|
|
567 |
file: UploadFile = File(...),
|
568 |
patient_id: Optional[str] = Form(None),
|
569 |
temperature: float = Form(0.5),
|
570 |
+
max_new_tokens: int = Form(1024),
|
571 |
+
current_user: dict = Depends(get_current_user)
|
572 |
):
|
573 |
+
logger.info(f"Report analysis initiated by {current_user['email']}")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
574 |
try:
|
|
|
575 |
content_type = file.content_type
|
576 |
allowed_types = [
|
577 |
'application/pdf',
|
|
|
585 |
detail=f"Unsupported file type: {content_type}. Supported types: PDF, TXT, DOCX"
|
586 |
)
|
587 |
|
|
|
588 |
file_content = await file.read()
|
589 |
|
|
|
590 |
if content_type == 'application/pdf':
|
591 |
text = extract_text_from_pdf(file_content)
|
592 |
elif content_type == 'text/plain':
|
|
|
597 |
else:
|
598 |
raise HTTPException(status_code=400, detail="Unsupported file type")
|
599 |
|
|
|
600 |
text = clean_text_response(text)
|
601 |
if len(text.strip()) < 50:
|
602 |
raise HTTPException(
|
|
|
604 |
detail="Extracted text is too short (minimum 50 characters required)"
|
605 |
)
|
606 |
|
|
|
607 |
analysis = await analyze_patient_report(
|
608 |
patient_id=patient_id,
|
609 |
report_content=text,
|
|
|
611 |
file_content=file_content
|
612 |
)
|
613 |
|
|
|
614 |
if "_id" in analysis and isinstance(analysis["_id"], ObjectId):
|
615 |
analysis["_id"] = str(analysis["_id"])
|
616 |
if "timestamp" in analysis and isinstance(analysis["timestamp"], datetime):
|
617 |
analysis["timestamp"] = analysis["timestamp"].isoformat()
|
618 |
|
|
|
619 |
return JSONResponse(content=jsonable_encoder({
|
620 |
"status": "success",
|
621 |
"analysis": analysis,
|
|
|
633 |
detail=f"Failed to analyze report: {str(e)}"
|
634 |
)
|
635 |
|
|
|
636 |
if __name__ == "__main__":
|
637 |
import uvicorn
|
638 |
uvicorn.run(app, host="0.0.0.0", port=8000)
|