Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,15 +1,17 @@
|
|
| 1 |
import os
|
| 2 |
import sys
|
| 3 |
import json
|
|
|
|
| 4 |
import logging
|
| 5 |
-
from fastapi import FastAPI, HTTPException
|
| 6 |
from fastapi.responses import JSONResponse
|
| 7 |
from fastapi.middleware.cors import CORSMiddleware
|
| 8 |
from typing import List, Dict, Optional
|
| 9 |
from datetime import datetime
|
| 10 |
from pydantic import BaseModel
|
|
|
|
| 11 |
|
| 12 |
-
# Configure logging
|
| 13 |
logging.basicConfig(
|
| 14 |
level=logging.INFO,
|
| 15 |
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
|
|
@@ -32,7 +34,7 @@ except ImportError as e:
|
|
| 32 |
app = FastAPI(
|
| 33 |
title="TxAgent API",
|
| 34 |
description="API for TxAgent medical document analysis",
|
| 35 |
-
version="
|
| 36 |
)
|
| 37 |
|
| 38 |
# CORS configuration
|
|
@@ -44,6 +46,47 @@ app.add_middleware(
|
|
| 44 |
allow_headers=["*"],
|
| 45 |
)
|
| 46 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
# Initialize agent at startup
|
| 48 |
agent = None
|
| 49 |
|
|
@@ -55,7 +98,7 @@ async def startup_event():
|
|
| 55 |
agent = TxAgent(
|
| 56 |
model_name="mims-harvard/TxAgent-T1-Llama-3.1-8B",
|
| 57 |
rag_model_name="mims-harvard/ToolRAG-T1-GTE-Qwen2-1.5B",
|
| 58 |
-
tool_files_dict={},
|
| 59 |
enable_finish=True,
|
| 60 |
enable_rag=False,
|
| 61 |
force_finish=True,
|
|
@@ -69,28 +112,39 @@ async def startup_event():
|
|
| 69 |
logger.error(f"Failed to initialize agent: {str(e)}")
|
| 70 |
raise RuntimeError(f"Failed to initialize agent: {str(e)}")
|
| 71 |
|
| 72 |
-
class ChatRequest(BaseModel):
|
| 73 |
-
message: str
|
| 74 |
-
temperature: float = 0.7
|
| 75 |
-
max_new_tokens: int = 512
|
| 76 |
-
history: Optional[List[Dict]] = None
|
| 77 |
-
|
| 78 |
@app.post("/chat")
|
| 79 |
async def chat_endpoint(request: ChatRequest):
|
| 80 |
-
"""Handle chat conversations"""
|
| 81 |
try:
|
| 82 |
-
logger.info(f"Chat request received: {request.
|
| 83 |
-
|
|
|
|
|
|
|
| 84 |
message=request.message,
|
| 85 |
history=request.history,
|
| 86 |
temperature=request.temperature,
|
| 87 |
max_new_tokens=request.max_new_tokens
|
| 88 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 89 |
return JSONResponse({
|
| 90 |
"status": "success",
|
| 91 |
-
"
|
| 92 |
-
"
|
|
|
|
|
|
|
| 93 |
})
|
|
|
|
| 94 |
except Exception as e:
|
| 95 |
logger.error(f"Chat error: {str(e)}")
|
| 96 |
raise HTTPException(status_code=500, detail=str(e))
|
|
@@ -100,7 +154,8 @@ async def service_status():
|
|
| 100 |
"""Check service status"""
|
| 101 |
return {
|
| 102 |
"status": "running",
|
| 103 |
-
"version": "
|
| 104 |
"model": agent.model_name if agent else "not loaded",
|
|
|
|
| 105 |
"timestamp": datetime.now().isoformat()
|
| 106 |
}
|
|
|
|
| 1 |
import os
|
| 2 |
import sys
|
| 3 |
import json
|
| 4 |
+
import re
|
| 5 |
import logging
|
| 6 |
+
from fastapi import FastAPI, HTTPException
|
| 7 |
from fastapi.responses import JSONResponse
|
| 8 |
from fastapi.middleware.cors import CORSMiddleware
|
| 9 |
from typing import List, Dict, Optional
|
| 10 |
from datetime import datetime
|
| 11 |
from pydantic import BaseModel
|
| 12 |
+
import markdown
|
| 13 |
|
| 14 |
+
# Configure logging
|
| 15 |
logging.basicConfig(
|
| 16 |
level=logging.INFO,
|
| 17 |
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
|
|
|
|
| 34 |
app = FastAPI(
|
| 35 |
title="TxAgent API",
|
| 36 |
description="API for TxAgent medical document analysis",
|
| 37 |
+
version="2.0.0"
|
| 38 |
)
|
| 39 |
|
| 40 |
# CORS configuration
|
|
|
|
| 46 |
allow_headers=["*"],
|
| 47 |
)
|
| 48 |
|
| 49 |
+
# Request models
|
| 50 |
+
class ChatRequest(BaseModel):
|
| 51 |
+
message: str
|
| 52 |
+
temperature: float = 0.7
|
| 53 |
+
max_new_tokens: int = 512
|
| 54 |
+
history: Optional[List[Dict]] = None
|
| 55 |
+
format: Optional[str] = "clean" # Options: raw, clean, structured, html
|
| 56 |
+
|
| 57 |
+
# Response cleaning functions
|
| 58 |
+
def clean_text_response(text: str) -> str:
|
| 59 |
+
"""Basic text cleaning"""
|
| 60 |
+
# Remove excessive newlines and whitespace
|
| 61 |
+
text = re.sub(r'\n\s*\n', '\n\n', text)
|
| 62 |
+
# Remove redundant spaces
|
| 63 |
+
text = re.sub(r'[ ]+', ' ', text)
|
| 64 |
+
# Clean markdown formatting
|
| 65 |
+
text = text.replace("**", "").replace("__", "")
|
| 66 |
+
return text.strip()
|
| 67 |
+
|
| 68 |
+
def structure_medical_response(text: str) -> Dict:
|
| 69 |
+
"""Structure medical content into categories"""
|
| 70 |
+
result = {"overview": "", "symptoms": [], "types": {}, "notes": ""}
|
| 71 |
+
|
| 72 |
+
# Extract overview
|
| 73 |
+
overview_end = text.find("Type 1 Diabetes:")
|
| 74 |
+
result["overview"] = clean_text_response(text[:overview_end])
|
| 75 |
+
|
| 76 |
+
# Extract by type
|
| 77 |
+
type_sections = re.split(r'(Type \d Diabetes:)', text[overview_end:])
|
| 78 |
+
current_type = None
|
| 79 |
+
|
| 80 |
+
for section in type_sections:
|
| 81 |
+
if section.startswith("Type "):
|
| 82 |
+
current_type = section.replace(":", "").strip().lower()
|
| 83 |
+
result["types"][current_type] = []
|
| 84 |
+
elif current_type:
|
| 85 |
+
points = [p.strip() for p in section.split('\n') if p.strip()]
|
| 86 |
+
result["types"][current_type] = points
|
| 87 |
+
|
| 88 |
+
return result
|
| 89 |
+
|
| 90 |
# Initialize agent at startup
|
| 91 |
agent = None
|
| 92 |
|
|
|
|
| 98 |
agent = TxAgent(
|
| 99 |
model_name="mims-harvard/TxAgent-T1-Llama-3.1-8B",
|
| 100 |
rag_model_name="mims-harvard/ToolRAG-T1-GTE-Qwen2-1.5B",
|
| 101 |
+
tool_files_dict={},
|
| 102 |
enable_finish=True,
|
| 103 |
enable_rag=False,
|
| 104 |
force_finish=True,
|
|
|
|
| 112 |
logger.error(f"Failed to initialize agent: {str(e)}")
|
| 113 |
raise RuntimeError(f"Failed to initialize agent: {str(e)}")
|
| 114 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 115 |
@app.post("/chat")
|
| 116 |
async def chat_endpoint(request: ChatRequest):
|
| 117 |
+
"""Handle chat conversations with formatting options"""
|
| 118 |
try:
|
| 119 |
+
logger.info(f"Chat request received (format: {request.format})")
|
| 120 |
+
|
| 121 |
+
# Get raw response from agent
|
| 122 |
+
raw_response = agent.chat(
|
| 123 |
message=request.message,
|
| 124 |
history=request.history,
|
| 125 |
temperature=request.temperature,
|
| 126 |
max_new_tokens=request.max_new_tokens
|
| 127 |
)
|
| 128 |
+
|
| 129 |
+
# Format response based on request
|
| 130 |
+
formatted_response = {
|
| 131 |
+
"raw": raw_response,
|
| 132 |
+
"clean": clean_text_response(raw_response),
|
| 133 |
+
"structured": structure_medical_response(raw_response),
|
| 134 |
+
"html": markdown.markdown(raw_response)
|
| 135 |
+
}
|
| 136 |
+
|
| 137 |
+
# Select the requested format
|
| 138 |
+
response_content = formatted_response.get(request.format, formatted_response["clean"])
|
| 139 |
+
|
| 140 |
return JSONResponse({
|
| 141 |
"status": "success",
|
| 142 |
+
"format": request.format,
|
| 143 |
+
"response": response_content,
|
| 144 |
+
"timestamp": datetime.now().isoformat(),
|
| 145 |
+
"available_formats": list(formatted_response.keys())
|
| 146 |
})
|
| 147 |
+
|
| 148 |
except Exception as e:
|
| 149 |
logger.error(f"Chat error: {str(e)}")
|
| 150 |
raise HTTPException(status_code=500, detail=str(e))
|
|
|
|
| 154 |
"""Check service status"""
|
| 155 |
return {
|
| 156 |
"status": "running",
|
| 157 |
+
"version": "2.0.0",
|
| 158 |
"model": agent.model_name if agent else "not loaded",
|
| 159 |
+
"formats_available": ["raw", "clean", "structured", "html"],
|
| 160 |
"timestamp": datetime.now().isoformat()
|
| 161 |
}
|