import sys
import os
import pandas as pd
import pdfplumber
import json
import gradio as gr
from typing import List, Dict, Any
from concurrent.futures import ThreadPoolExecutor, as_completed
import hashlib
import shutil
import re
import psutil
import subprocess
import logging
import traceback
from datetime import datetime
# Configure logging
logging.basicConfig(
level=logging.INFO,
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
handlers=[
logging.StreamHandler(),
logging.FileHandler('clinical_oversight.log')
]
)
logger = logging.getLogger(__name__)
# Persistent directory
persistent_dir = "/data/hf_cache"
os.makedirs(persistent_dir, exist_ok=True)
model_cache_dir = os.path.join(persistent_dir, "txagent_models")
tool_cache_dir = os.path.join(persistent_dir, "tool_cache")
file_cache_dir = os.path.join(persistent_dir, "cache")
report_dir = os.path.join(persistent_dir, "reports")
vllm_cache_dir = os.path.join(persistent_dir, "vllm_cache")
for directory in [model_cache_dir, tool_cache_dir, file_cache_dir, report_dir, vllm_cache_dir]:
os.makedirs(directory, exist_ok=True)
os.environ["HF_HOME"] = model_cache_dir
os.environ["TRANSFORMERS_CACHE"] = model_cache_dir
os.environ["VLLM_CACHE_DIR"] = vllm_cache_dir
os.environ["TOKENIZERS_PARALLELISM"] = "false"
os.environ["CUDA_LAUNCH_BLOCKING"] = "1"
current_dir = os.path.dirname(os.path.abspath(__file__))
src_path = os.path.abspath(os.path.join(current_dir, "src"))
sys.path.insert(0, src_path)
from txagent.txagent import TxAgent
MEDICAL_KEYWORDS = {'diagnosis', 'assessment', 'plan', 'results', 'medications',
'allergies', 'summary', 'impression', 'findings', 'recommendations'}
def sanitize_utf8(text: str) -> str:
"""Ensure text is UTF-8 encoded and clean."""
try:
return text.encode("utf-8", "ignore").decode("utf-8")
except Exception as e:
logger.error(f"UTF-8 sanitization failed: {str(e)}")
return ""
def file_hash(path: str) -> str:
"""Generate MD5 hash of file content."""
try:
with open(path, "rb") as f:
return hashlib.md5(f.read()).hexdigest()
except Exception as e:
logger.error(f"File hash generation failed for {path}: {str(e)}")
return ""
def extract_priority_pages(file_path: str, max_pages: int = 20) -> str:
"""Extract pages from PDF with priority given to pages containing medical keywords."""
try:
text_chunks = []
logger.info(f"Extracting pages from {file_path}")
with pdfplumber.open(file_path) as pdf:
# Always extract first 3 pages
for i, page in enumerate(pdf.pages[:3]):
try:
text = page.extract_text() or ""
text_chunks.append(f"=== Page {i+1} ===\n{text.strip()}")
except Exception as page_error:
logger.warning(f"Error processing page {i+1}: {str(page_error)}")
text_chunks.append(f"=== Page {i+1} ===\n[Error extracting content]")
# Extract remaining pages that contain medical keywords
for i, page in enumerate(pdf.pages[3:max_pages], start=4):
try:
page_text = page.extract_text() or ""
if any(re.search(rf'\b{kw}\b', page_text.lower()) for kw in MEDICAL_KEYWORDS):
text_chunks.append(f"=== Page {i} ===\n{page_text.strip()}")
except Exception as page_error:
logger.warning(f"Error processing page {i}: {str(page_error)}")
return "\n\n".join(text_chunks)
except Exception as e:
logger.error(f"PDF processing error for {file_path}: {str(e)}")
return f"PDF processing error: {str(e)}"
def convert_file_to_json(file_path: str, file_type: str) -> str:
"""Convert different file types to JSON format with caching."""
try:
h = file_hash(file_path)
if not h:
return json.dumps({"error": "Could not generate file hash"})
cache_path = os.path.join(file_cache_dir, f"{h}.json")
# Check cache first
if os.path.exists(cache_path):
try:
with open(cache_path, "r", encoding="utf-8") as f:
return f.read()
except Exception as cache_error:
logger.error(f"Cache read error for {file_path}: {str(cache_error)}")
result = {}
try:
if file_type == "pdf":
text = extract_priority_pages(file_path)
result = {
"filename": os.path.basename(file_path),
"content": text,
"status": "initial",
"file_type": "pdf"
}
elif file_type == "csv":
df = pd.read_csv(
file_path,
encoding_errors="replace",
header=None,
dtype=str,
skip_blank_lines=False,
on_bad_lines="skip"
)
content = df.fillna("").astype(str).values.tolist()
result = {
"filename": os.path.basename(file_path),
"rows": content,
"file_type": "csv"
}
elif file_type in ["xls", "xlsx"]:
try:
df = pd.read_excel(file_path, engine="openpyxl", header=None, dtype=str)
except Exception:
try:
df = pd.read_excel(file_path, engine="xlrd", header=None, dtype=str)
except Exception as excel_error:
logger.error(f"Excel read error for {file_path}: {str(excel_error)}")
raise
content = df.fillna("").astype(str).values.tolist()
result = {
"filename": os.path.basename(file_path),
"rows": content,
"file_type": "excel"
}
else:
result = {"error": f"Unsupported file type: {file_type}"}
json_result = json.dumps(result)
# Save to cache
try:
with open(cache_path, "w", encoding="utf-8") as f:
f.write(json_result)
except Exception as cache_write_error:
logger.error(f"Cache write error for {file_path}: {str(cache_write_error)}")
return json_result
except Exception as processing_error:
logger.error(f"Error processing {file_path}: {str(processing_error)}")
return json.dumps({"error": f"Error processing {os.path.basename(file_path)}: {str(processing_error)}"})
except Exception as e:
logger.error(f"Unexpected error in convert_file_to_json: {str(e)}")
return json.dumps({"error": f"Unexpected error processing file: {str(e)}"})
def log_system_usage(tag=""):
"""Log system resource usage including CPU, RAM, and GPU."""
try:
cpu = psutil.cpu_percent(interval=1)
mem = psutil.virtual_memory()
logger.info(f"[{tag}] CPU: {cpu}% | RAM: {mem.used // (1024**2)}MB / {mem.total // (1024**2)}MB")
try:
result = subprocess.run(
["nvidia-smi", "--query-gpu=memory.used,memory.total,utilization.gpu", "--format=csv,nounits,noheader"],
capture_output=True, text=True
)
if result.returncode == 0:
used, total, util = result.stdout.strip().split(", ")
logger.info(f"[{tag}] GPU: {used}MB / {total}MB | Utilization: {util}%")
except Exception as gpu_error:
logger.warning(f"[{tag}] GPU monitor failed: {gpu_error}")
except Exception as e:
logger.error(f"System usage logging failed: {str(e)}")
def init_agent():
"""Initialize the TxAgent with proper configuration."""
logger.info("Initializing model...")
log_system_usage("Before Load")
default_tool_path = os.path.abspath("data/new_tool.json")
target_tool_path = os.path.join(tool_cache_dir, "new_tool.json")
try:
if not os.path.exists(target_tool_path):
shutil.copy(default_tool_path, target_tool_path)
logger.info("Copied default tool configuration")
except Exception as e:
logger.error(f"Tool configuration copy failed: {str(e)}")
raise
try:
agent = TxAgent(
model_name="mims-harvard/TxAgent-T1-Llama-3.1-8B",
rag_model_name="mims-harvard/ToolRAG-T1-GTE-Qwen2-1.5B",
tool_files_dict={"new_tool": target_tool_path},
force_finish=True,
enable_checker=True,
step_rag_num=8,
seed=100,
additional_default_tools=[],
)
agent.init_model()
log_system_usage("After Load")
logger.info("Agent initialization successful")
return agent
except Exception as e:
logger.error(f"Agent initialization failed: {str(e)}")
raise
def save_report(content: str, file_hash_value: str = "") -> str:
"""Save analysis report to file and return path."""
try:
if not file_hash_value:
file_hash_value = hashlib.md5(content.encode()).hexdigest()
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
report_filename = f"report_{timestamp}_{file_hash_value[:8]}.txt"
report_path = os.path.join(report_dir, report_filename)
with open(report_path, "w", encoding="utf-8") as f:
f.write(content)
logger.info(f"Report saved to {report_path}")
return report_path
except Exception as e:
logger.error(f"Failed to save report: {str(e)}")
return ""
def clean_response(content: str) -> str:
"""Clean up model response by removing tool call artifacts."""
if not content:
return "⚠️ No content generated."
try:
# Remove tool call artifacts
cleaned = re.sub(r"\[TOOL_CALLS\].*?(?=(\[|\Z))", "", content, flags=re.DOTALL).strip()
# Remove excessive whitespace
cleaned = re.sub(r"\n{3,}", "\n\n", cleaned)
return cleaned or "⚠️ Empty response after cleaning."
except Exception as e:
logger.error(f"Response cleaning failed: {str(e)}")
return content
def process_model_response(chunk: Any, history: List[Dict[str, str]]) -> List[Dict[str, str]]:
"""Process model response chunk and update chat history."""
try:
if chunk is None:
return history
if isinstance(chunk, list) and all(hasattr(m, 'role') and hasattr(m, 'content') for m in chunk):
for m in chunk:
cleaned_content = clean_response(m.content)
history.append({"role": m.role, "content": cleaned_content})
elif isinstance(chunk, str):
cleaned_chunk = clean_response(chunk)
if history and history[-1]["role"] == "assistant":
history[-1]["content"] += cleaned_chunk
else:
history.append({"role": "assistant", "content": cleaned_chunk})
else:
logger.warning(f"Unexpected response type: {type(chunk)}")
return history
except Exception as e:
logger.error(f"Error processing model response: {str(e)}")
history.append({"role": "assistant", "content": f"⚠️ Error processing response: {str(e)}"})
return history
def analyze(message: str, history: list, files: list):
"""Main analysis function that processes files and generates responses."""
try:
# Initial response
new_history = history.copy()
new_history.append({"role": "user", "content": message})
new_history.append({"role": "assistant", "content": "⏳ Analyzing records for potential oversights..."})
yield new_history, None
# Process files
extracted = ""
file_hash_value = ""
if files:
logger.info(f"Processing {len(files)} files...")
with ThreadPoolExecutor(max_workers=4) as executor:
futures = []
for f in files:
try:
file_type = f.name.split(".")[-1].lower()
futures.append(executor.submit(convert_file_to_json, f.name, file_type))
except Exception as e:
logger.error(f"Error submitting file {f.name} for processing: {str(e)}")
new_history.append({"role": "system", "content": f"⚠️ Error processing {f.name}: {str(e)}"})
results = []
for f in as_completed(futures):
try:
results.append(sanitize_utf8(f.result()))
except Exception as e:
logger.error(f"Error getting file processing result: {str(e)}")
results.append(json.dumps({"error": "File processing failed"}))
extracted = "\n".join(results)
try:
file_hash_value = file_hash(files[0].name) if files else ""
except Exception as e:
logger.error(f"Error generating file hash: {str(e)}")
file_hash_value = ""
# Prepare prompt
prompt = f"""Review these medical records and identify EXACTLY what might have been missed:
1. List potential missed diagnoses
2. Flag any medication conflicts
3. Note incomplete assessments
4. Highlight abnormal results needing follow-up
Medical Records:
{extracted[:12000]}
### Potential Oversights:
"""
logger.info(f"Prompt length: {len(prompt)} characters")
# Initialize agent response
agent = init_agent()
response_content = ""
report_path = ""
# Process agent response
for chunk in agent.run_gradio_chat(
message=prompt,
history=[],
temperature=0.2,
max_new_tokens=2048,
max_token=4096,
call_agent=False,
conversation=[],
):
try:
new_history = process_model_response(chunk, new_history)
if isinstance(chunk, str):
response_content += clean_response(chunk)
yield new_history, None
except Exception as chunk_error:
logger.error(f"Error processing chunk: {str(chunk_error)}")
new_history.append({"role": "assistant", "content": f"⚠️ Error processing response chunk: {str(chunk_error)}"})
yield new_history, None
# Save final report
if response_content:
try:
report_path = save_report(response_content, file_hash_value)
except Exception as report_error:
logger.error(f"Error saving report: {str(report_error)}")
new_history.append({"role": "system", "content": "⚠️ Failed to save full report"})
yield new_history, report_path if report_path and os.path.exists(report_path) else None
except Exception as e:
logger.error(f"Analysis error: {str(e)}\n{traceback.format_exc()}")
error_history = history.copy()
error_history.append({"role": "assistant", "content": f"❌ Critical error occurred: {str(e)}"})
yield error_history, None
def create_ui(agent):
"""Create Gradio UI interface."""
with gr.Blocks(theme=gr.themes.Soft(), title="Clinical Oversight Assistant") as demo:
gr.Markdown("🩺 Clinical Oversight Assistant
")
gr.Markdown("""
Upload medical records and ask about potential oversights or missed diagnoses.
""")
with gr.Row():
with gr.Column(scale=2):
chatbot = gr.Chatbot(
label="Analysis Conversation",
height=600,
bubble_full_width=False,
show_copy_button=True
)
msg_input = gr.Textbox(
placeholder="Ask about potential oversights...",
show_label=False,
container=False
)
with gr.Row():
send_btn = gr.Button("Analyze", variant="primary")
clear_btn = gr.Button("Clear")
with gr.Column(scale=1):
file_upload = gr.File(
file_types=[".pdf", ".csv", ".xls", ".xlsx"],
file_count="multiple",
label="Upload Medical Records"
)
download_output = gr.File(
label="Download Full Report",
interactive=False
)
gr.Markdown("""
Note: The system analyzes PDFs, CSVs, and Excel files for potential clinical oversights.
""")
# Event handlers
send_btn.click(
analyze,
inputs=[msg_input, gr.State([]), file_upload],
outputs=[chatbot, download_output]
)
msg_input.submit(
analyze,
inputs=[msg_input, gr.State([]), file_upload],
outputs=[chatbot, download_output]
)
clear_btn.click(
lambda: ([], None),
inputs=[],
outputs=[chatbot, download_output]
)
# Add some examples
gr.Examples(
examples=[
["What potential diagnoses might have been missed in these records?"],
["Are there any medication conflicts I should be aware of?"],
["What abnormal results need follow-up in these reports?"]
],
inputs=msg_input,
label="Example Questions"
)
return demo
if __name__ == "__main__":
try:
logger.info("🚀 Launching Clinical Oversight Assistant...")
agent = init_agent()
demo = create_ui(agent)
demo.queue(
api_open=False,
concurrency_count=2
).launch(
server_name="0.0.0.0",
server_port=7860,
show_error=True,
allowed_paths=[report_dir],
share=False
)
except Exception as e:
logger.error(f"Application failed to start: {str(e)}\n{traceback.format_exc()}")
raise