Update app.py

app.py

@@ -1,4 +1,8 @@
-
+import sys
+import os
+import pandas as pd
+import pdfplumber
+import json
 import gradio as gr
 from typing import List
 from concurrent.futures import ThreadPoolExecutor, as_completed
@@ -7,6 +11,19 @@ import shutil
 import re
 import psutil
 import subprocess
+import logging
+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"
@@ -56,6 +73,7 @@ def extract_priority_pages(file_path: str, max_pages: int = 20) -> str:
                     text_chunks.append(f"=== Page {i} ===\n{page_text.strip()}")
         return "\n\n".join(text_chunks)
     except Exception as e:
+        logger.error(f"Error extracting pages from PDF: {str(e)}")
         return f"PDF processing error: {str(e)}"
 
 def convert_file_to_json(file_path: str, file_type: str) -> str:
@@ -83,29 +101,31 @@ def convert_file_to_json(file_path: str, file_type: str) -> str:
             result = json.dumps({"filename": os.path.basename(file_path), "rows": content})
         else:
             result = json.dumps({"error": f"Unsupported file type: {file_type}"})
+        
         with open(cache_path, "w", encoding="utf-8") as f:
             f.write(result)
         return result
     except Exception as e:
+        logger.error(f"Error converting {file_type} file to JSON: {str(e)}")
         return json.dumps({"error": f"Error processing {os.path.basename(file_path)}: {str(e)}"})
 
 def log_system_usage(tag=""):
     try:
         cpu = psutil.cpu_percent(interval=1)
         mem = psutil.virtual_memory()
-        print(f"[{tag}] CPU: {cpu}% | RAM: {mem.used // (1024**2)}MB / {mem.total // (1024**2)}MB")
+        logger.info(f"[{tag}] CPU: {cpu}% | RAM: {mem.used // (1024**2)}MB / {mem.total // (1024**2)}MB")
         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(", ")
-            print(f"[{tag}] GPU: {used}MB / {total}MB | Utilization: {util}%")
+            logger.info(f"[{tag}] GPU: {used}MB / {total}MB | Utilization: {util}%")
     except Exception as e:
-        print(f"[{tag}] GPU/CPU monitor failed: {e}")
+        logger.error(f"[{tag}] GPU/CPU monitor failed: {e}")
 
 def init_agent():
-    print("🔁 Initializing model...")
+    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")
@@ -124,33 +144,59 @@ def init_agent():
     )
     agent.init_model()
     log_system_usage("After Load")
-    print("✅ Agent Ready")
+    logger.info("✅ Agent Ready")
     return agent
 
-def create_ui(agent):
-    with gr.Blocks(theme=gr.themes.Soft()) as demo:
-        gr.Markdown("<h1 style='text-align: center;'>🩺 Clinical Oversight Assistant</h1>")
-        chatbot = gr.Chatbot(label="Analysis", height=600, type="messages")
-        file_upload = gr.File(file_types=[".pdf", ".csv", ".xls", ".xlsx"], file_count="multiple")
-        msg_input = gr.Textbox(placeholder="Ask about potential oversights...", show_label=False)
-        send_btn = gr.Button("Analyze", variant="primary")
-        download_output = gr.File(label="Download Full Report")
-
-        def analyze(message: str, history: list, files: list):
-            history = history + [{"role": "user", "content": message},
-                                 {"role": "assistant", "content": "⏳ Analyzing records for potential oversights..."}]
-            yield history, None
+def format_response_for_ui(response: str) -> str:
+    """Formats the raw response for clean display in the UI"""
+    # Remove any tool call metadata
+    cleaned = response.split("[TOOL_CALLS]")[0].strip()
+    
+    # If we have a structured response, format it nicely
+    if "Potential missed diagnoses" in cleaned or "Flagged medication conflicts" in cleaned:
+        # Add markdown formatting for better readability
+        formatted = []
+        for line in cleaned.split("\n"):
+            if line.startswith("Potential missed diagnoses"):
+                formatted.append(f"### 🔍 Potential Missed Diagnoses")
+            elif line.startswith("Flagged medication conflicts"):
+                formatted.append(f"\n### ⚠️ Flagged Medication Conflicts")
+            elif line.startswith("Incomplete assessments"):
+                formatted.append(f"\n### 📋 Incomplete Assessments")
+            elif line.startswith("Highlighted abnormal results"):
+                formatted.append(f"\n### ❗ Abnormal Results Needing Follow-up")
+            else:
+                formatted.append(line)
+        return "\n".join(formatted)
+    return cleaned
+
+def analyze(message: str, history: list, files: list):
+    start_time = datetime.now()
+    logger.info(f"Starting analysis for message: {message[:100]}...")
+    if files:
+        logger.info(f"Processing {len(files)} uploaded files")
+
+    history = history + [{"role": "user", "content": message},
+                         {"role": "assistant", "content": "⏳ Analyzing records for potential oversights..."}]
+    yield history, None
 
-            extracted = ""
-            file_hash_value = ""
-            if files:
-                with ThreadPoolExecutor(max_workers=4) as executor:
-                    futures = [executor.submit(convert_file_to_json, f.name, f.name.split(".")[-1].lower()) for f in files]
-                    results = [sanitize_utf8(f.result()) for f in as_completed(futures)]
-                    extracted = "\n".join(results)
-                    file_hash_value = file_hash(files[0].name)
+    extracted = ""
+    file_hash_value = ""
+    if files:
+        try:
+            with ThreadPoolExecutor(max_workers=4) as executor:
+                futures = [executor.submit(convert_file_to_json, f.name, f.name.split(".")[-1].lower()) for f in files]
+                results = [sanitize_utf8(f.result()) for f in as_completed(futures)]
+                extracted = "\n".join(results)
+                file_hash_value = file_hash(files[0].name)
+                logger.info(f"Processed {len(files)} files, extracted {len(extracted)} characters")
+        except Exception as e:
+            logger.error(f"Error processing files: {str(e)}")
+            history[-1] = {"role": "assistant", "content": f"❌ Error processing files: {str(e)}"}
+            yield history, None
+            return
 
-            prompt = f"""Review these medical records and identify EXACTLY what might have been missed:
+    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
@@ -161,54 +207,145 @@ Medical Records:
 
 ### Potential Oversights:
 """
-            response_chunks = []
+    logger.info(f"Generated prompt with {len(prompt)} characters")
+
+    response_chunks = []
+    try:
+        logger.info("Starting model inference...")
+        for chunk in agent.run_gradio_chat(
+            message=prompt,
+            history=[],
+            temperature=0.2,
+            max_new_tokens=1024,
+            max_token=4096,
+            call_agent=False,
+            conversation=[]
+        ):
+            if not chunk:
+                continue
+            if isinstance(chunk, str):
+                response_chunks.append(chunk)
+            elif isinstance(chunk, list):
+                response_chunks.extend([c.content for c in chunk if hasattr(c, 'content')])
+            
+            partial_response = "".join(response_chunks)
+            formatted_partial = format_response_for_ui(partial_response)
+            
+            if formatted_partial:
+                history[-1] = {"role": "assistant", "content": formatted_partial}
+                yield history, None
+
+        full_response = "".join(response_chunks)
+        logger.info(f"Full model response received: {full_response[:500]}...")
+        
+        final_output = format_response_for_ui(full_response)
+        if not final_output or len(final_output) < 20:  # Very short response
+            final_output = "No clear oversights identified. Recommend comprehensive review."
+            logger.info("No significant findings detected in analysis")
+        
+        history[-1] = {"role": "assistant", "content": final_output}
+
+        # Save report
+        report_path = None
+        if file_hash_value:
+            report_path = os.path.join(report_dir, f"{file_hash_value}_report.txt")
             try:
-                for chunk in agent.run_gradio_chat(
-                    message=prompt,
-                    history=[],
-                    temperature=0.2,
-                    max_new_tokens=1024,
-                    max_token=4096,
-                    call_agent=False,
-                    conversation=[]
-                ):
-                    if not chunk:
-                        continue
-                    if isinstance(chunk, str):
-                        response_chunks.append(chunk)
-                    elif isinstance(chunk, list):
-                        response_chunks.extend([c.content for c in chunk if hasattr(c, 'content')])
-                    partial_response = "".join(response_chunks)
-                    cleaned_partial = partial_response.split("[TOOL_CALLS]")[0].strip()
-                    if cleaned_partial:
-                        history[-1] = {"role": "assistant", "content": cleaned_partial}
-                        yield history, None
+                with open(report_path, "w", encoding="utf-8") as f:
+                    f.write(final_output)
+                logger.info(f"Saved report to {report_path}")
             except Exception as e:
-                history[-1] = {"role": "assistant", "content": f"❌ Error: {str(e)}"}
-                yield history, None
-                return
+                logger.error(f"Error saving report: {str(e)}")
 
-            full_response = "".join(response_chunks)
-            final_output = full_response.split("[TOOL_CALLS]")[0].strip()
-            if not final_output:
-                final_output = "No clear oversights identified. Recommend comprehensive review."
-            history[-1] = {"role": "assistant", "content": final_output}
+        elapsed = (datetime.now() - start_time).total_seconds()
+        logger.info(f"Analysis completed in {elapsed:.2f} seconds")
+        yield history, report_path if report_path and os.path.exists(report_path) else None
 
-            report_path = os.path.join(report_dir, f"{file_hash_value}_report.txt") if file_hash_value else None
-            yield history, report_path if report_path and os.path.exists(report_path) else None
+    except Exception as e:
+        logger.error(f"Error during analysis: {str(e)}", exc_info=True)
+        history[-1] = {"role": "assistant", "content": f"❌ Error during analysis: {str(e)}"}
+        yield history, None
+
+def create_ui(agent):
+    with gr.Blocks(theme=gr.themes.Soft(), title="Clinical Oversight Assistant") as demo:
+        gr.Markdown("<h1 style='text-align: center;'>🩺 Clinical Oversight Assistant</h1>")
+        gr.Markdown("""
+        <div style='text-align: center; margin-bottom: 20px;'>
+            Upload medical records and receive analysis of potential oversights, including:<br>
+            - Missed diagnoses - Medication conflicts - Incomplete assessments - Abnormal results needing follow-up
+        </div>
+        """)
+        
+        with gr.Row():
+            with gr.Column(scale=2):
+                file_upload = gr.File(
+                    label="Upload Medical Records",
+                    file_types=[".pdf", ".csv", ".xls", ".xlsx"],
+                    file_count="multiple",
+                    interactive=True
+                )
+                msg_input = gr.Textbox(
+                    placeholder="Ask about potential oversights...",
+                    show_label=False,
+                    lines=3,
+                    max_lines=5
+                )
+                send_btn = gr.Button("Analyze", variant="primary")
+            
+            with gr.Column(scale=3):
+                chatbot = gr.Chatbot(
+                    label="Analysis Results",
+                    height=600,
+                    bubble_full_width=False,
+                    show_copy_button=True
+                )
+                download_output = gr.File(
+                    label="Download Full Report",
+                    interactive=False
+                )
+
+        # Examples for quick testing
+        examples = gr.Examples(
+            examples=[
+                ["Are there any potential missed diagnoses in these records?"],
+                ["What medication conflicts should I be aware of?"],
+                ["Are there any incomplete assessments in this case?"]
+            ],
+            inputs=[msg_input],
+            label="Example Questions"
+        )
+
+        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]
+        )
+
+        # Add some footer text
+        gr.Markdown("""
+        <div style='text-align: center; margin-top: 20px; color: #666; font-size: 0.9em;'>
+            Note: This tool provides preliminary analysis only. Always verify findings with complete clinical evaluation.
+        </div>
+        """)
 
-        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])
     return demo
 
 if __name__ == "__main__":
-    print("🚀 Launching app...")
-    agent = init_agent()
-    demo = create_ui(agent)
-    demo.queue(api_open=False).launch(
-        server_name="0.0.0.0",
-        server_port=7860,
-        show_error=True,
-        allowed_paths=[report_dir],
-        share=False
-    )
+    logger.info("🚀 Launching Clinical Oversight Assistant...")
+    try:
+        agent = init_agent()
+        demo = create_ui(agent)
+        demo.queue(api_open=False).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"Failed to launch application: {str(e)}", exc_info=True)
+        raise