Update app.py

app.py

@@ -91,13 +91,13 @@ def log_system_usage(tag=""):
 def clean_response(text: str) -> str:
     text = sanitize_utf8(text)
     # Strip all tool and reasoning text
-    text = re.sub(r"\[TOOL_CALLS\].*?\n|\[.*?\].*?\n|(?:get_|tool\s|retrieve\s|use\s).*?\n", "", text, flags=re.DOTALL | re.IGNORECASE)
+    text = re.sub(r"\[TOOL_CALLS\].*?\n|\[.*?\].*?\n|(?:get_|tool\s|retrieve\s|use\s|rag\s).*?\n", "", text, flags=re.DOTALL | re.IGNORECASE)
     text = re.sub(r"\{'meta':\s*\{.*?\}\s*,\s*'results':\s*\[.*?\]\}\n?", "", text, flags=re.DOTALL)
     text = re.sub(
         r"(?i)(to\s|analyze|will\s|since\s|no\s|none|previous|attempt|involve|check\s|explore|manually|"
         r"start|look|use|focus|retrieve|tool|based\s|overall|indicate|mention|consider|ensure|need\s|"
         r"provide|review|assess|identify|potential|records|patient|history|symptoms|medication|"
-        r"conflict|assessment|follow-up|issue|reasoning|step|prompt|address|rag|thought|try).*?\n",
+        r"conflict|assessment|follow-up|issue|reasoning|step|prompt|address|rag|thought|try|john\sdoe|nkma).*?\n",
         "", text, flags=re.DOTALL
     )
     text = re.sub(r"\n{2,}", "\n", text).strip()
@@ -116,11 +116,9 @@ def clean_response(text: str) -> str:
     return "\n".join(lines).strip()
 
 def normalize_text(text: str) -> str:
-    # Normalize for deduplication
     return re.sub(r"\s+", " ", text.lower().strip())
 
 def consolidate_findings(responses: List[str]) -> str:
-    # Merge unique findings
     findings = defaultdict(set)
     headings = ["Missed Diagnoses", "Medication Conflicts", "Incomplete Assessments", "Urgent Follow-up"]
     
@@ -137,12 +135,10 @@ def consolidate_findings(responses: List[str]) -> str:
             elif current_heading and line.startswith("-"):
                 findings[current_heading].add(normalize_text(line))
     
-    # Format output
     output = []
     for heading in headings:
         if findings[heading]:
             output.append(f"**{heading}**:")
-            # Restore original case for display
             original_lines = {normalize_text(r): r for r in sum([r.split("\n") for r in responses], []) if r.startswith("-")}
             output.extend(sorted(original_lines.get(n, "- " + n) for n in findings[heading]))
     return "\n".join(output).strip() if output else "No oversights identified."
@@ -154,12 +150,11 @@ def init_agent():
         model_name="mims-harvard/TxAgent-T1-Llama-3.1-8B",
         rag_model_name="mims-harvard/ToolRAG-T1-GTE-Qwen2-1.5B",
         force_finish=True,
-        enable_checker=False,  # Disable to avoid repeat_thought
-        enable_rag=False,  # Disable all RAG
-        tool_files_dict=None,  # No tools
+        enable_checker=False,
+        enable_rag=False,
+        tool_files_dict=None,
         step_rag_num=0,
         seed=100,
-        enforce_eager=True,  # No torch.compile
     )
     agent.init_model()
     log_system_usage("After Load")
@@ -189,7 +184,6 @@ def create_ui(agent):
                     extracted = "\n".join([json.loads(r).get("content", "") for r in results if "content" in json.loads(r)])
                     file_hash_value = file_hash(files[0].name) if files else ""
 
-            # Smaller chunks
             chunk_size = 800
             chunks = [extracted[i:i + chunk_size] for i in range(0, len(extracted), chunk_size)]
             chunk_responses = []
@@ -209,7 +203,6 @@ Records:
 """
 
             try:
-                # Batch process with progress
                 for i in range(0, len(chunks), batch_size):
                     batch = chunks[i:i + batch_size]
                     batch_responses = []
@@ -239,18 +232,15 @@ Records:
                                     chunk_response += cleaned + "\n"
                         if chunk_response:
                             batch_responses.append(chunk_response)
-                        # Progress update
                         processed = min(i + j + 1, total_chunks)
                         history[-1]["content"] = f"🔄 Analyzing... ({processed}/{total_chunks} chunks)"
                         yield history, None
                     chunk_responses.extend(batch_responses)
 
-                # Final result
                 final_response = consolidate_findings(chunk_responses)
                 history[-1]["content"] = final_response
                 yield history, None
 
-                # Report
                 report_path = os.path.join(report_dir, f"{file_hash_value}_report.txt") if file_hash_value else None
                 if report_path and final_response != "No oversights identified.":
                     with open(report_path, "w", encoding="utf-8") as f: