Update app.py
Browse files
app.py
CHANGED
|
@@ -33,18 +33,15 @@ MAX_CHUNK_TOKENS = 8192
|
|
| 33 |
MAX_NEW_TOKENS = 2048
|
| 34 |
PROMPT_OVERHEAD = 500
|
| 35 |
|
| 36 |
-
|
| 37 |
def clean_response(text: str) -> str:
|
| 38 |
text = re.sub(r"\[.*?\]|\bNone\b", "", text, flags=re.DOTALL)
|
| 39 |
text = re.sub(r"\n{3,}", "\n\n", text)
|
| 40 |
text = re.sub(r"[^\n#\-\*\w\s\.,:\(\)]+", "", text)
|
| 41 |
return text.strip()
|
| 42 |
|
| 43 |
-
|
| 44 |
def estimate_tokens(text: str) -> int:
|
| 45 |
return len(text) // 3.5 + 1
|
| 46 |
|
| 47 |
-
|
| 48 |
def extract_text_from_excel(file_path: str) -> str:
|
| 49 |
all_text = []
|
| 50 |
xls = pd.ExcelFile(file_path)
|
|
@@ -55,7 +52,6 @@ def extract_text_from_excel(file_path: str) -> str:
|
|
| 55 |
all_text.extend(sheet_text)
|
| 56 |
return "\n".join(all_text)
|
| 57 |
|
| 58 |
-
|
| 59 |
def split_text_into_chunks(text: str, max_tokens: int = MAX_CHUNK_TOKENS, max_chunks: int = 30) -> List[str]:
|
| 60 |
effective_max = max_tokens - PROMPT_OVERHEAD
|
| 61 |
lines, chunks, curr_chunk, curr_tokens = text.split("\n"), [], [], 0
|
|
@@ -74,7 +70,6 @@ def split_text_into_chunks(text: str, max_tokens: int = MAX_CHUNK_TOKENS, max_ch
|
|
| 74 |
chunks.append("\n".join(curr_chunk))
|
| 75 |
return chunks
|
| 76 |
|
| 77 |
-
|
| 78 |
def build_prompt_from_text(chunk: str) -> str:
|
| 79 |
return f"""
|
| 80 |
### Unstructured Clinical Records
|
|
@@ -94,7 +89,6 @@ Analyze the following clinical notes and provide a detailed, concise summary foc
|
|
| 94 |
Respond in well-structured bullet points with medical reasoning.
|
| 95 |
"""
|
| 96 |
|
| 97 |
-
|
| 98 |
def init_agent():
|
| 99 |
tool_path = os.path.join(tool_cache_dir, "new_tool.json")
|
| 100 |
if not os.path.exists(tool_path):
|
|
@@ -111,7 +105,6 @@ def init_agent():
|
|
| 111 |
agent.init_model()
|
| 112 |
return agent
|
| 113 |
|
| 114 |
-
|
| 115 |
def process_final_report(agent, file, chatbot_state: List[Dict[str, str]]) -> Tuple[List[Dict[str, str]], Union[str, None]]:
|
| 116 |
messages = chatbot_state if chatbot_state else []
|
| 117 |
if file is None or not hasattr(file, "name"):
|
|
@@ -140,7 +133,7 @@ def process_final_report(agent, file, chatbot_state: List[Dict[str, str]]) -> Tu
|
|
| 140 |
response += r.content
|
| 141 |
return i, clean_response(response)
|
| 142 |
|
| 143 |
-
with ThreadPoolExecutor(max_workers=
|
| 144 |
futures = [executor.submit(analyze_chunk, i, c) for i, c in enumerate(chunks)]
|
| 145 |
for f in as_completed(futures):
|
| 146 |
i, result = f.result()
|
|
@@ -173,7 +166,6 @@ def process_final_report(agent, file, chatbot_state: List[Dict[str, str]]) -> Tu
|
|
| 173 |
messages.append({"role": "assistant", "content": f"✅ Report generated and saved: {os.path.basename(report_path)}"})
|
| 174 |
return messages, report_path
|
| 175 |
|
| 176 |
-
|
| 177 |
def create_ui(agent):
|
| 178 |
with gr.Blocks(css="""
|
| 179 |
body {
|
|
@@ -233,7 +225,6 @@ Upload clinical Excel records below and click **Analyze** to generate a medical
|
|
| 233 |
|
| 234 |
return demo
|
| 235 |
|
| 236 |
-
|
| 237 |
if __name__ == "__main__":
|
| 238 |
try:
|
| 239 |
agent = init_agent()
|
|
|
|
| 33 |
MAX_NEW_TOKENS = 2048
|
| 34 |
PROMPT_OVERHEAD = 500
|
| 35 |
|
|
|
|
| 36 |
def clean_response(text: str) -> str:
|
| 37 |
text = re.sub(r"\[.*?\]|\bNone\b", "", text, flags=re.DOTALL)
|
| 38 |
text = re.sub(r"\n{3,}", "\n\n", text)
|
| 39 |
text = re.sub(r"[^\n#\-\*\w\s\.,:\(\)]+", "", text)
|
| 40 |
return text.strip()
|
| 41 |
|
|
|
|
| 42 |
def estimate_tokens(text: str) -> int:
|
| 43 |
return len(text) // 3.5 + 1
|
| 44 |
|
|
|
|
| 45 |
def extract_text_from_excel(file_path: str) -> str:
|
| 46 |
all_text = []
|
| 47 |
xls = pd.ExcelFile(file_path)
|
|
|
|
| 52 |
all_text.extend(sheet_text)
|
| 53 |
return "\n".join(all_text)
|
| 54 |
|
|
|
|
| 55 |
def split_text_into_chunks(text: str, max_tokens: int = MAX_CHUNK_TOKENS, max_chunks: int = 30) -> List[str]:
|
| 56 |
effective_max = max_tokens - PROMPT_OVERHEAD
|
| 57 |
lines, chunks, curr_chunk, curr_tokens = text.split("\n"), [], [], 0
|
|
|
|
| 70 |
chunks.append("\n".join(curr_chunk))
|
| 71 |
return chunks
|
| 72 |
|
|
|
|
| 73 |
def build_prompt_from_text(chunk: str) -> str:
|
| 74 |
return f"""
|
| 75 |
### Unstructured Clinical Records
|
|
|
|
| 89 |
Respond in well-structured bullet points with medical reasoning.
|
| 90 |
"""
|
| 91 |
|
|
|
|
| 92 |
def init_agent():
|
| 93 |
tool_path = os.path.join(tool_cache_dir, "new_tool.json")
|
| 94 |
if not os.path.exists(tool_path):
|
|
|
|
| 105 |
agent.init_model()
|
| 106 |
return agent
|
| 107 |
|
|
|
|
| 108 |
def process_final_report(agent, file, chatbot_state: List[Dict[str, str]]) -> Tuple[List[Dict[str, str]], Union[str, None]]:
|
| 109 |
messages = chatbot_state if chatbot_state else []
|
| 110 |
if file is None or not hasattr(file, "name"):
|
|
|
|
| 133 |
response += r.content
|
| 134 |
return i, clean_response(response)
|
| 135 |
|
| 136 |
+
with ThreadPoolExecutor(max_workers=1) as executor:
|
| 137 |
futures = [executor.submit(analyze_chunk, i, c) for i, c in enumerate(chunks)]
|
| 138 |
for f in as_completed(futures):
|
| 139 |
i, result = f.result()
|
|
|
|
| 166 |
messages.append({"role": "assistant", "content": f"✅ Report generated and saved: {os.path.basename(report_path)}"})
|
| 167 |
return messages, report_path
|
| 168 |
|
|
|
|
| 169 |
def create_ui(agent):
|
| 170 |
with gr.Blocks(css="""
|
| 171 |
body {
|
|
|
|
| 225 |
|
| 226 |
return demo
|
| 227 |
|
|
|
|
| 228 |
if __name__ == "__main__":
|
| 229 |
try:
|
| 230 |
agent = init_agent()
|