Update app.py
Browse files
app.py
CHANGED
@@ -1,14 +1,16 @@
|
|
1 |
import sys
|
2 |
import os
|
3 |
import pandas as pd
|
4 |
-
import json
|
5 |
import gradio as gr
|
6 |
from typing import List, Tuple, Dict, Any, Union
|
7 |
-
import hashlib
|
8 |
import shutil
|
9 |
import re
|
10 |
from datetime import datetime
|
11 |
import time
|
|
|
|
|
|
|
|
|
12 |
|
13 |
# Configuration and setup
|
14 |
persistent_dir = "/data/hf_cache"
|
@@ -32,10 +34,22 @@ sys.path.insert(0, src_path)
|
|
32 |
from txagent.txagent import TxAgent
|
33 |
|
34 |
# Constants
|
35 |
-
MAX_MODEL_TOKENS =
|
36 |
-
MAX_CHUNK_TOKENS =
|
37 |
-
MAX_NEW_TOKENS =
|
38 |
-
PROMPT_OVERHEAD = 500
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
39 |
|
40 |
def clean_response(text: str) -> str:
|
41 |
try:
|
@@ -48,8 +62,10 @@ def clean_response(text: str) -> str:
|
|
48 |
return text.strip()
|
49 |
|
50 |
def estimate_tokens(text: str) -> int:
|
51 |
-
"""Estimate
|
52 |
-
|
|
|
|
|
53 |
|
54 |
def extract_text_from_excel(file_path: str) -> str:
|
55 |
"""Extract text from all sheets in an Excel file."""
|
@@ -67,10 +83,7 @@ def extract_text_from_excel(file_path: str) -> str:
|
|
67 |
return "\n".join(all_text)
|
68 |
|
69 |
def split_text_into_chunks(text: str, max_tokens: int = MAX_CHUNK_TOKENS) -> List[str]:
|
70 |
-
"""
|
71 |
-
Split text into chunks, ensuring each chunk is within token limits,
|
72 |
-
accounting for prompt overhead.
|
73 |
-
"""
|
74 |
effective_max_tokens = max_tokens - PROMPT_OVERHEAD
|
75 |
if effective_max_tokens <= 0:
|
76 |
raise ValueError(f"Effective max tokens ({effective_max_tokens}) must be positive.")
|
@@ -83,7 +96,7 @@ def split_text_into_chunks(text: str, max_tokens: int = MAX_CHUNK_TOKENS) -> Lis
|
|
83 |
for line in lines:
|
84 |
line_tokens = estimate_tokens(line)
|
85 |
if current_tokens + line_tokens > effective_max_tokens:
|
86 |
-
if current_chunk:
|
87 |
chunks.append("\n".join(current_chunk))
|
88 |
current_chunk = [line]
|
89 |
current_tokens = line_tokens
|
@@ -118,7 +131,7 @@ Please analyze the above and provide:
|
|
118 |
"""
|
119 |
|
120 |
def init_agent():
|
121 |
-
"""Initialize the TxAgent with
|
122 |
default_tool_path = os.path.abspath("data/new_tool.json")
|
123 |
target_tool_path = os.path.join(tool_cache_dir, "new_tool.json")
|
124 |
|
@@ -138,8 +151,47 @@ def init_agent():
|
|
138 |
agent.init_model()
|
139 |
return agent
|
140 |
|
141 |
-
def
|
142 |
-
"""Process
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
143 |
messages = chatbot_state if chatbot_state else []
|
144 |
report_path = None
|
145 |
|
@@ -152,57 +204,43 @@ def process_final_report(agent, file, chatbot_state: List[Dict[str, str]]) -> Tu
|
|
152 |
messages.append({"role": "assistant", "content": "⏳ Extracting and analyzing data..."})
|
153 |
|
154 |
# Extract text and split into chunks
|
|
|
155 |
extracted_text = extract_text_from_excel(file.name)
|
156 |
chunks = split_text_into_chunks(extracted_text, max_tokens=MAX_CHUNK_TOKENS)
|
157 |
-
|
158 |
-
|
159 |
-
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
-
|
169 |
-
|
170 |
-
|
171 |
-
|
172 |
-
|
173 |
-
|
174 |
-
|
175 |
-
|
176 |
-
|
177 |
-
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
-
response += result
|
182 |
-
elif hasattr(result, "content"):
|
183 |
-
response += result.content
|
184 |
-
elif isinstance(result, list):
|
185 |
-
for r in result:
|
186 |
-
if hasattr(r, "content"):
|
187 |
-
response += r.content
|
188 |
-
except Exception as e:
|
189 |
-
messages.append({"role": "assistant", "content": f"❌ Error analyzing chunk {i+1}: {str(e)}"})
|
190 |
-
continue
|
191 |
-
|
192 |
-
chunk_responses.append(clean_response(response))
|
193 |
-
messages.append({"role": "assistant", "content": f"✅ Chunk {i+1} analysis complete"})
|
194 |
-
|
195 |
if not chunk_responses:
|
196 |
messages.append({"role": "assistant", "content": "❌ No valid chunk responses to summarize."})
|
197 |
return messages, report_path
|
198 |
|
199 |
-
# Summarize chunk responses incrementally
|
200 |
summary = ""
|
201 |
current_summary_tokens = 0
|
202 |
for i, response in enumerate(chunk_responses):
|
203 |
response_tokens = estimate_tokens(response)
|
204 |
if current_summary_tokens + response_tokens > MAX_MODEL_TOKENS - PROMPT_OVERHEAD - MAX_NEW_TOKENS:
|
205 |
-
# Summarize current summary
|
206 |
summary_prompt = f"Summarize the following analysis:\n\n{summary}\n\nProvide a concise summary."
|
207 |
summary_response = ""
|
208 |
try:
|
@@ -270,13 +308,15 @@ def process_final_report(agent, file, chatbot_state: List[Dict[str, str]]) -> Tu
|
|
270 |
f.write(final_report)
|
271 |
|
272 |
messages.append({"role": "assistant", "content": f"✅ Report generated and saved: report_{timestamp}.md"})
|
|
|
273 |
|
274 |
except Exception as e:
|
275 |
messages.append({"role": "assistant", "content": f"❌ Error processing file: {str(e)}"})
|
|
|
276 |
|
277 |
return messages, report_path
|
278 |
|
279 |
-
def create_ui(agent):
|
280 |
"""Create the Gradio UI for the patient history analysis tool."""
|
281 |
with gr.Blocks(title="Patient History Chat", css=".gradio-container {max-width: 900px !important}") as demo:
|
282 |
gr.Markdown("## 🏥 Patient History Analysis Tool")
|
@@ -312,10 +352,15 @@ def create_ui(agent):
|
|
312 |
# State to maintain chatbot messages
|
313 |
chatbot_state = gr.State(value=[])
|
314 |
|
315 |
-
def update_ui(file, current_state):
|
316 |
-
messages
|
317 |
-
|
318 |
-
|
|
|
|
|
|
|
|
|
|
|
319 |
|
320 |
analyze_btn.click(
|
321 |
fn=update_ui,
|
@@ -329,7 +374,7 @@ def create_ui(agent):
|
|
329 |
if __name__ == "__main__":
|
330 |
try:
|
331 |
agent = init_agent()
|
332 |
-
demo = create_ui(agent)
|
333 |
demo.launch(
|
334 |
server_name="0.0.0.0",
|
335 |
server_port=7860,
|
|
|
1 |
import sys
|
2 |
import os
|
3 |
import pandas as pd
|
|
|
4 |
import gradio as gr
|
5 |
from typing import List, Tuple, Dict, Any, Union
|
|
|
6 |
import shutil
|
7 |
import re
|
8 |
from datetime import datetime
|
9 |
import time
|
10 |
+
from transformers import AutoTokenizer
|
11 |
+
import asyncio
|
12 |
+
import logging
|
13 |
+
from concurrent.futures import ThreadPoolExecutor, as_completed
|
14 |
|
15 |
# Configuration and setup
|
16 |
persistent_dir = "/data/hf_cache"
|
|
|
34 |
from txagent.txagent import TxAgent
|
35 |
|
36 |
# Constants
|
37 |
+
MAX_MODEL_TOKENS = 131072 # TxAgent's max token limit
|
38 |
+
MAX_CHUNK_TOKENS = 32768 # Larger chunks to reduce number of chunks
|
39 |
+
MAX_NEW_TOKENS = 512 # Optimized for fast generation
|
40 |
+
PROMPT_OVERHEAD = 500 # Estimated tokens for prompt template
|
41 |
+
MAX_CONCURRENT = 8 # High concurrency for A100 80GB
|
42 |
+
|
43 |
+
# Initialize tokenizer for precise token counting
|
44 |
+
try:
|
45 |
+
tokenizer = AutoTokenizer.from_pretrained("mims-harvard/TxAgent-T1-Llama-3.1-8B")
|
46 |
+
except Exception as e:
|
47 |
+
print(f"Warning: Could not load tokenizer, falling back to heuristic: {str(e)}")
|
48 |
+
tokenizer = None
|
49 |
+
|
50 |
+
# Setup logging
|
51 |
+
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
|
52 |
+
logger = logging.getLogger(__name__)
|
53 |
|
54 |
def clean_response(text: str) -> str:
|
55 |
try:
|
|
|
62 |
return text.strip()
|
63 |
|
64 |
def estimate_tokens(text: str) -> int:
|
65 |
+
"""Estimate tokens using tokenizer if available, else fall back to heuristic."""
|
66 |
+
if tokenizer:
|
67 |
+
return len(tokenizer.encode(text, add_special_tokens=False))
|
68 |
+
return len(text) // 3.5 + 1
|
69 |
|
70 |
def extract_text_from_excel(file_path: str) -> str:
|
71 |
"""Extract text from all sheets in an Excel file."""
|
|
|
83 |
return "\n".join(all_text)
|
84 |
|
85 |
def split_text_into_chunks(text: str, max_tokens: int = MAX_CHUNK_TOKENS) -> List[str]:
|
86 |
+
"""Split text into chunks within token limits, accounting for prompt overhead."""
|
|
|
|
|
|
|
87 |
effective_max_tokens = max_tokens - PROMPT_OVERHEAD
|
88 |
if effective_max_tokens <= 0:
|
89 |
raise ValueError(f"Effective max tokens ({effective_max_tokens}) must be positive.")
|
|
|
96 |
for line in lines:
|
97 |
line_tokens = estimate_tokens(line)
|
98 |
if current_tokens + line_tokens > effective_max_tokens:
|
99 |
+
if current_chunk:
|
100 |
chunks.append("\n".join(current_chunk))
|
101 |
current_chunk = [line]
|
102 |
current_tokens = line_tokens
|
|
|
131 |
"""
|
132 |
|
133 |
def init_agent():
|
134 |
+
"""Initialize the TxAgent with optimized vLLM settings for A100 80GB."""
|
135 |
default_tool_path = os.path.abspath("data/new_tool.json")
|
136 |
target_tool_path = os.path.join(tool_cache_dir, "new_tool.json")
|
137 |
|
|
|
151 |
agent.init_model()
|
152 |
return agent
|
153 |
|
154 |
+
async def process_chunk(agent, chunk: str, chunk_index: int, total_chunks: int) -> Tuple[int, str, str]:
|
155 |
+
"""Process a single chunk and return index, response, and status message."""
|
156 |
+
logger.info(f"Processing chunk {chunk_index+1}/{total_chunks}")
|
157 |
+
prompt = build_prompt_from_text(chunk)
|
158 |
+
prompt_tokens = estimate_tokens(prompt)
|
159 |
+
|
160 |
+
if prompt_tokens > MAX_MODEL_TOKENS:
|
161 |
+
error_msg = f"❌ Chunk {chunk_index+1} prompt too long ({prompt_tokens} tokens). Skipping..."
|
162 |
+
logger.warning(error_msg)
|
163 |
+
return chunk_index, "", error_msg
|
164 |
+
|
165 |
+
response = ""
|
166 |
+
try:
|
167 |
+
for result in agent.run_gradio_chat(
|
168 |
+
message=prompt,
|
169 |
+
history=[],
|
170 |
+
temperature=0.2,
|
171 |
+
max_new_tokens=MAX_NEW_TOKENS,
|
172 |
+
max_token=MAX_MODEL_TOKENS,
|
173 |
+
call_agent=False,
|
174 |
+
conversation=[],
|
175 |
+
):
|
176 |
+
if isinstance(result, str):
|
177 |
+
response += result
|
178 |
+
elif hasattr(result, "content"):
|
179 |
+
response += result.content
|
180 |
+
elif isinstance(result, list):
|
181 |
+
for r in result:
|
182 |
+
if hasattr(r, "content"):
|
183 |
+
response += r.content
|
184 |
+
status = f"✅ Chunk {chunk_index+1} analysis complete"
|
185 |
+
logger.info(status)
|
186 |
+
except Exception as e:
|
187 |
+
status = f"❌ Error analyzing chunk {chunk_index+1}: {str(e)}"
|
188 |
+
logger.error(status)
|
189 |
+
response = ""
|
190 |
+
|
191 |
+
return chunk_index, clean_response(response), status
|
192 |
+
|
193 |
+
async def process_final_report(agent, file, chatbot_state: List[Dict[str, str]]) -> Tuple[List[Dict[str, str]], Union[str, None]]:
|
194 |
+
"""Process the Excel file and generate a final report with asynchronous updates."""
|
195 |
messages = chatbot_state if chatbot_state else []
|
196 |
report_path = None
|
197 |
|
|
|
204 |
messages.append({"role": "assistant", "content": "⏳ Extracting and analyzing data..."})
|
205 |
|
206 |
# Extract text and split into chunks
|
207 |
+
start_time = time.time()
|
208 |
extracted_text = extract_text_from_excel(file.name)
|
209 |
chunks = split_text_into_chunks(extracted_text, max_tokens=MAX_CHUNK_TOKENS)
|
210 |
+
logger.info(f"Extracted text and split into {len(chunks)} chunks in {time.time() - start_time:.2f} seconds")
|
211 |
+
|
212 |
+
chunk_responses = [None] * len(chunks)
|
213 |
+
batch_size = MAX_CONCURRENT
|
214 |
+
|
215 |
+
# Process chunks in batches
|
216 |
+
for batch_start in range(0, len(chunks), batch_size):
|
217 |
+
batch_chunks = chunks[batch_start:batch_start + batch_size]
|
218 |
+
batch_indices = list(range(batch_start, min(batch_start + batch_size, len(chunks))))
|
219 |
+
logger.info(f"Processing batch {batch_start//batch_size + 1}/{(len(chunks) + batch_size - 1)//batch_size}")
|
220 |
+
|
221 |
+
with ThreadPoolExecutor(max_workers=MAX_CONCURRENT) as executor:
|
222 |
+
futures = [
|
223 |
+
executor.submit(lambda c, i: asyncio.run(process_chunk(agent, c, i, len(chunks))), chunk, i)
|
224 |
+
for i, chunk in zip(batch_indices, batch_chunks)
|
225 |
+
]
|
226 |
+
for future in as_completed(futures):
|
227 |
+
chunk_index, response, status = future.result()
|
228 |
+
chunk_responses[chunk_index] = response
|
229 |
+
messages.append({"role": "assistant", "content": status})
|
230 |
+
yield messages, None
|
231 |
+
|
232 |
+
# Filter out empty responses
|
233 |
+
chunk_responses = [r for r in chunk_responses if r]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
234 |
if not chunk_responses:
|
235 |
messages.append({"role": "assistant", "content": "❌ No valid chunk responses to summarize."})
|
236 |
return messages, report_path
|
237 |
|
238 |
+
# Summarize chunk responses incrementally
|
239 |
summary = ""
|
240 |
current_summary_tokens = 0
|
241 |
for i, response in enumerate(chunk_responses):
|
242 |
response_tokens = estimate_tokens(response)
|
243 |
if current_summary_tokens + response_tokens > MAX_MODEL_TOKENS - PROMPT_OVERHEAD - MAX_NEW_TOKENS:
|
|
|
244 |
summary_prompt = f"Summarize the following analysis:\n\n{summary}\n\nProvide a concise summary."
|
245 |
summary_response = ""
|
246 |
try:
|
|
|
308 |
f.write(final_report)
|
309 |
|
310 |
messages.append({"role": "assistant", "content": f"✅ Report generated and saved: report_{timestamp}.md"})
|
311 |
+
logger.info(f"Total processing time: {time.time() - start_time:.2f} seconds")
|
312 |
|
313 |
except Exception as e:
|
314 |
messages.append({"role": "assistant", "content": f"❌ Error processing file: {str(e)}"})
|
315 |
+
logger.error(f"Processing failed: {str(e)}")
|
316 |
|
317 |
return messages, report_path
|
318 |
|
319 |
+
async def create_ui(agent):
|
320 |
"""Create the Gradio UI for the patient history analysis tool."""
|
321 |
with gr.Blocks(title="Patient History Chat", css=".gradio-container {max-width: 900px !important}") as demo:
|
322 |
gr.Markdown("## 🏥 Patient History Analysis Tool")
|
|
|
352 |
# State to maintain chatbot messages
|
353 |
chatbot_state = gr.State(value=[])
|
354 |
|
355 |
+
async def update_ui(file, current_state):
|
356 |
+
messages = current_state if current_state else []
|
357 |
+
report_path = None
|
358 |
+
async for new_messages, new_report_path in process_final_report(agent, file, messages):
|
359 |
+
messages = new_messages
|
360 |
+
report_path = new_report_path
|
361 |
+
report_update = gr.update(visible=report_path is not None, value=report_path)
|
362 |
+
yield messages, report_update, messages
|
363 |
+
yield messages, gr.update(visible=report_path is not None, value=report_path), messages
|
364 |
|
365 |
analyze_btn.click(
|
366 |
fn=update_ui,
|
|
|
374 |
if __name__ == "__main__":
|
375 |
try:
|
376 |
agent = init_agent()
|
377 |
+
demo = asyncio.run(create_ui(agent))
|
378 |
demo.launch(
|
379 |
server_name="0.0.0.0",
|
380 |
server_port=7860,
|