Update app.py
Browse files
app.py
CHANGED
@@ -32,6 +32,7 @@ sys.path.insert(0, src_path)
|
|
32 |
|
33 |
from txagent.txagent import TxAgent
|
34 |
|
|
|
35 |
MAX_MODEL_TOKENS = 32768
|
36 |
MAX_CHUNK_TOKENS = 8192
|
37 |
MAX_NEW_TOKENS = 2048
|
@@ -68,26 +69,19 @@ def split_text_into_chunks(text: str, max_tokens: int = MAX_CHUNK_TOKENS) -> Lis
|
|
68 |
effective_max_tokens = max_tokens - PROMPT_OVERHEAD
|
69 |
if effective_max_tokens <= 0:
|
70 |
raise ValueError(f"Effective max tokens ({effective_max_tokens}) must be positive.")
|
71 |
-
|
72 |
lines = text.split("\n")
|
73 |
-
chunks = []
|
74 |
-
current_chunk = []
|
75 |
-
current_tokens = 0
|
76 |
-
|
77 |
for line in lines:
|
78 |
line_tokens = estimate_tokens(line)
|
79 |
if current_tokens + line_tokens > effective_max_tokens:
|
80 |
if current_chunk:
|
81 |
chunks.append("\n".join(current_chunk))
|
82 |
-
current_chunk = [line]
|
83 |
-
current_tokens = line_tokens
|
84 |
else:
|
85 |
current_chunk.append(line)
|
86 |
current_tokens += line_tokens
|
87 |
-
|
88 |
if current_chunk:
|
89 |
chunks.append("\n".join(current_chunk))
|
90 |
-
|
91 |
return chunks
|
92 |
|
93 |
def build_prompt_from_text(chunk: str) -> str:
|
@@ -113,10 +107,8 @@ Please analyze the above and provide:
|
|
113 |
def init_agent():
|
114 |
default_tool_path = os.path.abspath("data/new_tool.json")
|
115 |
target_tool_path = os.path.join(tool_cache_dir, "new_tool.json")
|
116 |
-
|
117 |
if not os.path.exists(target_tool_path):
|
118 |
shutil.copy(default_tool_path, target_tool_path)
|
119 |
-
|
120 |
agent = TxAgent(
|
121 |
model_name="mims-harvard/TxAgent-T1-Llama-3.1-8B",
|
122 |
rag_model_name="mims-harvard/ToolRAG-T1-GTE-Qwen2-1.5B",
|
@@ -141,7 +133,6 @@ def process_final_report(agent, file, chatbot_state: List[Dict[str, str]]) -> Tu
|
|
141 |
try:
|
142 |
messages.append({"role": "user", "content": f"Processing Excel file: {os.path.basename(file.name)}"})
|
143 |
messages.append({"role": "assistant", "content": "⏳ Extracting and analyzing data..."})
|
144 |
-
|
145 |
extracted_text = extract_text_from_excel(file.name)
|
146 |
chunks = split_text_into_chunks(extracted_text)
|
147 |
chunk_responses = [None] * len(chunks)
|
@@ -219,7 +210,6 @@ def process_final_report(agent, file, chatbot_state: List[Dict[str, str]]) -> Tu
|
|
219 |
except Exception as e:
|
220 |
messages.append({"role": "assistant", "content": f"❌ Error summarizing intermediate results: {str(e)}"})
|
221 |
return messages, report_path
|
222 |
-
|
223 |
summary += f"\n\n### Chunk {i+1} Analysis\n{response}"
|
224 |
current_summary_tokens += response_tokens
|
225 |
|
@@ -268,7 +258,7 @@ def process_final_report(agent, file, chatbot_state: List[Dict[str, str]]) -> Tu
|
|
268 |
def create_ui(agent):
|
269 |
with gr.Blocks(
|
270 |
title="Patient History Chat",
|
271 |
-
css
|
272 |
.gradio-container {
|
273 |
max-width: 900px !important;
|
274 |
margin: auto;
|
@@ -303,24 +293,27 @@ def create_ui(agent):
|
|
303 |
padding-left: 1.2em;
|
304 |
margin: 0.4em 0;
|
305 |
}
|
306 |
-
|
307 |
) as demo:
|
308 |
-
gr.Markdown(
|
|
|
|
|
|
|
309 |
|
310 |
with gr.Row():
|
311 |
with gr.Column(scale=3):
|
312 |
chatbot = gr.Chatbot(
|
313 |
-
label
|
314 |
show_copy_button=True,
|
315 |
height=600,
|
316 |
-
type
|
317 |
-
avatar_images=(None,
|
318 |
render_markdown=True
|
319 |
)
|
320 |
with gr.Column(scale=1):
|
321 |
-
file_upload = gr.File(label
|
322 |
-
analyze_btn = gr.Button(
|
323 |
-
report_output = gr.File(label
|
324 |
|
325 |
chatbot_state = gr.State(value=[])
|
326 |
|
@@ -328,24 +321,24 @@ def create_ui(agent):
|
|
328 |
messages, report_path = process_final_report(agent, file, current_state)
|
329 |
formatted_messages = []
|
330 |
for msg in messages:
|
331 |
-
role = msg.get(
|
332 |
-
content = msg.get(
|
333 |
-
if role ==
|
334 |
-
content = content.replace(
|
335 |
-
content = f
|
336 |
-
formatted_messages.append({
|
337 |
report_update = gr.update(visible=report_path is not None, value=report_path)
|
338 |
return formatted_messages, report_update, formatted_messages
|
339 |
|
340 |
-
analyze_btn.click(fn=update_ui, inputs=[file_upload, chatbot_state], outputs=[chatbot, report_output, chatbot_state], api_name
|
341 |
|
342 |
return demo
|
343 |
|
344 |
-
if __name__ ==
|
345 |
try:
|
346 |
agent = init_agent()
|
347 |
demo = create_ui(agent)
|
348 |
-
demo.launch(server_name
|
349 |
except Exception as e:
|
350 |
-
print(f
|
351 |
sys.exit(1)
|
|
|
32 |
|
33 |
from txagent.txagent import TxAgent
|
34 |
|
35 |
+
# Constants
|
36 |
MAX_MODEL_TOKENS = 32768
|
37 |
MAX_CHUNK_TOKENS = 8192
|
38 |
MAX_NEW_TOKENS = 2048
|
|
|
69 |
effective_max_tokens = max_tokens - PROMPT_OVERHEAD
|
70 |
if effective_max_tokens <= 0:
|
71 |
raise ValueError(f"Effective max tokens ({effective_max_tokens}) must be positive.")
|
|
|
72 |
lines = text.split("\n")
|
73 |
+
chunks, current_chunk, current_tokens = [], [], 0
|
|
|
|
|
|
|
74 |
for line in lines:
|
75 |
line_tokens = estimate_tokens(line)
|
76 |
if current_tokens + line_tokens > effective_max_tokens:
|
77 |
if current_chunk:
|
78 |
chunks.append("\n".join(current_chunk))
|
79 |
+
current_chunk, current_tokens = [line], line_tokens
|
|
|
80 |
else:
|
81 |
current_chunk.append(line)
|
82 |
current_tokens += line_tokens
|
|
|
83 |
if current_chunk:
|
84 |
chunks.append("\n".join(current_chunk))
|
|
|
85 |
return chunks
|
86 |
|
87 |
def build_prompt_from_text(chunk: str) -> str:
|
|
|
107 |
def init_agent():
|
108 |
default_tool_path = os.path.abspath("data/new_tool.json")
|
109 |
target_tool_path = os.path.join(tool_cache_dir, "new_tool.json")
|
|
|
110 |
if not os.path.exists(target_tool_path):
|
111 |
shutil.copy(default_tool_path, target_tool_path)
|
|
|
112 |
agent = TxAgent(
|
113 |
model_name="mims-harvard/TxAgent-T1-Llama-3.1-8B",
|
114 |
rag_model_name="mims-harvard/ToolRAG-T1-GTE-Qwen2-1.5B",
|
|
|
133 |
try:
|
134 |
messages.append({"role": "user", "content": f"Processing Excel file: {os.path.basename(file.name)}"})
|
135 |
messages.append({"role": "assistant", "content": "⏳ Extracting and analyzing data..."})
|
|
|
136 |
extracted_text = extract_text_from_excel(file.name)
|
137 |
chunks = split_text_into_chunks(extracted_text)
|
138 |
chunk_responses = [None] * len(chunks)
|
|
|
210 |
except Exception as e:
|
211 |
messages.append({"role": "assistant", "content": f"❌ Error summarizing intermediate results: {str(e)}"})
|
212 |
return messages, report_path
|
|
|
213 |
summary += f"\n\n### Chunk {i+1} Analysis\n{response}"
|
214 |
current_summary_tokens += response_tokens
|
215 |
|
|
|
258 |
def create_ui(agent):
|
259 |
with gr.Blocks(
|
260 |
title="Patient History Chat",
|
261 |
+
css="""
|
262 |
.gradio-container {
|
263 |
max-width: 900px !important;
|
264 |
margin: auto;
|
|
|
293 |
padding-left: 1.2em;
|
294 |
margin: 0.4em 0;
|
295 |
}
|
296 |
+
"""
|
297 |
) as demo:
|
298 |
+
gr.Markdown("""
|
299 |
+
<h2 style='color:#182848'>🏥 Patient History Analysis Tool</h2>
|
300 |
+
<p style='color:#444;'>Upload an Excel file containing clinical data. The assistant will analyze it for patterns, inconsistencies, and recommendations.</p>
|
301 |
+
""")
|
302 |
|
303 |
with gr.Row():
|
304 |
with gr.Column(scale=3):
|
305 |
chatbot = gr.Chatbot(
|
306 |
+
label="Clinical Assistant",
|
307 |
show_copy_button=True,
|
308 |
height=600,
|
309 |
+
type="messages",
|
310 |
+
avatar_images=(None, "https://i.imgur.com/6wX7Zb4.png"),
|
311 |
render_markdown=True
|
312 |
)
|
313 |
with gr.Column(scale=1):
|
314 |
+
file_upload = gr.File(label="Upload Excel File", file_types=[".xlsx"], height=100)
|
315 |
+
analyze_btn = gr.Button("🧠 Analyze Patient History", variant="primary", elem_classes="primary")
|
316 |
+
report_output = gr.File(label="Download Report", visible=False, interactive=False)
|
317 |
|
318 |
chatbot_state = gr.State(value=[])
|
319 |
|
|
|
321 |
messages, report_path = process_final_report(agent, file, current_state)
|
322 |
formatted_messages = []
|
323 |
for msg in messages:
|
324 |
+
role = msg.get("role")
|
325 |
+
content = msg.get("content", "")
|
326 |
+
if role == "assistant":
|
327 |
+
content = content.replace("- ", "\n- ")
|
328 |
+
content = f"<div class='chat-message-content'>{content}</div>"
|
329 |
+
formatted_messages.append({"role": role, "content": content})
|
330 |
report_update = gr.update(visible=report_path is not None, value=report_path)
|
331 |
return formatted_messages, report_update, formatted_messages
|
332 |
|
333 |
+
analyze_btn.click(fn=update_ui, inputs=[file_upload, chatbot_state], outputs=[chatbot, report_output, chatbot_state], api_name="analyze")
|
334 |
|
335 |
return demo
|
336 |
|
337 |
+
if __name__ == "__main__":
|
338 |
try:
|
339 |
agent = init_agent()
|
340 |
demo = create_ui(agent)
|
341 |
+
demo.launch(server_name="0.0.0.0", server_port=7860, show_error=True, allowed_paths=["/data/hf_cache/reports"], share=False)
|
342 |
except Exception as e:
|
343 |
+
print(f"Error: {str(e)}")
|
344 |
sys.exit(1)
|