Update app.py
Browse files
app.py
CHANGED
@@ -13,6 +13,7 @@ import gradio as gr
|
|
13 |
import torch
|
14 |
import matplotlib.pyplot as plt
|
15 |
from fpdf import FPDF
|
|
|
16 |
|
17 |
# === Configuration ===
|
18 |
persistent_dir = "/data/hf_cache"
|
@@ -223,17 +224,13 @@ Avoid repeating the same points multiple times.
|
|
223 |
final_response = remove_duplicate_paragraphs(final_response)
|
224 |
return final_response
|
225 |
|
226 |
-
def
|
227 |
-
|
228 |
-
|
229 |
-
def clean_for_pdf(text):
|
230 |
-
# Remove emojis and any non-latin characters
|
231 |
-
return ''.join(c for c in text if unicodedata.category(c)[0] != 'So')
|
232 |
|
|
|
233 |
chart_dir = os.path.join(os.path.dirname(report_path), "charts")
|
234 |
os.makedirs(chart_dir, exist_ok=True)
|
235 |
|
236 |
-
# Dummy chart
|
237 |
chart_path = os.path.join(chart_dir, "summary_chart.png")
|
238 |
categories = ['Diagnostics', 'Medications', 'Missed', 'Inconsistencies', 'Follow-up']
|
239 |
values = [4, 2, 3, 1, 5]
|
@@ -244,7 +241,6 @@ def generate_pdf_report_with_charts(summary: str, report_path: str):
|
|
244 |
plt.savefig(chart_path)
|
245 |
plt.close()
|
246 |
|
247 |
-
# PDF report
|
248 |
pdf_path = report_path.replace('.md', '.pdf')
|
249 |
pdf = FPDF()
|
250 |
pdf.add_page()
|
@@ -253,7 +249,8 @@ def generate_pdf_report_with_charts(summary: str, report_path: str):
|
|
253 |
pdf.ln(5)
|
254 |
|
255 |
for line in summary.split("\n"):
|
256 |
-
|
|
|
257 |
|
258 |
pdf.ln(10)
|
259 |
pdf.image(chart_path, w=150)
|
@@ -264,31 +261,43 @@ def process_report(agent, file, messages: List[Dict[str, str]]) -> Tuple[List[Di
|
|
264 |
if not file or not hasattr(file, "name"):
|
265 |
messages.append({"role": "assistant", "content": "β Please upload a valid file."})
|
266 |
return messages, None
|
|
|
267 |
start_time = time.time()
|
268 |
messages.append({"role": "user", "content": f"π Processing file: {os.path.basename(file.name)}"})
|
|
|
269 |
try:
|
270 |
extracted = extract_text(file.name)
|
271 |
if not extracted:
|
272 |
messages.append({"role": "assistant", "content": "β Could not extract text."})
|
273 |
return messages, None
|
|
|
274 |
chunks = split_text(extracted)
|
275 |
batches = batch_chunks(chunks, batch_size=BATCH_SIZE)
|
276 |
messages.append({"role": "assistant", "content": f"π Split into {len(batches)} batches. Analyzing..."})
|
|
|
277 |
batch_results = analyze_batches(agent, batches)
|
278 |
valid = [res for res in batch_results if not res.startswith("β")]
|
|
|
279 |
if not valid:
|
280 |
messages.append({"role": "assistant", "content": "β No valid batch outputs."})
|
281 |
return messages, None
|
|
|
282 |
summary = generate_final_summary(agent, "\n\n".join(valid))
|
|
|
283 |
report_path = os.path.join(report_dir, f"report_{datetime.now().strftime('%Y%m%d_%H%M%S')}.md")
|
284 |
with open(report_path, 'w', encoding='utf-8') as f:
|
285 |
f.write(f"# Final Medical Report\n\n{summary}")
|
|
|
286 |
pdf_path = generate_pdf_report_with_charts(summary, report_path)
|
|
|
287 |
end_time = time.time()
|
288 |
elapsed_time = end_time - start_time
|
|
|
289 |
messages.append({"role": "assistant", "content": f"π **Final Report:**\n\n{summary}"})
|
290 |
messages.append({"role": "assistant", "content": f"β
Report generated in **{elapsed_time:.2f} seconds**.\n\nπ₯ PDF report ready: {os.path.basename(pdf_path)}"})
|
|
|
291 |
return messages, pdf_path
|
|
|
292 |
except Exception as e:
|
293 |
messages.append({"role": "assistant", "content": f"β Error: {str(e)}"})
|
294 |
return messages, None
|
@@ -302,22 +311,27 @@ def create_ui(agent):
|
|
302 |
.gr-file, .gr-button { width: 100% !important; max-width: 400px; }
|
303 |
""") as demo:
|
304 |
gr.Markdown("""
|
305 |
-
<h2 style=
|
306 |
-
<p style=
|
307 |
""")
|
|
|
308 |
with gr.Column():
|
309 |
chatbot = gr.Chatbot(label="π§ CPS Assistant", height=480, type="messages")
|
310 |
upload = gr.File(label="π Upload Medical File", file_types=[".xlsx", ".csv", ".pdf"])
|
311 |
analyze = gr.Button("π§ Analyze")
|
312 |
download = gr.File(label="π₯ Download Report", visible=False, interactive=False)
|
|
|
313 |
state = gr.State(value=[])
|
|
|
314 |
def handle_analysis(file, chat):
|
315 |
messages, report_path = process_report(agent, file, chat)
|
316 |
return messages, gr.update(visible=bool(report_path), value=report_path), messages
|
|
|
317 |
analyze.click(fn=handle_analysis, inputs=[upload, state], outputs=[chatbot, download, state])
|
|
|
318 |
return demo
|
319 |
|
320 |
if __name__ == "__main__":
|
321 |
agent = init_agent()
|
322 |
ui = create_ui(agent)
|
323 |
-
ui.launch(server_name="0.0.0.0", server_port=7860, allowed_paths=["/data/hf_cache/reports"], share=False)
|
|
|
13 |
import torch
|
14 |
import matplotlib.pyplot as plt
|
15 |
from fpdf import FPDF
|
16 |
+
import unicodedata
|
17 |
|
18 |
# === Configuration ===
|
19 |
persistent_dir = "/data/hf_cache"
|
|
|
224 |
final_response = remove_duplicate_paragraphs(final_response)
|
225 |
return final_response
|
226 |
|
227 |
+
def remove_non_ascii(text):
|
228 |
+
return unicodedata.normalize('NFKD', text).encode('ascii', 'ignore').decode('ascii')
|
|
|
|
|
|
|
|
|
229 |
|
230 |
+
def generate_pdf_report_with_charts(summary: str, report_path: str):
|
231 |
chart_dir = os.path.join(os.path.dirname(report_path), "charts")
|
232 |
os.makedirs(chart_dir, exist_ok=True)
|
233 |
|
|
|
234 |
chart_path = os.path.join(chart_dir, "summary_chart.png")
|
235 |
categories = ['Diagnostics', 'Medications', 'Missed', 'Inconsistencies', 'Follow-up']
|
236 |
values = [4, 2, 3, 1, 5]
|
|
|
241 |
plt.savefig(chart_path)
|
242 |
plt.close()
|
243 |
|
|
|
244 |
pdf_path = report_path.replace('.md', '.pdf')
|
245 |
pdf = FPDF()
|
246 |
pdf.add_page()
|
|
|
249 |
pdf.ln(5)
|
250 |
|
251 |
for line in summary.split("\n"):
|
252 |
+
clean_line = remove_non_ascii(line)
|
253 |
+
pdf.multi_cell(0, 10, txt=clean_line)
|
254 |
|
255 |
pdf.ln(10)
|
256 |
pdf.image(chart_path, w=150)
|
|
|
261 |
if not file or not hasattr(file, "name"):
|
262 |
messages.append({"role": "assistant", "content": "β Please upload a valid file."})
|
263 |
return messages, None
|
264 |
+
|
265 |
start_time = time.time()
|
266 |
messages.append({"role": "user", "content": f"π Processing file: {os.path.basename(file.name)}"})
|
267 |
+
|
268 |
try:
|
269 |
extracted = extract_text(file.name)
|
270 |
if not extracted:
|
271 |
messages.append({"role": "assistant", "content": "β Could not extract text."})
|
272 |
return messages, None
|
273 |
+
|
274 |
chunks = split_text(extracted)
|
275 |
batches = batch_chunks(chunks, batch_size=BATCH_SIZE)
|
276 |
messages.append({"role": "assistant", "content": f"π Split into {len(batches)} batches. Analyzing..."})
|
277 |
+
|
278 |
batch_results = analyze_batches(agent, batches)
|
279 |
valid = [res for res in batch_results if not res.startswith("β")]
|
280 |
+
|
281 |
if not valid:
|
282 |
messages.append({"role": "assistant", "content": "β No valid batch outputs."})
|
283 |
return messages, None
|
284 |
+
|
285 |
summary = generate_final_summary(agent, "\n\n".join(valid))
|
286 |
+
|
287 |
report_path = os.path.join(report_dir, f"report_{datetime.now().strftime('%Y%m%d_%H%M%S')}.md")
|
288 |
with open(report_path, 'w', encoding='utf-8') as f:
|
289 |
f.write(f"# Final Medical Report\n\n{summary}")
|
290 |
+
|
291 |
pdf_path = generate_pdf_report_with_charts(summary, report_path)
|
292 |
+
|
293 |
end_time = time.time()
|
294 |
elapsed_time = end_time - start_time
|
295 |
+
|
296 |
messages.append({"role": "assistant", "content": f"π **Final Report:**\n\n{summary}"})
|
297 |
messages.append({"role": "assistant", "content": f"β
Report generated in **{elapsed_time:.2f} seconds**.\n\nπ₯ PDF report ready: {os.path.basename(pdf_path)}"})
|
298 |
+
|
299 |
return messages, pdf_path
|
300 |
+
|
301 |
except Exception as e:
|
302 |
messages.append({"role": "assistant", "content": f"β Error: {str(e)}"})
|
303 |
return messages, None
|
|
|
311 |
.gr-file, .gr-button { width: 100% !important; max-width: 400px; }
|
312 |
""") as demo:
|
313 |
gr.Markdown("""
|
314 |
+
<h2 style='text-align:center;'>π CPS: Clinical Patient Support System</h2>
|
315 |
+
<p style='text-align:center;'>Analyze and summarize unstructured medical files using AI (optimized for A100 GPU).</p>
|
316 |
""")
|
317 |
+
|
318 |
with gr.Column():
|
319 |
chatbot = gr.Chatbot(label="π§ CPS Assistant", height=480, type="messages")
|
320 |
upload = gr.File(label="π Upload Medical File", file_types=[".xlsx", ".csv", ".pdf"])
|
321 |
analyze = gr.Button("π§ Analyze")
|
322 |
download = gr.File(label="π₯ Download Report", visible=False, interactive=False)
|
323 |
+
|
324 |
state = gr.State(value=[])
|
325 |
+
|
326 |
def handle_analysis(file, chat):
|
327 |
messages, report_path = process_report(agent, file, chat)
|
328 |
return messages, gr.update(visible=bool(report_path), value=report_path), messages
|
329 |
+
|
330 |
analyze.click(fn=handle_analysis, inputs=[upload, state], outputs=[chatbot, download, state])
|
331 |
+
|
332 |
return demo
|
333 |
|
334 |
if __name__ == "__main__":
|
335 |
agent = init_agent()
|
336 |
ui = create_ui(agent)
|
337 |
+
ui.launch(server_name="0.0.0.0", server_port=7860, allowed_paths=["/data/hf_cache/reports"], share=False)
|