debug: add prints in new_question to trace call and LLM progress

app.py

@@ -269,14 +269,19 @@ def load_pdf(pdf_file, language):
 
 @spaces.GPU(duration=180)
 def new_question(chunks, language):
+    print(f"[new_question] called — {len(chunks)} chunks, lang={language}")
     s = UI[language]
     if not chunks:
+        print("[new_question] no chunks, returning early")
         return s["no_pdf"], "", s["no_pdf"]
     try:
         chunk = random.choice(chunks)
+        print(f"[new_question] chunk selected ({len(chunk)} chars), calling LLM…")
         question = generate_question(chunk, language=language)
+        print(f"[new_question] question generated: {question[:80]!r}")
         return question, chunk, ""
     except Exception as e:
+        import traceback; traceback.print_exc()
         err = f"Erreur : {e}"
         return err, "", err