Update app.py

app.py

@@ -28,6 +28,9 @@ client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.2")
 # --- RAG pipeline ---
 def generate_answer(messages):
     try:
+        sys.stderr.write("=== generate_answer called ===\n")
+        sys.stderr.flush()
+        
         user_query = messages[-1]["content"]
         query_embedding = embedding_model.encode([user_query], convert_to_numpy=True)
         faiss.normalize_L2(query_embedding)