Update app.py

app.py

@@ -18,6 +18,10 @@ import spaces
 import threading
 import sys
 import torch
+from llama_index.core.indices.query.query_transform.base import (
+    StepDecomposeQueryTransform,
+)
+
 
 head = """
     <link href="https://cdn.jsdelivr.net/npm/bootstrap@5.3.3/dist/css/bootstrap.min.css" rel="stylesheet" integrity="sha384-QWTKZyjpPEjISv5WaRU9OFeRpok6YctnYmDr5pNlyT2bRjXh0JMhjY6hW+ALEwIH" crossorigin="anonymous">
@@ -126,9 +130,10 @@ memory_bandi=None
 memory_bandi_sistema_puglia=None
 stream_response=None
 divDocumenti=None
-
+llm = None
 
 def main():
+    global llm
     setGPU()
     llm = setLLM()
     Settings.llm = llm
@@ -471,10 +476,14 @@ def main():
             else:
                 print("MODALITA STANDARD")
                 if(str(current_response_mode)=="tree_summarize"):
+                    step_decompose_transform = StepDecomposeQueryTransform(llm, verbose=True)
                     # define response synthesizer
                     response_synthesizer = get_response_synthesizer(streaming=True,response_mode="tree_summarize",text_qa_template=text_qa_template)
                     query_engine = None
                     query_engine = RetrieverQueryEngine(retriever=retriever, response_synthesizer=response_synthesizer)
+                    query_engine = MultiStepQueryEngine(
+                        query_engine, query_transform=step_decompose_transform
+                    )
                     stream_response = None
                     print(userMessage)
                     stream_response = query_engine.query(userMessage)