Update app.py

app.py

@@ -177,7 +177,6 @@ def phonetic_match(text, query, method='levenshtein_distance', apply_phonetic=Tr
         return jellyfish.levenshtein_distance(text_phonetic, query_phonetic)
     return 0
 
-#def optimize_query(query, llm_model):
 def optimize_query(
     query: str, 
     llm_model: str = "meta-llama/Llama-3.2-1B", 
@@ -185,10 +184,10 @@ def optimize_query(
     embedding_model: str = "sentence-transformers/all-MiniLM-L6-v2", 
     vector_store_type: str = "faiss", 
     search_type: str = "similarity", 
-    top_k: int = 5
+    top_k: int = 3  # Reduce top_k for quicker test
 ) -> List[str]:
     # Initialize the language model
-    #llm = HuggingFacePipeline(model=llm_model)
+    #llm = HuggingFacePipeline(pipeline(model=llm_model))
 
     # Create a temporary vector store for query optimization
     temp_vector_store = get_vector_store(vector_store_type, chunks, embedding_model)
@@ -202,10 +201,11 @@ def optimize_query(
         llm=llm
     )
 
-    # Use a NoOpRunManager as the run manager
-    optimized_queries = multi_query_retriever.invoke(query)
+    # Limit max time or set a timeout for LLM to avoid endless execution
+    optimized_queries = multi_query_retriever.invoke(query, max_time=30)  # Timeout in seconds
+
+    return optimized_queries
 
-    return optimized_queries 
     
 
 def create_custom_embedding(texts, model_type='word2vec', vector_size=100, window=5, min_count=1):