Hi! I'm trying to compare different retrieval strategies (MODES) using Arize and LlamaIndex. here's a snippet from my code:
for mode in MODES:
query_engine = vector_store_index.as_query_engine(similarity_top_k=5, vector_store_query_mode=mode)
px.close_app()
px.launch_app()
for query in queries[0:2]:
query_engine.query(query)
retrieved_documents = get_retrieved_documents(px.active_session())I'm relauncing Phoenix for each mode but I'm seeing that the length of retrieved_documents is incrementally increasing despite the reset which leads me to believe that it's not actually relaunching. Am I doing something wrong?
