storage_context = StorageContext.from_defaults(
    docstore=RedisDocumentStore.from_host_and_port(
        host=REDIS_HOST, port=REDIS_PORT, namespace="summary/product/11"
    ),
    index_store=RedisIndexStore.from_host_and_port(
        host=REDIS_HOST, port=REDIS_PORT, namespace="summary/product/11"
    ),
)
summary_index = SummaryIndex(nodes, storage_context=storage_context)

vertex_gemini = Vertex(
    model="gemini-1.5-pro",
    context_window=100000,
    temperature=0,
    additional_kwargs={},
)

# setup the index/query process, ie the embedding model (and completion if used)
Settings.llm = vertex_gemini


# setup the index/query process, ie the embedding model (and completion if used)
Settings.embed_model = embed_model


# Create a Pinecone-based vector store for detail
detail_vector_store = PineconeVectorStore(
    pinecone_index=pinecone_index,
    namespace=DETAIL_NAMESPACE
)
detail_storage_context = StorageContext.from_defaults(vector_store=detail_vector_store)
vector_index = VectorStoreIndex.from_vector_store(
    vector_store=detail_vector_store,
    storage_context=detail_storage_context
)
print("VectorStoreIndex (detail) created.")

detail_query_engine = vector_index.as_query_engine()
detail_tool = QueryEngineTool.from_defaults(
    query_engine=detail_query_engine,
    name="detail_vector_index",
    description="Use this to answer questions about the full text of the reviews."
)

agent = ReActAgent.from_tools(tools=[detail_tool], verbose=True)

@app.post("/chat-stream")
def chat_endpoint(chat_request: ChatRequest):
    user_query = chat_request.query

    def token_generator():
        response = agent.stream_chat(user_query)
        for delta in response.response_gen:
            print(delta)
            yield delta




    return StreamingResponse(token_generator(), media_type="text/plain")