Hi, I have a fairly simple use case

ŁŁukasz

Hi, I have a fairly simple use case - my dataset is small enough to always fit into a query, so I'm always preparing a single manual TextNode and creating a new index with that:

Plain Text

nodes = [TextNode(text=input)]
index = VectorStoreIndex(nodes=nodes, show_progress=True)
query_engine = index.as_query_engine()
prompt = get_prompt_from_promptlayer(name)
result = await query_engine.aquery(prompt)

I noticed that my llama_index is still creating embeddings for my query - why would it do that? How can I tell it to just use the nodes I have provided? Also, is there any direct benefit for using llama_index with such a use case?

6 comments

LLogan M

Don't use the vector index in this case, just use the SummaryIndex -- it will always fetch all nodes (in this case, the single node) without embeddings

LLogan M

The benefit here is mostly the storage and response synthesis code

LLogan M

tbh if you wanted, you could just use a response synthesizer directly

ŁŁukasz

Thanks, I was looking for that

LLogan M

Plain Text

from llama_index.core import get_response_synthesizer

synth = get_response_synthesizer(response_mode="compact", llm=llm)

response = synth.get_response(query_str, ["text1", "text2", ...])

LLogan M

probably the easiest setup tbh

Add a reply

Find answers from the community

Hi, I have a fairly simple use case - my