LlamaIndex
Log in

Find answers from the community

Updated 8 months ago

Chat

kkolaente
·
Is there a way to filter out retrieved documents with a low top_k value in chat? It seems like I can do that when querying, but not during chat (calling chat_engine.chat())): https://github.com/run-llama/llama_index/issues/905#issuecomment-1484288684
W
k
5 comments
WWhiteFang_Jr
You can do the same for chat engine as well.

You can set it like this
Plain Text
chat_engine = index.as_chat_engine(..., similarity_top_k=3)
kkolaente
Doesn't that only configure the number of results used in the response?
kkolaente
I'd like to prevent using results with a score of < 0.5 or similar
WWhiteFang_Jr
You can pass similarity node post processor alongside top_k.

https://docs.llamaindex.ai/en/stable/module_guides/querying/node_postprocessors/node_postprocessors/?h=similarity#similaritypostprocessor
kkolaente
That looks like what I need, thanks!
Add a reply
Sign up and join the conversation on Discord