S3

Question

Hey guys,I was trying to save a BM25Retriever into S3 following this doc: https://docs.llamaindex.ai/en/stable/module_guides/storing/save_load/But unfortunately, it's not working. Can anyone please help me out on how to save a retriever into S3?

Logan M · Answer

Hmm, it only saves to diskhttps://docs.llamaindex.ai/en/stable/api_reference/retrievers/bm25/#llama_index.retrievers.bm25.BM25RetrieverI don't think it's possible to use an fsspec either tbhYou'd need to put your nodes in a docstore, and save the docstore to s3 using s3-fs

datum · Answer

Yeah, just found that out the hard way.

datum · Answer

@Logan M I have another question. How can we convert RetrieverQueryEngine as ChatEngine? Like is it possible?

Logan M · Answer

You can put the query engine as a tool for an agent, or just use the retriever inside a CondensePlusContextChatEngine or ContextChatEngine

Logan M · Answer

The starter example actually shows setting up the agenthttps://docs.llamaindex.ai/en/stable/getting_started/starter_example_local/#adding-rag-capabilities

datum · Answer

@Logan M I couldn't find the documentation to add retriever to condense chat engine.

Logan M · Answer

Not sure what you meanfrom llama_index.core.chat_engines import CondensePlusContextChatEngine chat_engine = CondensePlusContextChatEngine.from_defaults(retriever, llm=llm)

datum · Answer

Thanks a lot 😁

Logan M · Answer

https://docs.llamaindex.ai/en/stable/api_reference/chat_engines/condense_plus_context/#llama_index.core.chat_engine.CondensePlusContextChatEngine

Find answers from the community

S3