LlamaIndex
Log in

Find answers from the community

Home
Members
ogg
o
ogg
Offline, last seen 3 months ago
Joined September 25, 2024
oogg
·

This is not clear from the documentation

This is not clear from the documentation, but when we add a document to the index, does that get automatically added to the vector db (I'm assuming yes from the way things are set up but good to ask) ? What about deletion and refresh ?
11 comments
o
W
oogg
·

Hi, this might be an easy one, I am

Hi, this might be an easy one, I am using fast API to allow users to access custom retrievers. I'd want to be able to cache or reuse a retriever if multiple questions are asked to it. how can I achieve that ?
3 comments
L
o
oogg
·

An agent similar to what is provided by

An agent (similar to what is provided by Llama index or Langchain) that has a role and is able to play questions answering. Yes an agent with Llama 2
7 comments
L
o
oogg
·

Retriver

Quick question I was wondering when we use a retriever - does the LLM see the metadata as well ?
1 comment
L
oogg
·

Can anyone explain to me in simple terms

Can anyone explain to me in simple terms when would we need to implement a treeindex vs a simple index retriever ?
2 comments
o
L
oogg
·

Hey guys I m a bit confused as to how

Hey guys, I'm a bit confused as to how VectorStoreIndex.from_documents work. I am using chromadb to have my code the way I do it is :
1- I upload my document,
2- split it into chunks
3- load it in the vector store using this code
Plain Text
def create_index_from_documents_background(docs, document, storage_context, service_context):
    
    index = VectorStoreIndex.from_documents(docs, storage_context=storage_context, service_context=service_context)
    print(index)
    index.set_index_id(str(document.id))

4- try to retrieve it later using a unified id. Issue is, it returns an empty list. Do I need to commit the changes someone to chromadb or something ?

I've attempted to visualize the VectorStoreIndex in Chroma and that's not possible apparently. Any guidance on how to manage this more efficiently / get me to update the doc_id to be linked to the right document
3 comments
o
L