Faiss

You can start looking from here: https://github.com/run-llama/llama_index/blob/main/llama-index-integrations%2Fvector_stores%2Fllama-index-vector-stores-faiss%2Fllama_index%2Fvector_stores%2Ffaiss%2Fbase.py

thanks i asked this question because i didnt understand how to use similarity_threshold the relevant document are not that relevant and im using voyage embedding for the embedding @WhiteFang_Jr

It’s great to hear you’re exploring similarity search with FAISS in LlamaIndex! The most challenging part of this process is often setting up the indexing correctly to ensure efficient searches. Essentially, FAISS allows you to create an index of your vectors, which can then be queried to find similar items based on distance metrics like Euclidean or cosine distance. In LlamaIndex, you can utilize the FaissVectorStore class to manage your vectors and perform searches, where the similarity calculations are handled in the query method of the FaissVectorStore class. If you need any help getting this set up or have specific questions, I’m here to assist!

i have documents stored on mongodb and the embedding is stored on the file faiss_index but when i query its like it checks only the first document and i dont know why

It sounds like you're facing a frustrating issue where your FAISS queries are only checking the first document in your MongoDB setup.

Right?

like for faiss the relevant documents are the 10 first chunk on mongodb if i ask about them the query answer correctly but if i ask something different it provide me the same source but its wrong

@FatherMonkey

yeah

I am here.

what do u thinks

@WhiteFang_Jr