Faiss Vector Search with Metadata Dictionary Filters

[theSJace]

Hi I am trying to filter the vector search results via the filter flag.

Documentation States that we can do filtering through a single key value pair - as seen below.

results = vector_store.similarity_search(query="thud",k=1,filter={"bar": "baz"})

Is it possible to filter through a nested dictionary within the metadata?

Example Vector:

..."doc_name":xyz , tags = {"color":"blue", "shape":"circle"} ...

Filter Search:
results = vector_store.similarity_search(query="thud",k=1,filter={tags = {"color":"blue", "shape":"circle"}})

Thanks

[asadoughi]

What library are you using? Faiss does not provide these interfaces.

[theSJace]

I am currently using Langchain library. It shows that there is a filter through the vector_store.similarity_search function here.

https://python.langchain.com/api_reference/community/vectorstores/langchain_community.vectorstores.faiss.FAISS.html

[asadoughi]

I suggest filing an issue with https://github.com/langchain-ai/langchain as Faiss does not provide this functionality directly.