Saving Vector Store as KB

Forermx

I was wondering if there was any way of saving a FAISS vector store I create in a python notebook as a knowledge bank I can use later on?

I created a vector store (see code below) which has summaries as the embedded objects, and the parent documents as the retrieved documents. I did this based on LangChain's MultiVectorRetriever. The vector store I created uses DKUEmbedding function as the embedding model and everything I want to do works within the notebook. The problem is, I can't actually use this vector store because the FAISS knowledge bank can't be saved into my flow. Any thoughts? I would be open to work arounds like overwriting an empty knowledge bank with the custom vector store I create, if that's necessary.

Sample Code:

sample_query = "hello world"

sample_embedding = embedding_function.embed_query(sample_query)

embedding_dimension = len(sample_embedding) # Simple way to get embedding model output dimension

new_index = faiss.IndexFlatL2(embedding_dimension)

new_docstore = InMemoryDocstore() # Initialize an in-memory docstore

vector_store = FAISS(

embedding_function=embedding_function,

index=new_index,

docstore=new_docstore,

index_to_docstore_id={}

)
extracted_loader = DataFrameLoader(extracted_df, page_content_column="extracted_text")

docs.extend(extracted_loader.load())

doc_ids = [str(uuid.uuid4()) for _ in docs]


summary_docs = [

    Document(page_content=s, metadata={id_key: doc_ids[i]})

    for i, s in enumerate(summary_df['text_summary'])

]
retriever = MultiVectorRetriever(

    vectorstore=vector_store,

    byte_store=store,

    id_key=id_key,

)


retriever.vectorstore.add_documents(summary_docs)

retriever.docstore.mset(list(zip(doc_ids, docs)))