How to use a redis or other database to store docstores with Multi-modal RAG in OpenAI and LangChain?

API
,
1

Hi everyone, I was watching a tutorial about Multi-modal RAG, but I only found examples with ‘InMemoryStore’ in langchain. I have pdfs with tables and images, that’s why I need to use Multi-modal approach.

Searching in the documentation only there are these 3 classes:
1 - EncoderBackedStore.
2 - LocalFileStore.
3 - InMemoryStore.

But seems that them all are for local instances, but in production is supposed to I have to store docstore in a remote database, isn´t it?.

What can I do?. Thank you.

from langchain_community.vectorstores import Chroma
from langchain.retrievers.multi_vector import MultiVectorRetriever
from langchain.storage import InMemoryStore
from langchain_openai import OpenAIEmbeddings

# The vectorstore to use to index the child chunks
vectorstore = Chroma(collection_name="summaries", embedding_function=OpenAIEmbeddings())

# The storage layer for the parent documents
store = InMemoryStore()
id_key = "doc_id"

# The retriever (empty to start)
retriever = MultiVectorRetriever(
    vectorstore=vectorstore,
    docstore=store, # ====> I want to change this to a remote storage
    id_key=id_key,
)
2

from langchain_community.storage import RedisStore

The vectorstore to use to index the child chunks

vectorstore = Chroma(collection_name=“summaries”, embedding_function=OpenAIEmbeddings())

redis_store = RedisStore(
redis_url=f’redis://{os.getenv(“REDIS_URL”)}:{os.getenv(“REDIS_PORT”)}',
namespace=redis_namespace
)

The storage layer for the parent documents

store = InMemoryStore()
id_key = “doc_id”

retriever = MultiVectorRetriever(
vectorstore=vectorstore,
byte_store=redis_store,
id_key=id_key,
)

pattern = f"{redis_namespace}*"

Keys that we are looking for to understand if the cache is empty

keys = list(redis_store.yield_keys(prefix=pattern))
if len(keys) == 0:
# redis_store.mset(list(zip(doc_ids, documents)))
retriever.docstore.mset(list(zip(doc_ids, documents)))
else:
print(“Cache in Redis is found…”)

1 Like