I am having an index having 170000

HK · 2024-06-12T20:11:16.612Z

I am having an index having 170000 embeddings and if I use query engine then it takes 15-20 seconds to give me an answer. What is the most appropriate way to handle these many embeddings index. I want to have results within 2-3 seconds.

LLogan M

Use a vector db like qdrant or some other integration

HHK

Thanks @Logan M for the reply. I am trying to use FAISS index now and I have created an index like this;

d = 1024
faiss_index = faiss.IndexFlatL2(d)

nodes = []
vector_store = FaissVectorStore(faiss_index=faiss_index)
storage_context = StorageContext.from_defaults(vector_store=vector_store)

for index, row in data.iterrows():
    text_embedding = json.loads(row["list_embedding"])
    text = row["text"]
    node = TextNode(text=text, metadata={"id":row["id"]}, embedding=text_embedding)
    nodes.append(node)

vector_index = VectorStoreIndex(nodes=nodes, storage_context=storage_context)

# save index to disk
vector_index.storage_context.persist(persist_dir="./vector_store__faiss_index_50000")

And loaded the index like this;

# load index from disk
vector_store = FaissVectorStore.from_persist_dir("./vector_store__faiss_index_50000")
storage_context = StorageContext.from_defaults(
    vector_store=vector_store, persist_dir="./vector_store__faiss_index_50000"
)
newwindex = load_index_from_storage(storage_context=storage_context)

But the problem is when I try to query with this index;

query_engine = newwindex.as_query_engine()
response = query_engine.query("Can you show me all the products related to bikes?")

It is giving me **AssertionError: **

LLogan M

hmmm, not sure whats going on there. Is there a full traceback? (FAISS is honestly pretty annoying to use lol)

HHK

Thanks Logan, I have used FAISS implementation from their github and it working good now. will share the compelte traceback in a few hours.

Find answers from the community

I am having an index having 170000