And one more question I m following the

At a glance

The community member is following the FAISS vectorstore documentation and has successfully indexed documents, but is encountering issues with retrieval. They are getting a KeyError: '1' error when trying to query the index. The community members discuss potential solutions, such as checking how the index was saved, using index.insert_nodes(nodes) instead, and ensuring there are at least 10 nodes in the index store. The issue is eventually resolved, with the community member finding that the error was caused by having only one node in the index store, which is not enough for the query to work.

Useful resources

BBaygon

And one more question. I'm following the FAISS vectorstore documentation and manage to index documents.

However for retrieval I encounter some issues:

service_context = ServiceContext.from_defaults(
    llm=llm,
    embed_model=base_embeddings,
    chunk_size=512
)
vector_store = FaissVectorStore.from_persist_dir('./storage')
storage_context = StorageContext.from_defaults(
    vector_store=vector_store, persist_dir='./storage'
)
index = load_index_from_storage(
    storage_context=storage_context, 
    service_context=service_context
)
query_engine = index.as_query_engine()
response = query_engine.query(query)

And I get the following error:

self._index.index_struct.nodes_dict[idx] for idx in query_result.ids
KeyError: '1'

Any idea how to fix that?

14 comments

LLogan M

How did you save the index? I just ran the example notebook and it worked fine
https://gpt-index.readthedocs.io/en/stable/examples/vector_stores/FaissIndexDemo.html

BBaygon

vector_store = FaissVectorStore.from_persist_dir(FAISS_INDEX_PATH)
storage_context = StorageContext.from_defaults(
    vector_store=vector_store, persist_dir=FAISS_INDEX_PATH
)

# generate embeddings for each node
    for node in nodes:
        node_embedding = base_embeddings.get_text_embedding(
            node.get_content(metadata_mode="all")
        )
        node.embedding = node_embedding
        print(node_embedding)

# index the message in the vector db
    vector_store.add(nodes)
    index.storage_context.persist()

BBaygon

I'm mixing both the RAG from scratch and FAISS notebooks

LLogan M

Should probably use index.insert_nodes(nodes) I think?

LLogan M

That updates both the docstore, index store, and vector store

BBaygon

hmmm I've just copied paste the notebook in a local script, setup a new venv and it fails with same issue.

BBaygon

Here's my full script, I'm using my own embeddings and Azure OpenAI as llm, but the error seems to come from the index:

BBaygon

Name: faiss-cpu
Version: 1.7.4

BBaygon

Name: llama-index
Version: 0.8.28

BBaygon

Python 3.8.10

LLogan M

hmm not really sure. The only mistake I notice is that you should pass the service context back in when loading

index = load_index_from_storage(storage_context, service_context=service_context)

BBaygon

nope, same stuff....
Did you run the notebook on your local? or can you try to run the script above on your local to see if that's an environment issue from my side?

BBaygon

here is the full stacktrace:

BBaygon

I found the issue. You need at least 10 nodes in your index store. Since I was only loading one document with one sentence only, it creates only one node and fails.

Add a reply

Find answers from the community

And one more question I m following the