[Bug]: FAISS failed to read from the pat...

At a glance

Hey all!

Toying with some examples and using a SimpleVectorStore saved to disk. Generating the embeddings and issuing the first query seems to function appropriately but when issuing another query I am receiving the following error:

Plain Text

 "/root/.cache/pypoetry/virtualenvs/pa-MPXmvNdN-py3.11/lib/python3.11/site-packages/llama_index/vector_stores/simple.py", line 306, in from_persist_path
    raise ValueError(
ValueError: No existing llama_index.vector_stores.simple found at ./storage/vector_store.json, skipping load.

I've found what looks like a similar issue in the repo: https://github.com/run-llama/llama_index/issues/9110

Has anyone seen this before?

7 comments

eedhenry

I've rolled around versions of llama_index between 0.9.7 and 0.9.12 and the issue seems to persist.

LLogan M

ah, a recent change to the default vector_store json name is causing this.

I can submit a fix, but there's also a workaround

Plain Text

# load index from disk
vector_store = FaissVectorStore.from_persist_path("./storage/default__vector_store.json")
storage_context = StorageContext.from_defaults(
    vector_store=vector_store, persist_dir="./storage"
)
index = load_index_from_storage(storage_context=storage_context)

LLogan M

good catch

LLogan M

oof, two week old gh issue got lost in the abyss I see

LLogan M

https://github.com/run-llama/llama_index/pull/9330

eedhenry

Thanks again, @Logan M ! You're a bit of a machine, clearly. 🙂

LLogan M

If I know the fix will take less than 10 minutes, I'm stopping what I'm doing to fix 😆👏

Add a reply

Find answers from the community

[Bug]: FAISS failed to read from the pat...