**ValueError: Invalid query mode: hybrid

At a glance

The community member is encountering a ValueError: Invalid query mode: hybrid error when trying to query a Vector Store Index with the "hybrid" query mode. They believe this may be because they are using a simple vector store built from the Documents.example() function.

The community members discuss various approaches to resolving the issue, including trying to import the VectorStoreQueryMode object, updating to a newer version of the library, and using a free Pinecone instance. Ultimately, the community member determines that the issue was likely due to the lack of text data in the vector store index they had built, which prevented the hybrid retriever from operating correctly. They were able to resolve the issue by using the Pinecone instance, which supported both vector and hybrid search.

nno_dice

ValueError: Invalid query mode: hybrid

I am running into an error where I cannot seem to query/retrieve a Vector Store Index with the query mode "hybrid". I believe it may be because I'm just using a simple vector store built from the Documents.example() function.

My code:

Plain Text

## setup
@pytest.fixture
def setup() -> dict:

    os.environ["OPENAI_API_KEY"] = str(settings.openai_api_key)

    service_context = ServiceContext.from_defaults(
        embed_model=OpenAIEmbedding(
            model="text-embedding-ada-002"
        ),
        llm=OpenAI(
            model="gpt-3.5-turbo"
        )
    )

    shots = AlphaMatrix(data=DEFAULT_CATEGORIES)

    vector_index = VectorStoreIndex.from_documents(
        [Document.example()]
        , service_context=service_context
    )

    reranker = LLMRerank(service_context=service_context)

    retriever = CustomRetriever(
        index=vector_index,
        llm=service_context.llm,
        reranker=reranker,
        matrix=shots,
        verbose=True,
    )

    return {
        "retriever": retriever,
        "service_context": service_context,
        "vector_index": vector_index,
        "matrix": shots,
    }

#Where the error occurs:
retriever = VectorIndexRetriever(
            index=index,
            vector_store_query_mode="hybrid",
            alpha=default_alpha,
            **kwargs,  # filters, etc, added here
        )

def test_retrieve(setup):

    retriever = setup.get("retriever")
    query = "What are LLMs good at?"
    results = retriever.retrieve(query)

Error:

Plain Text

ValueError: Invalid query mode: hybrid

18 comments

LLogan M

thats kind of weird 🤔

LLogan M

Attachment

LLogan M

Attachment

nno_dice

Yeah and this part of my code I can't really compromise on like - I'm building a custom hybrid retriever for my project. It needs to be a hybrid retriever

LLogan M

right right

LLogan M

what happens if you do

Plain Text

from llama_index.core.vector_stores.types import VectorStoreQueryMode

print(VectorStoreQueryMode)
mode = VectorStoreQueryMode("hybrid")
print(mode)

nno_dice

how would I import this object? The VectorStoreQueryMode - having a hard time importing it

nno_dice

I was about to try that yeah

nno_dice

If I update from from 0.9.47 to 0.9.48 is my code likely to break? It looked like lots of the directory changed in the recent update haha

nno_dice

I think I'll need to if I want to try this otherwise the path to import looks different as its not being recognized when I try to import that.

nno_dice

Nevermind I think I found the old import path

nno_dice

Same error. I tried VectorStoreQueryMode.HYBRID and VectorStoreQueryMode("hybrid")

LLogan M

v0.10.0 is the breaking change

LLogan M

thats.... impossible? 😅

LLogan M

let me try

nno_dice

Damn just as the big update comes out

nno_dice

I finally figured it out. Its because the documents/vector store index I built doesn't (I presume) store any text data for a Hybrid Retriever to operate on. So when I swapped it over to normal vector search, it passed.

But also, I spun up a free pinecone instance and that worked on both Vector and Hybrid Search. So I'm almost positive its the dummy implementation I went with. I'm going to stick with my free pinecone instance for testing for now. But all is good now 😄

LLogan M

great!

Add a reply

Find answers from the community

**ValueError: Invalid query mode: hybrid