Weaviate

ggarnizzle

Is there a way to create a VectorStore that can query a collection/index in a vector DB that is already populated with data which was not ingested via LlamaIndex?

For example, something like this:

Plain Text

vector_store = WeaviateVectorStore(weaviate_client = client, index_name="test", text_key="source_text")

# setting up the storage for the embeddings
storage_context = StorageContext.from_defaults(vector_store = vector_store)

# set up the index
index = VectorStoreIndex([], storage_context = storage_context)

but this tries to create the index for me under the hood. What I want is to point at an already existing index called "test" which is already full of vectors and query against it using a query engine or a retriever.

if i do something like this vector_store.index_name = "Test" it will point at my index but it won't return any results. I suspect this is because its tracking the nodes and I initialized the VectorStoreIndex without any nodes.

Any help would be greatly appreciated!

10 comments

LLogan M

When you construct the vector store object, set the text_field kwarg to be the field in your weaviate db that contains the text

https://github.com/run-llama/llama_index/blob/97b684b7570a5ca48e01ee4e20c5bea68a9e4c71/llama_index/vector_stores/weaviate.py#L86

ggarnizzle

I tried that, in this code block down below and it returns an empty response:

Plain Text

# initialize it to nothing so that it doesn't attempt to create it
vector_store = WeaviateVectorStore(weaviate_client=client, index_name=None, text_key="source_data", text_field="source_data")

embeddings
storage_context = StorageContext.from_defaults(vector_store = vector_store)

index = VectorStoreIndex([], storage_context = storage_context)

vector_store.index_name = "Test"

query_engine = index.as_query_engine()
response = query_engine.query("What is the general theme of movie reviews?")

print(response)

ggarnizzle

also tried it without text_key

LLogan M

oh I think you also need to set index name still, otherwise it doesn't connect to anything

ggarnizzle

tried that and it still returns an an empty response.
do you know if this line here:
index = VectorStoreIndex([], storage_context = storage_context)
where I pass the index an empty list as the value for the nodes, is causing the retrieval to fail somehow?

ggarnizzle

Here it says that the Base retriever tries to retrieve nodes given a query - https://gpt-index.readthedocs.io/en/stable/api_reference/query/retrievers.html - can I interpret this to mean that if I don't pass in any nodes to the index, it cannot perform the operation?

LLogan M

It ends up calling query on the vector store. Your setup should be fine 🤔

LLogan M

You can maybe try debugging the query. My guess is either the index name or text field is wrong

https://github.com/run-llama/llama_index/blob/a009f98aa3c2515b9187673e935f020f884793b2/llama_index/vector_stores/weaviate.py#L272

ggarnizzle

I debugged it successfully. It was actually an issue with the way the weaviate class was set up

ggarnizzle

thanks for your help

Add a reply

Find answers from the community

Weaviate