search

langchain-ai / langchain-weaviate

MIT License
35 stars 15 forks source link

Documention on Weaviate does not discuss saving and reloading later. Getting index_name/text_key errors when i test. #197

Open heyjohnlim opened 4 months ago

heyjohnlim commented 4 months ago

The weaviate documentation does not explain how to save a database with an index_name nor load an existing vector database and query it. It assumes that we are just using from_documents or from_text.

Also when i was trying to load an existing vector database, i got errors saying index_name and text_key were required parameters to WeaviateVectorStore

Reference: https://python.langchain.com/v0.2/docs/integrations/vectorstores/weaviate/

Thanks again!

hsm207 commented 4 months ago

It seems that based on issue https://github.com/langchain-ai/langchain/issues/11471, the vector store interface does not specify a standard way to reuse an index and it is up to the implementation to define it.

@efriis can you confirm if this is true?

efriis commented 4 months ago

the parameters are integration-specific (e.g. weaviate has different parameters required to connect vs. chroma), and I would recommend adding documentation on how to connect to an existing store since it's not there yet! Maybe add it to the "connecting to weaviate" section?

Good issue!