Add support for Milvus vector DB about llm-chain HOT 1 CLOSED

Hello,
I would love to contribute to this project and this issue. I have some experience working with Milvus in both Python and GO, I see that you support qdrant as a vector store for now. I have some questions regarding the implementation :

• As I can see from the qdrant implementation, all the collection creation details are given to the user (embedding_dim size, distance, index type ... ) and the Qdrant struct only has the client instance. Should the Milvus offer the same flexibility or hide the collection creation details, the index type, etc... from the user?
• Index creation is an additional step in Milvus, should we integrate this from the llm-milvus side: for example at query time check if an index on the collection exists and is loaded, if not create it.
• The add_text and add_documents methods return Vec<String>, where string are UUIDs generated by the client in Qdrant. Milvus has an auto_id FieldSchema, we could parse the response or create an additional Id field with UUID to have the same interface ?

I really hope I could help on this issue. Thanks for your answers !

from llm-chain.