search

infiniflow / ragflow

RAGFlow is an open-source RAG (Retrieval-Augmented Generation) engine based on deep document understanding.
https://ragflow.io
Apache License 2.0
23.53k stars 2.31k forks source link

[Question]: Do you have plan to support other vector databases? #491

Open shaoyie opened 7 months ago

shaoyie commented 7 months ago

Describe your problem

Noticed now the backend is coupling with ElasticSearch. Do you have plan to support some other vector stores, such as Pinecone, Milvus, etc.?

KevinHuSh commented 7 months ago

Do these support hybrid search like weighted vector similarity and keyword similarity? I don't think they do.

jlcbj commented 7 months ago

Milvus 2.4 support hybrid search

KevinHuSh commented 6 months ago

Milvus 2.4 support hybrid search

Can Milvus support text similarity calculation such as BM25 and TFIDF?

jlcbj commented 6 months ago

https://milvus.io/docs/embed-with-bm25.md

Nuclear6 commented 6 months ago

I feel that the latest vector database is better than the currently used hybrid retrieval. Everyone has updated and iterated to the ANN algorithm, and the ragflow project is still using the KNN algorithm.

nadirvishun commented 3 months ago

I have tested the pgvector vector store of fastgpt and the weaviate vector store of dify, both using bce-embedding-base. The retrieval effects of both are better than the es vector store of ragflow, but I am not very sure if it is due to the vector store.

taowang1993 commented 2 months ago

I have tested the pgvector vector store of fastgpt and the weaviate vector store of dify, both using bce-embedding-base. The retrieval effects of both are better than the es vector store of ragflow, but I am not very sure if it is due to the vector store.

how do you compare Dify vs Ragflow?

I am using Dify with Milvus to build a RAG application.

But I am not sure if ragflow is a better option than Dify.

seems like Dify is going to update to RAG2.0 at the end of this year.