Integrate Qdrant In-Memory Storage into LangChain

[slavakurilyak]

I propose integrating Qdrant's in-memory storage capabilities into LangChain to enable running Auto-GPT in local mode without requiring a server and enhance the memory feature of Auto-GPT.

Qdrant is an open-source vector search engine and vector database with in-memory payload storage for fast search speeds. By integrating Qdrant's in-memory storage into LangChain, we can enable local mode for Auto-GPT without the need for a server. This would not only benefit developers who want to run Auto-GPT locally without incurring server costs but also enhance the memory feature of Auto-GPT.

[davelm42]

Do you think Qdrant or Chroma would be the best choice for this? I've been experimenting with Chroma and the API and client are MUCH cleaner than Qdrant.

[slavakurilyak]

@davelm42 Perhaps a new feature?

[davelm42]

Yea, I think that's what I'm going to work on this weekend unless someone beats me to it.

[dahifi]

I was going to suggest Weaviate, Llama-index/gpt-index already supports it -- I am unfamiliar with these others.

I am starting a local branch to add the Weaviate client, I'll abstract the memory interface and see if I can make it work without breaking anything.

[dahifi]

This appears to be a sample of the current langchain vector stores:

    "ElasticVectorSearch",
    "FAISS",
    "VectorStore",
    "Pinecone",
    "Weaviate",
    "Qdrant",
    "Milvus",
    "Chroma",
    ```
    
    
    
    

[davelm42]

Looks like someone has a Weaviate PR out there right now:

#424

[Pwuts]

Closing as resolved with #424. Please reopen if you disagree.