The Rag technology stack of .NET is somewhat scattered, does the official have a clear roadmap?

[sheng-jie]

Currently, the .NET technology stack about RAG is scattered in three different github repositories （M.E.DataIngestion , M.E.VectorData and KM ❷ ）. This is not friendly to developers, does the official have a clear development roadmap? For example, will M.E.VD migrate to this repository, and will the connectors for vector databases in SK also be migrated?

[stephentoub]

While we do plan on moving the MEVD source, can you elaborate on why the repo matters to you? It's irrelevant to most developers, where the library is simply consumed via nuget.

[sheng-jie]

At the beginning, we used KM to develop the Rag application, which supported the complete RAG process, including data ingestion, embedding, and retrieval. Although the function is not very perfect and is not maintained now, at least for developers, knowing that KM can build RAG is enough.

But now, if you want to build an RAG application using the latest .NET technology stack, it is difficult to find clear guidance because it is scattered, and I believe that most developers do not even know about M.E.DataIngection. I understand that the official may want it to be like M.E.AI, as an extension of the .NET platform, but really, it's best to put highly related content in a centralized repository. Personally, I don't quite understand this scattered approach, and I'm looking forward to KM2 making a good integration.