Abstractions and libraries for RAG with different file formats

[ManojG1978]

https://github.com/microsoft/semantic-kernel/tree/main/dotnet/samples/Demos/VectorStoreRAG

I see that this is the only reference I could find which brings Vector search for PDF files through the abstractions in Semantic Kernel.
Compared to the abstractions created in the official RAG experiment accelerator (https://github.com/microsoft/rag-experiment-accelerator/tree/development), I sense the support in Semantic Kernel seems a bit weak. For instance, is UglyToad the only way to load and chunk PDFs. Many classes/abstractions are in the sample folder, which ideally should be part of the SK libraries - like RawTextContent etc,?

When would we have proper abstractions for dealing with files (this is a very important aspect of RAG)? Also, any guidance on loading other file formats, like Docx, md etc? Which is the recommended loader by the SK team? Any roadmap for these?

[sophialagerkranspandey]

Tagging @westey-m

[sophialagerkranspandey]

We have added this item to our SK backlog and appreciate you bringing this up.