New Feature: A sample that leverage SK implementing a documentation AI Assistant

[MithrilMan]

This feature request is not about a SK feature per se but could be helpful for two reasons:

• showing best practice on SK programming
• add value to the community

Basically what I propose is levaraging SK to implement a sample that uses chat completion, memory, RAG, and function calling in order to implement an application that help you teach how to use semantic kernel.

Ingesting all the updated semantic kernel documentation, blog post and samples, the user can ask the AI how to accomplish a specific task in semantic kernel and it would point out to relative documentation/posts and give back snippets.

This could be an idea that the community could implement itself, why not, but some internal help would be helpful about following best practices and updated "way to do it"

What do you think?

[JadynWong]

https://github.com/microsoft/chat-copilot

[matthewbolanos]

This would make a good sample someone could build during a hackathon, @alliscode

[matthewbolanos]

This would actually be a good RAG sample as part of the search graduation.

[MithrilMan]

This would actually be a good RAG sample as part of the search graduation.

Which graduation are you talking about? Is something public? I'd like to see what's talking about because I'm interested to dig into knowledge graph, something I haven't found already in sk

[westey-m]

We have an example now that addresses most of the asks here. See: https://github.com/microsoft/semantic-kernel/tree/main/dotnet/samples/Demos/VectorStoreRAG

There are some good ideas here about also including blog posts and samples as RAG input that would be good to consider for us to add.

[sorin-costea]

I'm also eager to see the kernel service generating the embeddings, and an agent being able to use them natively, not over tens of intermediate steps (and/or hacks). I mean I'd love to have it as basic functionality, as most use cases I see around start with augmenting from some documents - so this definitely deserves some builtin facilities