NextJS Web App

[Konstantinos Passadis | Azure MVP | MCT (passadis)]

Hello !

I have already build a NextJS Client in a classic Web Chat - Assistant Format where we get info from the Server.

I am wondering if it is interesting , and whether we could collaborate to filter the responses a bit better.

Currently working on a "Read More..." hyperlink but the results are enormous . I am starting to formulate an idea to get a middleware model to correctly match the Question with the results ...

Thoughts ?

[Pieter de Bruin (pdebruin)]

Interesting, thanks for sharing! 🙂
From your message it looks like you got your webapp connected to the MCP server, and that the challenge is that the response is bigger than useful. This seems to be the expected behavior as described in the readme: "10 high-quality content chunks (each max 500 tokens)". I think the current scenario we want to support is that AI agents use a model grounded in additional Learn context to answer user questions, like GitHub Copilot, Claude, etc. Your webapp doesn't have a model to interpret the results, but you are thinking about adding one. That seems to be the right path forward.
Specifically on filtering, we are considering to potentially allow a limited scope for certain products. I don't think we are considering reducing the number of results.
cc Tianqi Zhang (@TianqiZhang)

[Konstantinos Passadis | Azure MVP | MCT (passadis)]

Perfect ! Thank you Pieter de Bruin (@pdebruin) !
That's exactly what my way of thinking is , so i am going to continue working on it
If someone / anyone would like to participate let me know, the codebase is yet to be uploaded in GitHub
Cheers !

[Konstantinos Passadis | Azure MVP | MCT (passadis)]

Hello !
I was able to work out a method to clear the results and apply a nice formatted response !

I will push more tests with complex Questions to see how it goes but i think we found the way !

Thank you Pieter de Bruin (@pdebruin) for the positive feedback and encouragement !

Cheers !

[Konstantinos Passadis | Azure MVP | MCT (passadis)]

It seems results are well structured and properly arranged 👍

I will post the solution ASAP !

Cheers

[Konstantinos Passadis | Azure MVP | MCT (passadis)]

Hello !
I have gone through an exciting and interesting journey to make this flexible and usable !
I hope you guys find it useful !

https://github.com/passadis/mslearn-mcp-chat

BlogPost