search

Azure-Samples / Azure-Functions-OpenAI-Demo

This C# demo is based on azure-search-openai-demo and uses a static web app for the frontend and Azure functions for the backend API's. This solution uses the Azure Functions OpenAI triggers and binding extension for the backend capabilities.
MIT License
20 stars 11 forks source link

Proposal - refactor into 3 smaller samples #24

Open paulyuk opened 1 month ago

paulyuk commented 1 month ago

The proposal is to split this very good sample into 3 smaller fit for purpose samples. The current sample is great but it is too complex, too hard to set up locally, and too hard to find. By doing this proposal, we will have easy to search samples that each do one job well.

Proposed smaller samples: 1) ChatGPT Chatbot - using assistants 2) RAG - using embeddings output and semantic search 3) Function calling - using Assistant and Assistant Skills trigger

Note, I observe we are already down this path. e.g.

So we'd just need to fill out matrix a bit more with RAG and the other languages.

For languages I recommend: P1 - Python, Javascript/Node, C#/Dotnet P2 - TypeScript, PowerShell

@eamonoreilly @BigMorty @MadhuraBharadwaj-MSFT please weigh in

paulyuk commented 1 month ago

I also wonder if the SWA frontend UI is worth the complexity (all the extra files and dev ops tasks/scripts to deploy). I could see this demo still existing for full end to end, whereas the small samples are lightweight. would love opinions.

BigMorty commented 1 month ago

Definitely agree in splitting this sample into three samples, that will make them much easier to grasp for folks new to our AI extension. Regarding the SWA frontend, agree we should just use http files like we do in the other samples, assuming we can do the file upload fairly easy still.

MadhuraBharadwaj-MSFT commented 1 month ago

Yes, agreed on this proposal. It makes sense to do a logical split and have them in smaller, easier to consume scenarios. I'm currently working on the RAG + AI Search sample for JavaScript.