In this AI workshop, Sarmad Qadri, Co-founder/CEO of LastMile AI, shows how to build and evaluate a RAG application using AIConfig — a configuration format for generative AI. AIConfig allows developers to decouple prompt engineering from application development by saving prompts and model settings in a YAML config which can be independently versioned and edited.
The challenge with model-graded RAG evaluation is doing prompt engineering to reliably evaluate application behavior (such as answer succinctness, relevance, faithfulness), and quickly trying different models to identify the best performing one. In this presentation Sarmad shows how to do this with AIConfig Editor extension in VS Code, which turns VS Code into a universal prompt playground.
Install the FREE AIConfig extension to get started: https://marketplace.visualstudio.com/items?itemName=lastmile-ai.vscode-aiconfig&ssr=false#qna
-----------------------
AIConfig links:-
GitHub: https://github.com/lastmile-ai/aiconfig
Workshop code: https://github.com/lastmile-ai/aiconfig/tree/main/cookbooks/RAG-with-Model-Graded-Eval-v2
Chapters:
0:06 - Introducing LastMile AI
08:47 - AI config walkthough
15:00 - Evaluating performance
24:00 - Preparing your prompts for production use
37:30 - Performance optimization and cost reduction
#prompt #ai #rag