I'm building my first real app that uses LLM output to power the entire thing. Not just another chatbot, but something that combines a more traditional UI and more traditional set of software engineering practices with the new possibilities of LLMs.
0:00 Intro
0:29 Why this type of app?
1:36 Generating JSON in the Groq Playground
3:21 Tradeoffs of a smaller/faster model
4:30 App architecture for generating recipe ideas
7:18 Recipe Generation
8:25 Validation using LLMs and JSON
9:50 Extracting data from the recipe
10:37 Handling failure when generating JSON
12:48 Other cool features
https://x.com/jeffreybiles