The Shiny AI Stack: shinychat, querychat, and the LLM Plumbing Beneath
Conference
Slides
Abstract
LLM-powered chat is becoming a core pattern in modern Shiny apps, but building a good chat experience — with streaming, history, file attachments, editing, and cancellation — is harder than it looks. In this talk, we’ll tour the Shiny AI stack: shinychat’s getting-started API, which gives you a polished chat UI with all of these niceties out of the box, and the lower-level plumbing beneath it (ellmer and chatlas) that handles the connection to the LLM.
We’ll also see how these building blocks power domain-specific applications like querychat, and why they’re designed to be general enough for essentially any LLM-powered Shiny app. Examples are in R, with Python equivalents available for everything we’ll cover.