Anatomy of a conversation
Programmatically conversing with LLMs
Slides
Outline
Anatomy of a conversation - HTTP requests, message roles, ellmer basics - Activity: 02_word-game (5 min) - Demo: 03_clearbot
Understanding statelessness - LLMs don’t remember - you send full conversation history each time
How LLMs work - Training on text data, tokens, probabilistic generation - Demo: 04_token-possibilities
Interactive chatting - Demo: 05_live - live_console() and live_browser()
How to think about LLMs - Think empirically, LLMs are jagged, embrace experimentation
Shinychat basics - Building apps with chat_mod_ui() and chat_mod_server() - Activity: 03_word-games (7 min)