Building an AI Travel Agent That Never Hallucinates
summary
Hands-on tutorial shows a running AI travel agent that avoids hallucinations by separating execution from understanding: CALM and Rasa provide a flow engine that runs predefined business logic while LLMs handle language. Demo uses a GitHub Codespace, a seeded flights DB, Rasa train/actions and the inspector UI, and offers templated responses (rephrasing optional) so answers are deterministic. Alternative branch uses fine-tuned self‑hosted models like llama 8b to cut cost and latency. End‑to‑end tests, slot validation (catching invalid dates) and per‑turn token metrics are included; the demo works live.
AI-written summary · about
chapters (10)
- 0:00Welcome
- 3:28AI agents need reliable logic
- 11:18Prompt and prey limitations
- 23:59AI travel agent demo
- 33:17Testing the travel assistant
- 39:04Com separates conversation logic
- 51:15File structure and flows
- 1:00:50Testing AI assistants
- 1:09:13Fine-tuning models
- 1:17:25Q&A
related talks
01h 30mHow to build AI Agents: A step by step guide to creating AI agents
26mAI agents 101
53mHands-on with AI agents: From scratch to LangGraph
31mHow to Build Autonomous AI Agents That Run Anywhere
02h 08mIntroduction to AI Agents
30mWhen AI Takes Over: DevOps Nightmare Unleashed
27mHow to Build AI Agents and Keep Your Sanity
01h 04mBuilding Expertise Into Generated Conversations