Build an AI Receptionist with Vapi, Twilio & Codex | Full Tutorial

I built an AI receptionist for restaurants after noticing how much revenue can disappear when calls go unanswered during busy service. What started as a university Pattern Recognition project became a working AI phone system that can answer common guest questions, collect reservation details, connect to backend tools, and pass more complicated requests to the restaurant team. In this video, I show the full story and the complete setup I used, including: • Creating the AI assistant in Vapi • Adding the restaurant knowledge base• Configuring the prompt, transcriber, model, and voice • Creating the knowledge-base tool • Getting a phone number through Twilio • Using Codex or Claude Code to plan and build the backend• Connecting the tools and testing the final receptionist • The mistakes, latency, broken versions, and everything that went wrong This is part devlog, part comedy, and part full free tutorial. The first version paused for six seconds before answering. That was not latency. That was realistic employee hesitation. Resources GitHub project and files: https://github.com/utanvir71/AI-Recep... My website:https://utanvir.com Watch the original AI receptionist video:   • receptionist demo video   The repository includes the knowledge-base example, Vapi receptionist prompt, backend structure, and other files shown in the video. This project is still being improved. It may misunderstand names, dates, accents, or very fast callers, so test everything carefully before using it for a real business. I am now trying to turn this university project into a real product and get my first restaurant customer. Subscribe to see whether I get paid for it… or continue washing dishes forever. Graphic designed by Freepik https://www.freepik.com Music designed by Pixabay https://www.pixabay.com #AIReceptionist #Vapi #Twilio #Codex #ClaudeCode #AIAutomation #AITutorial #RestaurantTechnology