Why I chose this

I didn’t want to build something that someone has to review, but instead a project someone wants to review. AI agents are technically powerful, but not very fun to use as a text-based interface. What excites me most about Tavus is it’s as useful as other agents AND tells a story, one like “I’m talking to a real person”. You know you’re not, but your brain goes with it anyway. Why not have fun with that story?

I documented my process here. I think in terms of failure points, user experience, expectation and delivery. I try to be thorough, but it’s a lot to read, so I hid some with toggles. Feel free to explore and press me on any points!

Architecture

Framework

• Next.js 14 (App Router)
• TypeScript
• React 18

Styling

• Tailwind CSS

AI / Video

• Tavus API — AI video agent (replica + persona)

Code Editor

• highlight.js — syntax highlighting (atom-one-dark theme)

Improvements

• Supabase for real-time subscriptions
• Ngrok for local dev on webhooks

My Process