Inspiration

Live Twitch chats move quickly. Viewers share useful ideas, ask questions, react to moments, and contribute to the stream, but much of that activity gets lost in the flow of messages. Streamers also have to balance entertaining their audience with keeping chat useful and manageable.

We built Breadcrumb to help streamers turn that fast-moving chat into a more meaningful community experience—without taking control away from the streamer.

What it does

Breadcrumb is an AI-powered Twitch community companion with three core features:

  • Hint Ladder: Viewers submit private clues, and the system creates three progressive hint levels. Streamers can review, edit, approve, reject, and reveal hints to their audience.
  • Chat Guard: Identifies potentially disruptive patterns such as repeated messages, duplicate content, copypasta, promotional links, and chat flooding. It supports manual review rather than automatically banning viewers.
  • Chat Pulse: Helps streamers understand ongoing chat activity by surfacing summaries, recurring topics, repeated questions, and suggestions.

Breadcrumb also includes an OBS Browser Source overlay so selected content can appear on stream. Private submissions and unrevealed hints are kept out of the public overlay.

How we built it

We built Breadcrumb as a monorepo with a Next.js and React frontend, and a Node.js backend using Express and Socket.IO.

The application connects with Twitch through OAuth and EventSub to receive live chat events. Supabase provides persistent storage, while the OpenAI API is configured as the AI provider for the application. The OBS Browser Source connects the streamer's dashboard experience with the live broadcast.

We used OpenAI Codex during development to help implement, refine, and test the different parts of the application.

Challenges we faced

One of our biggest challenges was designing interactions that work with fast-moving live chat while keeping the streamer in control.

We also had to think carefully about privacy: a viewer's private clue or an unrevealed hint must never accidentally appear on the public overlay. Another challenge was connecting several systems—Twitch authentication, live EventSub events, persistent storage, AI workflows, and OBS—into one working application.

Deploying the frontend and backend separately also required careful configuration of API URLs, environment variables, OAuth redirects, and production services.

What we learned

Building Breadcrumb taught us how to connect real-time event-driven systems with an interactive web application. We gained practical experience with Twitch OAuth and EventSub, Socket.IO, Supabase, AI-provider integration, OBS Browser Sources, and production deployment.

Most importantly, we learned that useful AI in a live community should support human decisions—not replace them. Breadcrumb is designed to give streamers better context and tools

Built With

Share this project:

Updates

posted an update —

Introducing Breadcrumb — Making Twitch Communities More Interactive

I built Breadcrumb to help Twitch streamers turn their live chats into more interactive, manageable, and meaningful community experiences.

Here’s what Breadcrumb brings together:

  • Hint Ladder: Viewers submit private clues, and AI helps transform them into progressive hints that streamers can review and reveal.
  • Chat Guard: Identifies spam, repeated messages, copypasta, and promotional links while keeping moderation decisions in the streamer's hands.
  • Chat Pulse: Highlights recurring questions, topics, and patterns from live chat.
  • OBS Overlay: Displays approved interactions directly on stream.

The goal is simple: help streamers engage with their communities while staying in control.

This is the beginning of Breadcrumb. I'm excited to keep improving it and explore what's possible when AI and live community interaction come together.

Check out the project and share your feedback!

Log in or sign up for Devpost to join the conversation.

Submission history