-
-
StreamForge Live Overlay: Final view of stream widgets unified in OBS.
-
Interactive Community Voting & Stream Recap: Streamers can launch live chat polls and trigger real-time session summaries instantly.
-
Stream Control Panel & Hype Tracking: Real-time monitoring of chat activity, hype bursts, and basic configurable safety filters.
-
Codex Integration & Plugin Contract: The core engine showing how plain language prompts generate sandboxed, instant-load overlay plugins.
-
Live Control Dashboard: Easy-to-use interface for configuring polls, test commands, and managing active overlay slots
-
OpenAI Codex in Action: AI generating and modifying plugin architecture.
Inspiration & The Problem
Twitch live-streaming is dynamic, but customization has always been rigid. Streamers want custom overlay widgets, interactive chat plugins, and automated moderation tools, but coding them from scratch mid-stream or hiring developers is nearly impossible. We asked ourselves: What if a streamer could build fully functional Twitch extensions just by describing them in plain words? That is why we built StreamForge.
What It Does
StreamForge is an interactive Twitch overlay and plugin ecosystem powered by OpenAI Codex. It allows streamers to instantly generate, deploy, and manage live overlay widgets—such as real-time chat voting, hype trackers, and session summaries—using natural language commands, with zero downtime or restarts required.
How We Built It
- Core Intelligence: OpenAI Codex translates natural language descriptions into clean, sandboxed plugin logic and widget structures.
- Twitch Integration: Leveraged Twitch EventSub and WebSocket/Chat APIs to capture real-time audience events.
- Architecture: Built a modular system where generated plugins load dynamically into the main overlay canvas safely.
- Safety & Moderation: Integrated a basic configurable chat filter to handle blocked words and link replacements cleanly.
Challenges We Faced
Ensuring that newly generated code/plugins could load dynamically into a live overlay without requiring a full browser or server restart was a massive hurdle. We solved this by designing a secure sandboxed plugin loader architecture that compiles and mounts widgets on the fly.
What We Learned
We mastered the intricacies of real-time event streaming on Twitch combined with LLM-driven code generation, proving that prompt-driven development can seamlessly bridge the gap between non-technical creators and powerful custom software.
Transparency & Tech Stack
- AI Engine: OpenAI Codex
- APIs: Twitch EventSub, WebSockets
- Note on Workflow: Certain development and generation sequences in the demo video were sped up for timing efficiency.
Built With
- ai
- html
- javascript
- node.js
- openai
- twitch
