GlassKit ships a growing set of runnable examples that each demonstrate a distinct slice of the smart-glasses development stack. Every example is a complete, self-contained project you can copy and adapt — Android app, backend, and configuration included. The sections below describe what each one does, what technologies it relies on, and which GlassKit building blocks it exercises.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/RealComputer/GlassKit/llms.txt
Use this file to discover all available pages before exploring further.
Examples
Rokid Feature Demo
What it demonstrates: Device-feature reference app for Rokid Glasses and emulator/phone testing. Covers touchpad navigation, offline Vosk voice commands, live camera preview, microphone level meter, and audio test tones.Technologies: Kotlin, Vosk offline ASR, Android AudioRecord, AVD emulator.GlassKit building blocks:
ScreenController pattern, HUD viewport layout, touchpad + voice command input mapping, Vosk model setup.IKEA Assembly Assistant
What it demonstrates: Voice-first assembly assistant that streams mic and camera from the glasses to the OpenAI Realtime API via WebRTC. Set up with IKEA wooden box instructions out of the box.Technologies: Kotlin, Node.js 24, OpenAI Realtime API, WebRTC, sideband WebSocket tool calls.GlassKit building blocks: WebRTC audio + video streaming, OpenAI Realtime session brokering, sideband function calls,
SESSION_INSTRUCTIONS customization.Speedrun Timer (RF-DETR)
What it demonstrates: Vision-driven speedrun HUD for physical tasks. RF-DETR detects configured objects and automatically advances HUD split timers with a two-hit confirmation rule. Includes a sushi speedrun out of the box.Technologies: Kotlin, FastAPI, aiortc, RF-DETR, Roboflow
inference library.GlassKit building blocks: WebRTC video + data channel messaging, object detection inference loop, configurable speedrun state machine, annotated frame capture.Live Scene Reader (Overshoot)
What it demonstrates: Minimal real-time scene reader that continuously describes what the wearer is looking at. Inference text streams to the HUD as a rolling, auto-scrolling log.Technologies: Kotlin, FastAPI, Overshoot VLM API, WebRTC, WebSocket relay.GlassKit building blocks: Overshoot stream lifecycle management, WebRTC brokering, HUD text rendering and auto-scroll, backend WebSocket relay.
Proactive Drink-making Coach
What it demonstrates: Full-stack AI assistant that scans ingredients, picks a recipe automatically, and guides each step with spoken instructions. Combines Overshoot visual inference, OpenAI Realtime speech, and a data-driven recipe workflow.Technologies: Kotlin, FastAPI, Overshoot VLM API, OpenAI Realtime API, WebRTC (video + audio), sideband WebSocket.GlassKit building blocks: Multi-connection session orchestration, Overshoot prompt switching, OpenAI sideband speech control, recipe JSON workflow, server-authoritative HUD state.
IKEA Assistant + RF-DETR Variant
What it demonstrates: An updated version of the IKEA assembly assistant that adds RF-DETR object detection on the backend. Detected, annotated frames are injected into the Realtime API conversation for more accurate spatial understanding.Technologies: Kotlin, Node.js, OpenAI Realtime API, RF-DETR, WebRTC.GlassKit building blocks: All blocks from the base IKEA assistant, plus RF-DETR inference and annotated-frame injection into a Realtime session.
This variant lives in
examples/rokid-openai-realtime-rfdetr in the repository. It does not have a separate documentation page — see the IKEA Assembly Assistant page and the README.md inside that directory for the full setup guide.Common Requirements
Most examples share the same hardware and toolchain requirements. Each example README documents its exact steps and environment variables.| Requirement | Details |
|---|---|
| Rokid Glasses + dev cable | Required for on-device testing. See docs/how-to-get-rokid-glasses.md. |
Android Studio / adb | Required for all Android app builds and deployments. |
uv | Used for all Python backends (FastAPI examples). |
| Node.js 24 | Used for the TypeScript backend in rokid-openai-realtime. |
OPENAI_API_KEY | Required by the OpenAI Realtime examples. |
OVERSHOOT_API_KEY | Required by the Overshoot examples. |
ROBOFLOW_API_KEY | Required by the RF-DETR examples (only for initial weight download). |
Copy an Example
To copy any example and start building on top of it:rokid-feature-demo with the name of any example directory.