xg-glass-sample

This directory contains a set of sample apps built with the xg.glass SDK, to help developers quickly understand:

• How to use the unified APIs across different smart glasses
• How to build, install, and run a working glasses app from a single Kotlin entry file

These samples target xg.glass SDK 0.3.0.

If you're new to the SDK, start with the main documentation (see developer guide).

Examples

Example	What it demonstrates	Run command
photo_translator	Capture a photo, call an OpenAI-compatible vision model, and display the translation.	xg-glass run photo_translator/PhotoTranslatorEntry.kt --sdk /path/to/xg-glass-sdk
exam_solver	Auto-capture loop with streaming AI answers and conversation memory.	xg-glass run exam_solver/ExamSolverEntry.kt --sdk /path/to/xg-glass-sdk
teleprompter	Simulator-runnable display teleprompter with streaming/paged display degradation and tap/long-press controls.	xg-glass run teleprompter/TeleprompterEntry.kt --sim --sdk /path/to/xg-glass-sdk
voice_notes	Simulator-runnable microphone capture with transcription when AI settings are configured, otherwise an honest audio summary.	xg-glass run voice_notes/VoiceNotesEntry.kt --sim --sdk /path/to/xg-glass-sdk
ai_assistant	Standalone simulator phone-host app: video stream awareness, latest-frame snapshot on tap/button, OpenAI-compatible vision answer, and display on glasses.	cd ai_assistant && xg-glass run --sim --local_video /path/to/sample.mp4
play_feature_delivery	Standalone Android App Bundle sample that keeps Even + Simulator in the base app and loads the Meta adapter through an on-demand Play Feature Delivery split.	cd play_feature_delivery && ./gradlew :app:bundleDebug

Prerequisites

Install the CLI and clone the SDK checkout once:

pip install xg-glass
git clone https://github.com/hkust-spark/xg-glass-sdk

The SDK checkout is needed when the PyPI-installed CLI runs a single Kotlin entry file. In the examples below, replace /path/to/xg-glass-sdk with your checkout path. If you are running from inside an SDK checkout, the same commands still work without --sdk.

---

photo_translator (Photo Translator)

Location: xg-glass-sample/photo_translator

This sample demonstrates a minimal end-to-end flow: capture photo → LLM translate → display on glasses.

• Capture a photo from the glasses camera
• Encode the image as base64 and call OpenAI Chat Completions for image-text translation
• Display the translated result on the glasses

Quick run (recommended)

Run the single-file entry directly from this directory:

cd xg-glass-sample/photo_translator
xg-glass run PhotoTranslatorEntry.kt --sdk /path/to/xg-glass-sdk

Notes:

• xg-glass is installed with pip install xg-glass; for PyPI installs, pass --sdk /path/to/xg-glass-sdk when running a single .kt entry.
• Before running, replace YOUR_OPENAI_API_KEY_HERE in PhotoTranslatorEntry.kt with your own key (this is a placeholder; for real apps, inject secrets securely).

Core logic (you can build this app in ~10 lines)

In PhotoTranslatorEntry.kt, the core logic that implements capture → translate → display is essentially just the snippet below (you only need ~10 lines like this to build the full app):

override suspend fun run(ctx: UniversalAppContext): Result<Unit> {
    val img = ctx.client.capturePhoto().getOrThrow()
    val b64 = Base64.getEncoder().encodeToString(img.jpegBytes)
    val req = chatCompletionRequest {
        model = ModelId("gpt-4o-mini")
        messages { user { content { text("Translate the text in this image to Chinese. Output only the result."); image("data:image/jpeg;base64,$b64") } } }
    }
    val text = openAI.chatCompletion(req).choices.firstOrNull()?.message?.content.orEmpty().ifBlank { "No text" }
    return ctx.client.display(text, DisplayOptions())
}

---

ai_assistant (AI Assistant)

Location: xg-glass-sample/ai_assistant

This standalone Android phone-host app demonstrates the flagship video-stream loop: connect to simulator glasses, start a LOW-tier camera stream, capture the latest stream frame when the glasses tap event or phone button fires, send it to an OpenAI-compatible vision endpoint, and display the answer on the glasses.

Configuration lives in untracked ai_assistant/local.properties or environment variables:

ai.baseUrl=http://10.0.2.2:8765/v1
ai.apiKey=mock-key
ai.model=mock-vision

The endpoint must accept POST /v1/chat/completions with an image_url data URL. OpenAI, DashScope's OpenAI-compatible mode, and Ollama's OpenAI-compatible API can be used by changing ai.baseUrl, ai.apiKey, and ai.model. If any setting is missing, the app logs and displays a clear configuration message and does not crash.

Run hardware-free on an emulator:

cd xg-glass-sample/ai_assistant
xg-glass run --sim --local_video /path/to/sample.mp4

Without --local_video, the simulator uses the emulator camera. The sample was verified against a local mock endpoint; any OpenAI-compatible provider with vision support should work.