Trillboards Sensing SDK for Android

agent-core is the Android sensing primitive that powers Trillboards on-device audience intelligence: face detection, audio classification, speech recognition with diarization, BLE / WiFi / mDNS / SSDP / ARP / HTTP discovery, native sensor reads, and Vertex multimodal embeddings. Targets Android 8.0+ (minSdk=26), built against compileSdk=35, ARM64 only.

Apache 2.0. Public. Anonymous-readable via JitPack.

Install

settings.gradle.kts (or settings.gradle):

dependencyResolutionManagement {
    repositories {
        google()
        mavenCentral()
        maven { url = uri("https://jitpack.io") }
    }
}

app/build.gradle.kts:

dependencies {
    implementation("com.github.trillboards:agent-core:v1.0.0")
}

(Or pin to a specific commit SHA: com.github.trillboards:agent-core:<sha>.)

Quick start

import com.trillboards.ctv.core.AgentConfig
import com.trillboards.ctv.core.TrillboardsSensingSdk

val sdk = TrillboardsSensingSdk(
    context = applicationContext,
    config = AgentConfig.Builder()
        .partnerApiKey("YOUR_PARTNER_API_KEY")
        .deviceCode("YOUR_DEVICE_CODE")  // your fleet's device identifier
        .enableFaceSensing(true)
        .enableAudioSensing(true)
        .build()
)
sdk.start()

What ships in the AAR

• Face: FaceXFormer demographics (age / gender bucketed when on-device model loaded; null otherwise), gaze estimation, per-track dwell + emotion classifier
• Audio: MediaPipe AudioClassifier (yamnet labels), Moonshine ASR with speaker diarization, voice-activity detection
• Discovery: BLE scan + Phase 6 (UWB / Auracast / ChannelSounding API 36+), WiFi BSSID/SSID hashing + nearby networks, mDNS / SSDP / ARP / HTTP probes
• Native sensors: ambient light, barometer, IMU
• Cloud inference: Vertex multimodal embeddings via Trillboards API (no GCP credentials needed on device)

Configuration

The full surface (~30+ config knobs: enable/disable each sensing primitive, sampling rates, diarization on/off, FaceXFormer on/off, debug logging, etc.) is documented inline in AgentConfig.kt. Common partner integrations:

• Camera + mic + face/audio sensing (default — full surface): leave defaults; SDK adapts to device capabilities
• Audio-only (no camera): set enableFaceSensing(false), leave enableAudioSensing(true)
• Discovery-only (no camera, no mic): set enableFaceSensing(false), enableAudioSensing(false), enableDiscovery(true) — useful for kiosks where audio/video privacy is restricted

Privacy

• All face / audio / image inference happens on-device. No raw frames or audio buffers leave the device.
• Cloud calls send only structured features (face counts, attention level, speech transcripts when diarization is enabled, etc.) — never images.
• GDPR / CCPA opt-out is per-screen via the Trillboards portal; SDK respects opt-out flags from the heartbeat response.

License

Apache License, Version 2.0. See LICENSE. The bundled ML models (Moonshine ASR, MediaPipe AudioClassifier, MediaPipe pose landmarker, yamnet) ship under their respective licenses (Apache 2.0 for all).

Reporting issues

File an issue on this repo, or email engineering@trillboards.com.