@trillboards/edge-core

Core runtime for Trillboards Edge AI SDK. Device identity, Socket.io connectivity, configuration, signal buffering, health monitoring, and auto-recovery for DOOH/CTV devices.

Install

npm install @trillboards/edge-core

What This Does

Provides the foundational infrastructure that all other @trillboards/edge-* packages depend on:

• Device Identity — generates stable device fingerprints and session IDs from hardware characteristics
• Socket.io Manager — real-time connection to Trillboards API with auto-reconnection and telemetry emission
• Signal Buffer — SQLite-backed offline queue that caches audience metrics when connectivity drops
• Health Registry — tracks 6 subsystems (socket, player, camera, microphone, ML pipeline, memory)
• Recovery Ladder — 6-level auto-recovery from reconnect through service restart to rollback
• Capability Tiers — classifies devices into TIER_1 through TIER_4 based on RAM and NPU availability
• Sensing Config — deserializes 70+ sensing parameters for audience measurement tuning

Usage

import {
  DeviceIdentity,
  SocketManager,
  SignalBuffer,
  SubsystemHealthRegistry,
  SensingConfig,
  computeTier,
  classifyObservationSignal
} from '@trillboards/edge-core';

// Device identity
const identity = new DeviceIdentity();
const fingerprint = await identity.getFingerprint();

// Capability detection
const tier = computeTier({ ramMB: 4096, hasNPU: false });
// → TIER_2

// Socket.io connectivity
const socket = new SocketManager({
  url: 'https://api.trillboards.com',
  deviceToken: 'your-token'
});
await socket.connect();

// Offline-resilient signal buffering
const buffer = new SignalBuffer({ dataDir: './data' });
await buffer.push({ type: 'audience_metrics', payload: { faceCount: 3 } });

// Signal classification
const signal = classifyObservationSignal(rawPayload);
// → { family: 'SCREEN_AUDIENCE', evidenceGrade: 'FACE_BACKED', ... }

Subsystem Health Monitoring

const health = new SubsystemHealthRegistry();

health.update('camera', 'healthy');
health.update('ml_pipeline', 'degraded');

const report = health.getReport();
// → { camera: 'healthy', ml_pipeline: 'degraded', ... }

Capability Matrix

import { getCapabilityMatrix } from '@trillboards/edge-core';

const matrix = getCapabilityMatrix(tier);
// → { models: ['blazeface'], inferenceHz: 2, resolution: 320, ... }

Key Types

• PlatformAdapter — interface for platform-specific adapters (Linux, Windows)
• CameraProvider — interface for camera capture implementations
• AudioProvider — interface for audio capture implementations
• AudienceMetricsPayload — structured audience measurement data
• VASResult — Viewability Attention Score output
• DetectedFace — face detection result with bounding box, age, gender, emotion
• EdgeQualityReport — inference quality telemetry

License

MIT