The Sentinel — AI-Powered Edge Security Monitoring Platform
A cloud-native, privacy-first security monitoring system that runs real-time AI object detection on live video streams, syncs events to Azure Blob Storage, and supports multi-tenant deployments across distributed edge devices.
PRIVASEE AI is a single-page application with an Express backend, designed for edge deployment on any device with a camera.
git clone https://github.com/aurelianware/privaseeAI.git
cd privaseeAI
docker build -t privaseeai .
docker run -p 8080:8080 privaseeai# Prerequisites: Node.js 20+, npm
git clone https://github.com/aurelianware/privaseeAI.git
cd privaseeAI
npm install
# Copy and fill in environment variables
cp .env.example .env.local
# Start Vite dev server (frontend)
npm run dev
# In a second terminal — start Express backend
npm startFrontend: http://localhost:3000 · Backend: http://localhost:8080
- COCO-SSD / TensorFlow.js — in-browser WebGL object detection, zero server-side inference latency
- Detection overlays — colour-coded bounding boxes rendered on a canvas stream capture
- Confidence scoring — per-object confidence percentages, configurable threshold
- Annotated captures — JPEG snapshots and WebM/MP4 recordings with detection overlays baked in
- Severity classification — critical / high / medium / low with Sentinel colour coding
- IndexedDB local storage — full offline capability with proper IDB initialisation and reopen on close
- Azure Blob sync — background upload queue with SAS token auth; media blobs preserved with original MIME types
- Real-time event list — live "time ago" ticker, media playback modal, error surface for
MediaErrorcodes
- Per-user encrypted settings — AES-256-GCM server-side settings API, keyed by Azure AD
oid - Azure PostgreSQL —
UserSettingstable, Prisma schema, raw SQL migration included - MSAL v5 —
@azure/msal-browser+@azure/msal-react; Auth0 also supported
- IP / RTSP camera management — add cameras by ID + RTSP URL; HLS transcoding via server
- AGM Taipan V2 thermal — auto-detect probe
- Autel EVO Lite drone SDK — takeoff/landing, waypoint missions, RTH, emergency land via WebSocket relay
| Layer | Technology |
|---|---|
| Frontend framework | React 18 + TypeScript 5 |
| Build tool | Vite 7 |
| Styling | Tailwind CSS + inline Sentinel design tokens |
| AI inference | TensorFlow.js · COCO-SSD (WebGL) |
| Authentication | MSAL v5 (Azure AD) · Auth0 |
| Backend | Node.js · Express |
| Database ORM | Prisma 7 |
| Database | Azure Database for PostgreSQL (Flexible Server) |
| Media storage | Azure Blob Storage (SAS, CORS) |
| Containerisation | Docker (multi-stage, node:20-alpine runtime) |
| Deployment | Azure Container Apps (via GitHub Actions CI/CD) |
| Container registry | Azure Container Registry (ACR) |
| IaC | Azure Bicep (deploy/azure-app-service.bicep) |
privaseeAI/
├── src/
│ ├── App.tsx # Root component — Sentinel UI shell
│ ├── components/
│ │ ├── CameraStream.tsx # WebRTC camera + TensorFlow inference loop
│ │ ├── DetectionOverlay.tsx # Canvas bounding box renderer
│ │ ├── EventsList.tsx # Security event feed with media playback
│ │ ├── HlsVideoPlayer.tsx # IP camera HLS stream player
│ │ ├── MissionDashboard.tsx # Drone mission control UI
│ │ ├── SettingsPanel.tsx # User settings + Azure config
│ │ └── Auth.tsx / Auth0Components.tsx / AuthProvider.tsx
│ ├── hooks/
│ │ └── useUserSettings.ts # AES-256-GCM encrypted settings hook
│ ├── drone/ # Autel EVO Lite SDK adapters
│ └── utils/
│ ├── storage.ts # IndexedDB local event store
│ └── syncQueue.ts # Azure Blob upload queue
├── prisma/
│ ├── schema.prisma # UserSettings model
│ └── migrations/ # Raw SQL migrations
├── public/
│ ├── logo/ # Animated logo (WebM + MP4 + poster)
│ ├── privaseeai-kubrick.png # Brand hero image
│ └── privaseeai-brand.png # Chrome 3D brand render
├── deploy/
│ └── azure-app-service.bicep # Azure Bicep IaC
├── infra/
│ └── blob-lifecycle.json # Blob storage lifecycle policy
├── docs/
│ ├── AZURE_DEPLOYMENT.md
│ ├── DEPLOYMENT_GUIDE.md
│ ├── SECURITY.md
│ ├── DRONE_INTEGRATION.md
│ ├── DRONE_SETUP.md
│ └── STRIPE_SETUP.md
├── .github/workflows/ # CI/CD: build-and-push → deploy-aca
├── server.js # Express backend (settings API, HLS proxy)
├── Dockerfile # Multi-stage production build
└── vite.config.ts
Push to main triggers a two-stage GitHub Actions pipeline:
build-and-push.yml— builds the Docker image, tags with$GITHUB_SHA, pushes to ACRdeploy-aca.yml— updates the Azure Container App with the new image tag
Required GitHub secrets: AZURE_CLIENT_ID, AZURE_TENANT_ID, AZURE_SUBSCRIPTION_ID, AZURE_RESOURCE_GROUP, ACR_NAME, ACA_ENVIRONMENT.
See docs/AZURE_DEPLOYMENT.md for full setup.
- Real-time COCO-SSD detection with canvas overlay recording
- Azure Blob Storage cloud sync with SAS auth
- Multi-tenant AES-256-GCM encrypted user settings
- Azure PostgreSQL backend with Prisma
- MSAL v5 Microsoft Entra ID authentication
- Sentinel brand redesign (dark glass-morphism, Kubrick aesthetic)
- Animated logo video integration
- IP / RTSP multi-camera management with HLS
- Autel EVO Lite drone SDK integration
- Docker + Azure Container Apps CI/CD pipeline
- YOLO v8 model swap for higher accuracy
- WebRTC peer-to-peer multi-device streaming
- Push notifications (Azure Notification Hubs)
- Mobile native app (Capacitor / iOS)
- Stripe billing for SaaS tiers
- Audit log export (SIEM integration)
- Cloud Dental Office — SaaS dental practice management with integrated privaseeAI vision service for narcotics cabinet monitoring, consent recording, and insurance card OCR
- Together, privaseeAI + Cloud Dental Office form a full provider-side AI vision + compliance stack
Apache License 2.0 — Copyright 2026 Aurelianware