System Architecture
Overview
Arvist Stack is an on-premises computer vision and safety platform designed for industrial environments. The system provides real-time monitoring, AI-powered analytics, and automated safety detection capabilities through a containerized microservices architecture.
All services run on edge devices deployed at customer facilities, with selective cloud integration for storage and remote access. The web application is accessible on port 80/443 (HTTP/HTTPS)* within the local network.
Architecture Diagram
System Components
Core Platform Features
The Arvist Stack provides three main categories of functionality:
🛡️ Safety & Compliance
- Configurable Safety Monitoring: Track, detect, save, and alert for safety incidents throughout your warehouse
- Existing Camera Infrastructure: Works with your current camera setup - no new camera installation required
- Incident Management: Automatic capture and storage of safety events with real-time notifications
- Compliance Reporting: Historical data and analytics for safety compliance
📦 Quality & Inventory
- PalletScan: Automated pallet and shipment inspection using computer vision
- Damage Detection: AI-powered quality checks for identifying damaged goods (uses External AI API)
- On-Device Processing: OCR, detection, and most processing happens locally on the edge device
- Inventory Tracking: Visual verification of shipment contents and conditions
📊 Productivity
- Worker Tracking: Monitor worker movement and productivity throughout the facility
- Analytics Dashboard: Real-time metrics and historical reporting
Technical Components
| Component | Purpose | Notes |
|---|---|---|
| Web Application | User interface and control panel | Accessible via browser on your network (WiFi or wired) |
| Arvist NVR | Video recording and camera management | Connects to your existing NVR system |
| AI/ML Services | Computer vision and object detection | Runs locally on edge device |
| Database | Event storage and historical data | PostgreSQL for structured data |
Storage Options
| Storage Type | Description | Retention |
|---|---|---|
| Cloud Storage (Default) | Media stored in Cloudflare R2 | Extended retention periods |
| On-Device Storage | Media stored locally on edge device | Limited retention due to device capacity |
Note: Storage configuration is determined during initial discussions based on customer requirements. Cloud storage is the default option; on-device storage available for security-sensitive environments with reduced retention times.
External Services (Optional)
| Service | Purpose | When Required |
|---|---|---|
| Cloudflare R2 | Cloud storage media files | Optional - can store locally only as well |
| Cloudflare Zero Trust | Secure remote access for Arvist team | Optional - alternative VPN solutions or on-premises monitoring |
| External AI API | Advanced damage detection for quality checks | Required only for PalletScan damage detection feature |
Network Access
The Arvist system integrates into your existing network infrastructure with minimal configuration.
Local Network Access
- Web application accessible on port 80/443 (HTTP/HTTPS) on your local network
- Users access the dashboard via standard web browsers on any network-connected device
- Edge device connects to your network to enable wireless access throughout your facility
- No special firewall rules required for internal operation
- Supports standard IP camera protocols (RTSP very highly recommended)
- Integrates with your existing Network Video Recorder (NVR)
Remote Access (Optional)
- Remote Maintenance Options:
- Cloudflare Zero Trust VPN (preferred by Arvist team)
- Alternative VPN solutions provided by customers
- On-premises monitoring (customer-provided or Arvist on-site support)
- Used by Arvist team for maintenance, updates, and support
- Encrypted connections without exposing services to the public internet
- SSH access for system maintenance and diagnostics
Network Requirements
- Gigabit Ethernet recommended for optimal video streaming performance
- Static IP addresses recommended for:
- Edge device (highly recommended)
- All IP cameras (recommended)
- Network access for cameras, edge device, and user workstations
- WiFi or wired connectivity for user devices accessing the web dashboard
- Bandwidth Requirements:
- Safety/Productivity cameras: 2-4 Mbps mainstream (1080p), 0.5-1 Mbps substream
- Quality cameras: 8-12 Mbps mainstream (4K), 0.5-1 Mbps substream
- Example: 10 cameras at 1080p = ~30-40 Mbps total
- Local network traffic only; camera streams don't traverse the internet
- Internet connection required for:
- Damage detection in quality inspections (External AI API)
- Cloud storage (when configured based on customer preference)
- Remote access by Arvist team for maintenance and support (optional - can use on-prem monitoring)
Data Flow
Safety & Productivity Workflow
Quality Inspection Workflow
Flow Characteristics
Safety & Productivity:
- Fully operates on-premises without internet dependency (when using on-device storage)
- Real-time processing with low latency
- Data stored based on configured storage option (cloud or on-device)
Quality Inspection:
- Most processing (detection, OCR) happens on-device
- External AI API used only for damage detection/quality assessment
- Media stored based on configured storage option (cloud or on-device)
- Detailed inspection reports and historical tracking
Deployment Model
Deployment Characteristics
- Edge-First Architecture: All core processing happens locally on customer premises
- Low Latency Processing: Video analysis and ML inference run on-device for real-time response
- Partial Offline Operation:
- Safety and productivity features work without internet (when using on-device storage)
- Quality inspection detection and OCR work offline; damage detection requires External AI API
- Cloud storage requires internet; on-device storage operates fully offline
- Storage Options:
- Cloud storage (default): Extended retention via Cloudflare R2
- On-device storage: Limited retention for security-sensitive environments
- Secure Remote Management: Remote access via VPN (Cloudflare Zero Trust or alternatives) for Arvist team maintenance and support, or on-premises monitoring if remote access is not permitted
- Containerized Deployment: All services run as Docker containers for easy updates and management
- Scalable Architecture: Modular design supports adding new features and capabilities
Security & Access
Local Access
- Web application accessible on port 80/443 (HTTP/HTTPS) within local network
- Direct LAN access for on-site workers and administrators
- No external firewall rules required for basic operation
Remote Access
- Cloudflare Zero Trust provides secure VPN and SSH tunneling
- Used by Arvist team for maintenance, updates, and support
- Alternative VPN solutions can be used if preferred
- On-premises monitoring available if remote access is not permitted (customer-provided or Arvist on-site support)
- Encrypted connections for remote administration
- No need to expose services directly to the internet
Container Security
- Services run in isolated Docker containers with network segmentation
- Separation between application services, database, and external access
- Resource limits and security constraints per container
Data Security
- Storage Configuration: Determined during initial discussions based on customer requirements (cloud default or on-device for security)
- Cloud Storage: Files encrypted during transmission and at rest in Cloudflare R2
- On-Device Storage: All media remains on customer premises with limited retention
- Database Security: Isolated database access with role-based permissions
- Access Control: Role-based access control (RBAC) in web application
- Audit Logging: Comprehensive activity logs for compliance and troubleshooting
Performance Considerations
Hardware Requirements
- GPU Acceleration:
- Intel GPU: Sufficient for Safety-only deployments
- NVIDIA GPU: Required for Quality features and/or warehouses with 40+ cameras
- Memory: Minimum 16GB RAM for ML workloads and video processing
- Storage: SSD recommended for video storage and database performance
- Network: Gigabit Ethernet for IP camera streams and smooth video playback
Performance Features
- Optimized ML Models: Computer vision models optimized for edge inference
- Efficient Processing: Background processing for non-critical tasks
- Resource Management: Docker resource constraints ensure system stability
- Caching: Smart caching for faster model loading and response times
Monitoring & Operations
System Monitoring
- Real-time system health monitoring via web dashboard
- Service status indicators and alerts
- Resource utilization tracking (CPU, memory, storage)
Logging & Diagnostics
- Centralized logging for troubleshooting
- Configurable log retention policies
- Remote log access via Cloudflare Zero Trust VPN
Maintenance & Updates
- Software Updates: Docker-based updates for easy deployment
- Backup & Recovery: Automated and manual backup options
- Configuration Management: Web-based configuration interface
- Health Checks: Automatic service restart on failures
Future Extensibility
The modular architecture of Arvist Stack supports continuous evolution:
- New Detection Models: Add custom computer vision models for specific use cases
- Custom Plugins: Develop safety or productivity plugins tailored to your operations
- API Integration: Connect to existing warehouse management or ERP systems
- Multi-site Management: Potential for centralized management of multiple installations
- Advanced Analytics: Expand reporting and analytics capabilities