Our Enterprise-Grade Architecture
Engineered over seven years, our powerful platform uses a C and Rust tech stack for flexible deployment on the edge, on-premise, or multi-cloud. It's agnostic to AI chips, camera brands, and connectivity, built to orchestrate workloads across a decentralized network of 1,000 to 50,000 cameras.
High-Level Architecture Diagram
Get a complete overview of our 21 microservices, distributed across three distinct layers: Edge, Cloud, and Client Apps.
Hover over any component to see its description.
Explore Each System Layer
Select a layer to explore its components in detail. Each microservice has a specific job within the overall architecture.
VSaaS.ai Edge Software Suite
Our suite of microservices deployed on-site—on anything from OpenWRT routers to GPU servers. It captures video, runs AI detections, and syncs results with the cloud, all designed to run smoothly with limited or intermittent connectivity.
XM-Cameras
C / RTSP / ONVIFOur video capture module connects to IP cameras, NVRs, DVRs, and VMS via RTSP, Snapshot, and HTTP. It manages video streams and routes them to the right processing modules.
XM-Detections
C / RustThe AI detection engine processes each video frame and executes assigned detection models. It works with the inference server to balance the load between the GPU, CPU, and NPU.
XM-Sensors
C / JSON ConfigManages virtual sensors and camera-based triggers. You can define detection zones, tripwires, areas of interest, and activation rules for every analytic.
XM-Upload
Rust / HTTP2This data upload module sends processed data to the cloud. It manages the send queue, compresses evidence images, and automatically retries if connectivity fails.
XM-Events
C / Event BusGenerates and manages events from detections. It applies business rules, confidence filters, and time-based grouping to produce meaningful alerts.
AI Inference Server
C / CUDA / TensorRTOur AI inference server runs compiled C models on GPU, CPU, or NPU. It supports multiple simultaneous models and assigns tasks to the optimal processor.
Sync Service
Rust / gRPCA two-way service that syncs data between the edge and the cloud. It keeps configurations, models, and rules updated at each site, even with spotty connectivity.
Data Flow
See how data travels from the camera to the end-user, passing through edge processing and cloud intelligence.
Video Capture
IP cameras send RTSP/HTTP streams to the XM-Cameras module, which decodes and extracts frames for processing.
AI Detection at the Edge
XM-Detections runs AI models compiled in C on the AI Inference Server. Virtual sensors (XM-Sensors) define what to detect and where.
Event Generation
XM-Events processes detections, applies business rules, and generates significant events. XM-Upload then preps the data for cloud upload.
Edge-to-Cloud Sync
Data travels from the edge to the Bridge API in the cloud, while the Sync Service keeps configurations and models updated in both directions.
Cloud Processing
The Main DB stores the data, the False Positive service filters bad alerts, Object Storage saves evidence, and the ETL service crunches data for analytics.
User Access
Users access the platform via the Web App or Mobile App. The Real-time Service sends instant push notifications for new events.
A Robust and Scalable Tech Stack
Every component in our architecture was chosen and developed to deliver maximum performance, reliability, and scalability.
7+ Years of Development
An architecture matured over seven years of continuous development, production-proven with thousands of cameras across multiple industries.
Multi-Cloud & On-Premise
Deploy in the public cloud (AWS, GCP, Azure), a private cloud, or completely on-premise. The same codebase works in any scenario.
Hardware-Agnostic
Works with any AI chip (NVIDIA, Blaize, Axelera, Hailo, DeepX), any camera brand, and any network setup.
Scales from 1K to 50K Cameras
Decentralized workload orchestration lets you scale from 1,000 to 50,000 cameras with distributed computing across multiple nodes.
High-Performance C & Rust
Critical components are built in C and Rust for maximum performance, memory safety, and minimal resource consumption.
SQL + MongoDB
A hybrid data architecture using PostgreSQL for transactional and relational data, and MongoDB for flexible metadata and dynamic configurations.
From Router to Server
Runs on everything from OpenWRT routers to servers with Linux/Ubuntu and dedicated GPUs. Adapts to any infrastructure.
Decentralized Orchestration
A decentralized network of edge nodes that operate autonomously and sync with the cloud. Each node can run independently during connectivity loss.
Compatible with Any Infrastructure
Ready to see our architecture in action?
Request a personalized demo to see how VSaaS.ai can transform your video security. Our team will show you how our technology adapts to your specific needs.