ViPPET’s Architecture#

This section summarizes the core components ViPPET is built from. These components are grouped into ViPPET application specific services and foundational services, which provide platform-wide capabilities.

ViPPET architecture

For the Voice integration, see STT, TTS and metrics architecture for C4 System Context, Container and Component diagrams and runtime flows.

Application Specific Microservices#

Application-specific microservices implement ViPPET’s end-user workflows. Together they provide the web UI, the backend control plane for pipelines and jobs, and the camera discovery path used to onboard USB and ONVIF-compatible devices.

Microservice

Description

Docs

API

alt text

ViPPET UI microservice provides the web-based React interface for user interaction. It integrates with backend and foundation services through secure API calls.

Docs

N/A

alt text

ViPPET BE (Backend) microservice orchestrates workflows and exposes core application APIs. It manages user requests, coordinates jobs, and routes analytics results to UI.

Docs

API

alt text

Sensor Manager microservice discovers USB cameras and ONVIF-capable network cameras on the local subnet and loads their ONVIF media profiles.

Docs

API

Middleware Microservices#

Middleware microservices provide shared infrastructure capabilities reused across all Edge AI suites, including stream ingestion, model lifecycle management, and operational observability.

Microservice

Description

Docs

API

alt text

Model Download microservice is a centralized model management system that downloads AI or machine learning models from various model hubs while ensuring consistency and simplicity across applications, stores the models, accepts custom model uploads, and handles optional format conversions.

Docs

API

alt text

Metrics Manager service for unified collection, ingestion, and real-time relay of system and application metrics on edge and cloud nodes. It bundles Telegraf-based hardware telemetry (CPU, memory, temperature, Intel® GPU, Intel® NPU) with a FastAPI REST surface that accepts custom metrics in JSON, InfluxDB Line Protocol, and OpenTelemetry formats, and exposes them through a Prometheus-compatible endpoint as well as a Server-Sent Events (SSE) stream suitable for live dashboards.

Docs

API