Welcome to Acusight
Acusight is an integrated MLOps and container management platform designed to streamline the entire workflow from edge data collection to model training and infrastructure management. It provides a unified interface to manage Docker environments via Portainer while simultaneously offering a powerful, end-to-end data lifecycle pipeline for computer vision tasks.Key Features
Unified Management
A single pane of glass for container orchestration (via Portainer) and MLOps data workflows.
Edge Device Integration
Securely register, approve, and manage edge devices with automated batch creation for incoming data.
Video & Timeline
CCTV-style multi-device WebRTC streaming grid with a global timeline for browsing historical events.
End-to-End Data Lifecycle
Sophisticated two-tier system moving data from raw ingestion through curated project pipeline.
Dataset Versioning
Create atomic, point-in-time snapshots with train/val/test splits for reproducible training.
Real-time Observability
Live monitoring via WebSockets with integrated Prometheus metrics and OpenTelemetry tracing.
Architecture Overview
The Acusight platform is built on a microservices architecture:- Core Service: API gateway handling auth, device management, and Portainer proxy
- Data Service: MLOps data pipeline for batches, annotations, and datasets
- ML Service: Model training orchestration with MLFlow integration
- Portainer: Container management for edge devices
Technology Stack
| Component | Technology |
|---|---|
| Frontend | React, TypeScript, Vite, Tailwind CSS |
| Backend Services | Go, Gin, GORM |
| Database | PostgreSQL |
| Cache / Jobs | Redis |
| Object Storage | MinIO (S3-compatible) |
| Monitoring | Prometheus, Grafana, Alertmanager |