Skip to main content

Welcome to Observable

Observable is an industrial-grade, 24/7 network monitoring platform designed to observe the uptime and latency of the OpenML ecosystem.

Unlike traditional status pages that rely on stagnant lists and spreadsheets, Observable treats microservices as living, interacting entities. By leveraging human visual processing through a 3D knowledge graph, it allows operators to pinpoint outages and performance degradation within a fraction of a second.

System Architecture​

Observable is built as a decoupled, high-performance system composed of two primary layers:

1. The Backend (The Measurement Apparatus)​

Built with Go, the backend is a high-concurrency daemon designed for 24/7 operation.

  • Concurrent Polling: Utilizes goroutines to monitor dozens of endpoints simultaneously, ensuring that the time taken to check the entire ecosystem is only as long as the slowest single request.
  • Embedded Persistence: Uses a pure-Go SQLite implementation (modernc.org/sqlite) to store high-velocity time-series data without the overhead of external database servers.
  • Stateless API: Exposes a lean JSON API that serves the latest telemetry data to the visual layer.

2. The Frontend (The HUD)​

A high-fidelity Next.js application that serves as the operator's Head-Up Display (HUD).

  • 3D Topology: Uses Three.js and react-force-graph-3d to render the ecosystem as an organic, physics-driven knowledge graph.
  • Quantum Aesthetic: Implements a "Quantum Void" design language featuring sharp glassmorphism, military-grade HUD brackets, and monospace telemetry readouts.
  • Static Export: Optimized for performance and security using Next.js static HTML export.

Key Features​

  • Real-Time Observation: Continuous background monitoring independent of user activity.
  • Interactive Physics: Nodes and links react organically like physical springs, allowing for intuitive exploration of complex network topologies.
  • High-Velocity Analytics: Efficiently aggregates millions of rows of data to calculate 24-hour average latency and stability metrics.
  • Zero-Dependency Deployment: Statically compiled binaries and containerized workflows for seamless operations.

Getting Started​

Ready to deploy the Observer? Head over to the Deployment Guide to stand up your own monitoring stack using Podman.