I built an open-source, real-time MQTT viewer for UNS (github.com)

A developer released an open-source, containerized real-time MQTT viewer tailored to Unified Namespace (UNS) deployments that visualizes MQTT topic trees and drives dynamic SVG floorplans. The single Docker image runs two Node.js processes: a web server that subscribes to MQTT, broadcasts messages to the frontend via WebSockets, manages a persistent DuckDB message history, and a child MCP Server that exposes structured tools and controls for LLM agents. Key features include automatic hierarchical topic discovery with live filtering, a dynamic SVG view with a History Mode for time-replay, advanced history filtering (keyword highlighting + dual time-range slider), optional Sparkplug B decoding, built-in data simulator, optional MTLS/ALPN support, HTTP Basic Auth, and a web-based configuration editor. For the AI/ML community this is significant because it bridges real-time industrial/IoT telemetry and LLM-based agents: the MCP Server provides a programmable, structured interface so agents can query state, invoke controls, or reason over historical traces. DuckDB persistence and configurable pruning make it practical for dataset capture and offline analysis, while the simulator helps generate synthetic data for model testing. The SVG-mapping approach (SVG group IDs map to MQTT topics; data-key attributes bind JSON fields to text elements) yields a low-effort digital-twin visualization pipeline suitable for rapid prototyping, observability, and agent-in-the-loop automation.