Every factory generates vast amounts of data — but it's fragmented across disconnected systems that don't talk to each other. The UNS Framework replaces point-to-point integrations with a single, shared data layer — giving every system, dashboard, and AI agent real-time access to your shop floor.
The Unified Namespace (UNS) replaces point-to-point integrations with a single, shared data layer. Every system — machines, sensors, applications, databases — publishes and subscribes to a common MQTT topic hierarchy.
Every new connection adds complexity. Every system has its own data model. Getting a simple answer — "what's the utilisation of cnc-01 this shift?" — requires pulling data from multiple systems.
Any system can publish data. Any system can subscribe. No point-to-point wiring. No vendor lock-in. Adding a new consumer means subscribing to existing topics — not building a new integration.
Data is organised using the ISA-95 international standard — a structured path from enterprise down to individual data points. The hierarchy is self-describing.
AI in manufacturing fails when it can't access clean, structured, real-time data. The UNS solves this — it gives AI agents the same structured context that humans get, through the same APIs, in real time.
Every machine's state, every stoppage reason, every production run — structured in PostgreSQL with precise timestamps. AI agents read the same data as your KPI dashboards. No data wrangling, no export pipelines, no stale spreadsheets.
AI results publish back to MQTT — so an anomaly detection agent or a shift summary agent is just another participant in the UNS. Other systems can subscribe to AI output topics the same way they subscribe to machine data. The AI becomes part of the architecture, not bolted on top.
Every AI call is logged — the data that went in, the response that came out, token counts, latency, and the complete structured result. Traceability from raw machine data to AI recommendation.
Supports OpenAI and Anthropic · Structured JSON responses · Results publish to MQTT
The UNS isn't just a technical architecture — it's a business tool. Real-time visibility, reduced integration cost, and faster time to insight for every role on the factory floor.
| Metric | Before UNS | With UNS |
|---|---|---|
| Time to answer "what's machine utilisation?" | Hours (manual data pull) | Seconds (API call) |
| New integration setup | Weeks (custom development) | Minutes (subscribe to topic) |
| Data freshness | Hours/days (batch) | Real-time (< 1 second) |
| Systems with access to machine data | 1-2 (SCADA, historian) | Unlimited (any MQTT subscriber) |
| Cost of adding a new KPI | Significant (cross-system) | Minimal (new SQL query) |
Real-time dashboards showing machine utilisation, production progress, and stoppage reasons across the entire shop floor.
Automatic MTBF and MTTR calculations per machine. Alarm history with durations. Stoppage pareto charts showing where to focus.
Actual throughput data — parts per hour, target attainment — compared against planned schedules. Identify bottleneck machines.
Scrap tracking linked to specific machines, programs, and operators. Quality check pass rates over time.
Data-driven kaizen. Every state change, every stoppage, every production run is recorded with timestamps and durations.
A clean, maintainable architecture. Each function is independent, version-controlled, and deployable via git push. No vendor lock-in.
An open standard for structuring manufacturing data, an open platform for deploying serverless functions, and a reference implementation that brings them together.
An open standard that defines how to structure manufacturing data in MQTT using YAML files. ISA-95 topic hierarchy, versioned namespaces, payload conventions, and best practices. Vendor-neutral — any system that follows the standard can participate.
A lightweight Functions-as-a-Service platform that runs anywhere Containers run. Write functions in any language — Go, Node.js, Python — scaffold with one command, deploy via git push. No cloud vendor required.
A complete reference implementation of a UNS — open source serverless functions you can learn from, fork, and extend. Covers the full pipeline from machine simulation through to AI-powered analysis, dashboards, and a data historian. Built on fnkit, follows the UNS Framework standard.
Learn what a Unified Namespace is, read the open standard, or dive straight into the fn-uns reference implementation with real serverless functions.