Skip to Content
Physical AIOverview

Physical AI

IronFlock includes an AI assistant built for industrial operations. It connects directly to your edge devices, your production data, and the apps running across your fleet. It can do three things that set it apart from generic AI tools:

  • Query — Ask questions about your production data in natural language. The AI explores your time-series database, investigates incidents, analyzes historical trends, and generates live charts — all without writing a single query.
  • Act — Go beyond answers. The AI can call functions running on your edge devices in real time: read a sensor, send a control command, scan a network, adjust a setpoint. It closes the loop between insight and action.
  • Extend — Install an app that ships with AI agents, and the assistant automatically gains new domain-specific skills. No configuration required — new capabilities appear the moment the app runs on a device.

Query Your Production Data

The IronFlock AI has direct access to all time-series data collected by your devices. Ask a question in plain language, and the AI translates it into a database query, runs it, and presents the result — as a number, a table, or a live chart widget embedded right in the conversation.

This covers the everyday questions that otherwise require manual dashboard browsing or custom SQL:

  • “What was the peak temperature on line 3 last night?”
  • “Show me power consumption across all devices for the past week.”
  • “Were there any error codes reported by device A yesterday between 8 and 10 AM?”
  • “Compare vibration levels on machine 1 vs. machine 2 over the last 30 days.”

The AI works with the Built-in Tools — Data Explorer, Widget Generator, and Dashboard Generator — which are always available in every project regardless of which apps are installed.

Act on Physical Devices

Most AI assistants stop at answering questions. IronFlock’s AI can execute functions on real devices at the edge of your network.

Every app deployed on an IronFlock device can register AI-callable functions via the IronFlock SDK. When the app runs, those functions become available to the AI over the WAMP messaging layer. The AI invokes them in real time, just like a human operator would — but faster, and in response to natural language.

Ask → AI reasons and selects an action → Device function executes → result flows back → AI responds

Read a sensor value. Send a control command. Scan a production network. Adjust a setpoint. The AI doesn’t just tell you what happened — it can intervene.

Extend via App Agents

The AI assistant is not a fixed set of capabilities. It grows with your fleet.

When you install an app that includes AI agents, those agents automatically register with the IronFlock AI. The assistant delegates questions to the right agent based on context — you don’t select or configure anything manually. Each agent brings a set of tools that map directly to functions running on your devices.

For example, an OPC UA connectivity app exposes tools for scanning production networks, reading live registers, and configuring data collection. Ask “which PLCs are reachable on the line 3 subnet?” and the AI calls the scan function on the relevant edge device and reports back with real results.

Install more apps, and the AI learns more skills — automatically.

See App Agents for details on how agent delegation works.

For Developers

If you’re building apps and want to add AI agents that execute functions on your devices, see:

Last updated on
IoT AlarmsBuilt-in Tools