Run Industrial AI
on Your Own Server

May 26, 9AM CST

Industrial AI is moving fast. Most of the conversation assumes the AI lives in the cloud. That assumption is no longer necessary.

FrameworX 10.1.5 ships with a built-in AI Runtime that runs a language model directly on the same machine as the SCADA runtime. No cloud subscription required. No data leaving the plant network. Configurable for cloud LLMs only when you want them.

This works alongside the other major advance in 10.1.5: a new Knowledge Graph layer that makes your Unified Namespace ontology-aware. The AI Runtime reads tags that carry meaning, not just names. Search by what it is, not by what it is called.

What you will see live

  • Local AI Runtime, running on the SCADA server. Install Ollama, run the setup script, and your plant has an AI assistant on the same machine as the runtime. Connected to live tag values, active alarms, and historian queries. The AI reads your data and answers questions about your actual plant.
  • Industrial Ontology import — ISA-88, IOF, and any RDF/OWL graph. Bring established industrial vocabularies directly into your Unified Namespace. The Designer wizard or an AI assistant takes the ontology file and FrameworX materializes typed UDTs, typed tags, and AssetTree folders automatically. No manual folder writing.
  • Live knowledge graph on the operator screen. Drop the knowledge graph chart onto any HMI screen and the imported ISA-88 model renders as an interactive view — pan, zoom, click through the equipment hierarchy. Same control in the WPF rich client and in the HTML5 browser. Closes the loop: import the ontology, then see it on the screen the operator uses.
  • GIS and AutoCAD import to live HMI. Drop in GeoJSON, Shapefile, KML, KMZ, DXF, or DWG files. Tags, Asset Tree folders, and auto-generated Displays come out the other side. The structure comes from the ontology. The content comes from your existing engineering drawings.
  • MongoDB and QuestDB across three roles each. The same MongoDB or QuestDB instance can serve as a Dataset source, a Device with UNS TagProvider, and a Historian Storage Location. Two new database integrations, three substrates each, one architecture.
  • Two AI assistants. Two Designer instances. One plant. Watch concurrent Claude sessions edit the same FrameworX project through the DesignerMCP and ConsoleMCP servers. The workspace flow turns a folder of JSON files into a running solution.
  • AI safety boundary by architecture. The Runtime AI reads tags, alarms, and history. It does not write setpoints. It does not acknowledge alarms. The safety property is built into the platform, not bolted on.

 

Who should attend

This workshop is for engineers, plant managers, IT leaders, system integrators, OEMs, and anyone evaluating where industrial AI is going. You do not need prior FrameworX experience to follow along. Live demos are presented for engineers and architects, but the implications matter for operations, security, and procurement teams as well.

Presenter

Marc Taccolini — Founder and CEO of Tatsoft, founder of InduSoft (1997, the engine that became AVEVA Edge HMI), 30+ years of industrial software architecture experience.