Ai Agents
An AI agent worker inside OTOM AFP Web — state-of-the-art models, running locally on your own PC or in the cloud, driving the simulation while you stay in the conversation.
Setting up an AFP simulation used to mean knowing where every parameter lives. In AFP Web you can simply say what you want — “place the whole programme, then run the thermal analysis” — and an AI agent does the work inside the application while you watch it happen.
Local LLM or cloud model — your choice
Local LLM
Point AFP Web at a state-of-the-art open model running on your own machine. Your geometry, G-code and process settings never leave the building — which is what most production data policies require.
Cloud models
Or connect a hosted frontier model with your own API key when you want the strongest reasoning on a complex lay-up. Both routes drive the same application, and you can switch between them whenever you like.
What the agent worker does
The agent is not a chatbot bolted onto a help page. It reads the live state of the application — the loaded tool geometry, the G-code programme and its courses, what is already placed, the analysis range and the head geometry — and then acts on it:
- Builds the ply — selects a course, maps its trajectory onto the tool, and accepts it into the lay-up; or processes the whole programme in the correct order in one go.
- Edits the lay-up — places or removes specific courses on request, such as “ply path 8 and 10” or “remove path 13”, and can reset back to the bare geometry.
- Changes settings — process parameters, head geometry, obstacles and the analysis range, set by name instead of hunted for in panels.
- Runs the analyses — optical analysis at the nip point, an optical sweep over the analysis range, a transient thermal run, and defect analysis for coverage, gaps, overlap and steering.
- Reports honestly — every action comes back as success or failure with a reason, so the agent corrects itself and tells you what actually happened rather than claiming work it did not do.
Why it changes the user experience
The value of a physics-based model is in the answer it gives, not in the clicking that precedes it. Letting an agent handle the mechanics shortens the distance between a question and its answer: an engineer new to AFP can produce a meaningful lay-up on the first afternoon, and an experienced user can run through variants far faster than by hand.
Because the agent works from the application’s own state rather than from a script, it also stops guessing: it reads what is loaded before it acts, and asks only the questions the state cannot answer.