The Asset Wizard turns a process that usually takes days — identify equipment, find documentation, organize knowledge — into a guided, AI-assisted experience. No empty forms. The system anticipates, suggests, and validates at every step.
Every step feeds the next. Zero manual configuration.
Take a photo of your equipment's nameplate. Claude Vision extracts the manufacturer, model, serial number, and year of manufacture automatically. In seconds, with zero manual entry.
Aviation, CNC Machining, Automotive, Robotics, Marine, Medical Devices, Energy, and more. The moment you select a domain, the system adapts manufacturers, models, subsystems, and search angles. An aviation technician never sees CNC manufacturers.
An LLM call generates the list of relevant manufacturers — global leaders, mid-tier players, and niche specialists. Each manufacturer is classified by tier with a visual badge. A free-text field lets you enter any unlisted manufacturer.
The LLM generates a complete model list grouped by category, tagged entry/prosumer/pro. Both current and legacy models still in service are included — because many machines in the field are 10, 20, 30 years old.
For each piece of equipment, the LLM identifies 8 to 12 subsystems. Each chip is clickable to enable or disable. Only activated subsystems are used for documentation search — less noise, more relevance.
The Corpus Builder launches 44 queries per angle — Maintenance, Operation, Training, Parts, and All — covering manufacturer, model, subsystems, and technical terminology. 30 to 80 PDFs found and ranked in 5 to 15 minutes.
The KG Pipeline extracts knowledge from each document into structured Knowledge Units. Three presets: Economy (fast), Optimal (balanced), Max Quality (multi-pass with validation). Each KU contains a fact, its source, and confidence metadata.
Drag & drop zone for PDF, Word, Excel, images, CSV, TXT. Associate each document at the fleet level, per model, or per specific machine. Internal documents contain the information that public documentation doesn't have.
The RAG Chat includes an asset selector. Context — manufacturer, model, callsign, subsystems — is injected into every question. Switch equipment in one click. Every answer cites its source PDF with the exact page.
Score out of 100 with a GO / CAUTION / NO GO verdict. Five criteria evaluated: assets added, sufficient corpus, internal documents, subsystems covered, KG generated. Detailed breakdown per asset with clear actions for gaps.
The Wizard anticipates what you need at every step.
Every step uses previous choices to preconfigure the next. Domain determines manufacturers. Manufacturer determines models. Model determines subsystems. Nothing is generic.
Language model calls run in parallel while you interact. You never see an empty screen or a loading spinner blocking your workflow.
At every step, a free-text field lets you bypass the suggested list. The LLM adapts to any input — listed or not.
Chips that change opacity, real-time progress bars, status badges, counters. Every action produces an immediate visual response.
Asset deletion requires password confirmation in a secure modal. Destructive actions are deliberately slowed down. Your data is safe.
The LLM understands French, English, and German. The Knowledge Graph is multilingual. Ask in any language, get grounded answers.
Try the live demo with 35,606 Knowledge Units, or request a demo with your own technical documents.