Nox-Lumen Mfgimage_tools
In one sentence
Three native vision tool groups—invoked uniformly—so the AI need not care which model backs them:
| Tool | Purpose | Typical scenarios |
|---|---|---|
image_understanding | Describe images, OCR, layout | Patent figures, CAD screenshots, flows, tables |
image_modification | Edit images (annotate, crop, stitch, repaint) | Arrows on figures, highlight defect photos |
image_storage | Store and retrieve images in the KB | Persist session images into KB search |
Relationship to Parser SKILL
For figures inside documents, the platform auto-runs image_understanding with the KB’s chosen VLM prompt pack.
Calling image_tools explicitly in a SKILL targets individual images with finer control over prompts and output shape.
Industry usage
- Patents: figure numbering, block diagrams, trademark comparison
- Automotive: circuit markup, ECU pinout reading, CAN waveform screenshots
- Manufacturing: CAD dimension extraction, defect vision, SCADA alarm screenshots