| name | 3d-modeler |
| description | Generates 3D interior models from photos and injects them into Mapbox GL JS map instances. |
| version | 1.0.0 |
| author | Ahmed Qaddoura |
| dependencies | ["@lumaai/luma-web","three","threebox-plugin"] |
3D Modeler Skill
This skill allows agents to convert photographs of interiors into 3D models (GLB/GLTF) and inject them into web-based Mapbox 3D styles using Three.js and the Threebox plugin.
Capabilities
- Photo-to-3D Generation: Uses
photoTo3D.js to call 3D reconstruction APIs (e.g., Luma AI / Polycam) to process an image and return a URL to a GLB file.
- Mapbox Injection Prep: Uses
mapboxInjector.js to prepare a GLB file for Mapbox insertion, calculating the necessary MercatorCoordinates and Three.js scale/rotation matrices to align the interior model perfectly with the real-world map.
Usage
When a user uploads a photo of a room and asks to place it on the map, trigger this skill:
- Pass the image URL to
photoTo3D.js. Wait for the 3D generation to complete.
- Pass the resulting GLB URL and target LngLat coordinates to
mapboxInjector.js to get the integration payload.
- Send the payload back to the frontend to render the model inside the Mapbox custom layer.
