getting-started_skill

Overview

This skill provides a quick start and basic usage patterns for Vuer, a Python-based real-time 3D visualization tool used in robotics and AI projects. It explains the minimal async backend setup, how to create a scene with basic primitives, and the core session API for driving live updates. The goal is practical: get a working browser-based 3D view quickly and show common update patterns.

How this skill works

Vuer runs an async Python backend (aiohttp) that communicates with a browser client over WebSocket, rendering via Three.js. A VuerSession exposes simple commands to set the root scene and to add, update, or remove named elements. The typical flow is: spawn a session, send an initial scene, then push incremental updates while the session runs.

When to use it

• Rapidly prototype 3D scenes for robotics simulations or visualization pipelines
• Stream sensor data or state updates to a web-based viewer in real time
• Inspect spatial relationships and debugging visuals during algorithm development
• Share interactive views of robot state across a team without heavy tooling
• Integrate lightweight WebXR/Three.js visuals into Python apps

Best practices

• Use descriptive keys for elements so updates and removals are unambiguous
• Initialize the scene once with session.set and use upsert/update for frequent changes
• Keep element payloads small; send transforms and state deltas instead of full meshes when possible
• Run the session loop in an async task and await session.forever() to maintain the websocket
• Use add/remove for topology changes and update/upsert for per-frame changes to avoid flicker

Example use cases

• Visualize robot poses and sensor frames as boxes, spheres, and lines streamed from a controller node
• Show live object detections by upserting bounding primitives keyed by object id
• Prototype VR inspection tools by exposing the same scene to WebXR-enabled browsers
• Debug kinematics by animating joint transforms and highlighting collisions
• Create dashboards that display 3D maps and tracked targets with minimal front-end code