System overview
Scenaven splits the work in two. The server keeps your simulation, converts it, and runs filters and Python. The browser shows an interactive 3D view of the surfaces the server sends back.
┌─────────────────────────────────────────────────────────────┐
│ Browser │
│ 3D view · Scene (OpenUSD tree) · Viewport · time controls │
└───────────────────────────┬─────────────────────────────────┘
│ upload, filters, geometry
┌───────────────────────────▼─────────────────────────────────┐
│ Server │
│ Convert · filter · time · Python │
└───────────────────────────┬─────────────────────────────────┘
│
┌─────────────┴─────────────┐
▼ ▼
Cloud storage On-prem storage
Why the work is split
| Work | Where it runs | Why |
|---|---|---|
| Reading VTK files, clip, warp, extract | Server | The full simulation is often too large for a browser tab |
| Building the scene the viewer can load | Server | One description of the session, shared by the view and by Python |
| Camera, colors, probes, and the scene tree | Browser | These stay responsive because they do not recompute the simulation |
| Keeping the converted scene | Cloud or on-prem storage | The same viewer, with storage that matches where you run Scenaven |
What you work with
The viewer is the full-screen 3D view and the controls around it: the Scene card (OpenUSD stage tree and Selected prim inspector), Viewport settings, viewport toolbar (camera, display, time), and dataset info where enabled. After you sign in, amAIra and the Python session sit on the right.
The server accepts VTK files (.vtk, .vtkhdf, .vtp, .vtu), turns them into a scene, applies clip, warp, and extract, and runs Python against the simulation you loaded.
The home page signs you in and opens that viewer with your session.
Cloud and on-prem
The same viewer and the same conversion run in both places. What changes is where the files live and which features your connection allows.
| Mode | Storage | Typical use |
|---|---|---|
| Cloud | Hosted storage | Examples on hosted cloud; uploads, server-side clip/warp/extract, Python, and amAIra on your files on on-prem |
| On-prem | Disk on the Docker host | Your own files; backend via Infrastructure → On-prem |
Related
- VTK, OpenUSD & Three.js — what each technology is for
- Why OpenUSD? — why the server describes the scene this way
- Simulation-to-view handoff — what the server–browser contract preserves (and what it does not yet)
- Why Three.js? — why the browser draws with Three.js
- Data flow — what happens when you load, filter, or play time
- Large meshes in the browser — display-style limits for dense surface meshes
- Cloud connection / On-prem connection
- Trust — data privacy and on-prem boundaries