Demos
Inference • model status
UNKNOWNIDs: anonymized
Model Gallery
Live view of the cluster's model endpoints. In production, the backend and identifiers are anonymized before this page sees them.
Active
—
running models
Catalog
—
total endpoints
Cluster
— GPU
— pods running
Quick tour
What it shows
- A live view of model endpoints exposed by the cluster.
- Type + status badges and a rough parameter-size visualization.
- Anonymized identifiers so the public feed stays safe to share.
How it’s built
- A read-only public inventory endpoint that strips sensitive fields.
- Client-side filtering and lightweight motion for instant feedback.
- Optional sample models so you can try the UI when the feed is empty.
What to look for
- Running vs loading/stopped mix (and whether it matches expectations).
- Type distribution (LLMs vs embeddings vs image models).
- Empty feeds that usually mean “exporters aren’t configured”, not “nothing is running”.
Notes
- If you toggle sample models, the link includes sample=1 so it’s shareable.
- Filters are stored in the URL so you can bookmark a specific view.