Overview
Status (Building, Sleeping, Ready, Failed), source (Hub or artifact), pipeline, and pricing plan (Per hour / Per token). See Core concepts.Usage
Cost, request count, workers over time, tokens (LLMs), latency, and pipeline metrics. The URL may say?tab=usage; older copy called this Analytics.
Use this with Autoscaling and Optimizing costs.
Logs
Operational events and errors. Start here for launch or inference failures — Troubleshooting.Instances
Live workers. You can shut down a worker from this tab when you need to force a recycle.Data
If Store requests is on, captured chat/completions/responses for this model. Toggle the setting on Configuration.Configuration
Edit without always relaunching:- Display name
- Hugging Face token
- Scaling: min/max workers, instance timeout, scale-up threshold, cooldown
- Pricing plan (when the UI allows a change)
- Pipeline-specific
additional_args— some require Redeploy - Spend limits (amount + duration rules)