System & Task Queue API

Backs the Monitor page and operational tooling.

System — prefix /api/v1/system

Method	Path	Description
GET	/health	Backend health check
POST	/restart	Restart the backend process
GET	/gpu-list	All discovered GPUs with static info
GET	/gpu-metrics	Metrics for one GPU (?gpu_id=)
GET	/gpu-metrics/all	Metrics for every GPU
GET	/gpu-info	Detailed GPU device info
GET	/gpu-processes	Processes currently holding GPU memory
GET	/metrics	Host CPU/memory snapshot
GET	/disk-usage	Disk usage for the data volumes
GET	/disk-rates / /network-rates	I/O rates
GET	/all	Everything above in one call (what the Monitor page fetches on load)
GET	/resource-estimate	Estimate resources for a prospective job config

Live metrics stream over the system/* WebSocket channels every 2 seconds — the REST endpoints are the polling fallback and scripting interface. See the WebSocket reference.

Task queue — prefix /api/v1/task-queue

The persistent record of background jobs, powering the Monitor page's Active Operations and Failed Operations sections.

Method	Path	Description
GET	``	List task records (paginated)
GET	/active	Currently running/queued tasks
GET	/failed	Failed tasks with error details
GET	/{id}	One task record
POST	/{id}/retry	Retry a failed task
DELETE	/{id}	Delete a task record (204)

Workers — prefix /api/v1/workers

Celery introspection:

Method	Path	Description
GET	/queues	Queue depths
GET	/active	Tasks currently executing on workers
GET	/stats	Worker statistics
GET	/health	Worker connectivity health

Version — GET /api/v1/version

Returns the running application version.