YOU

Metrics

Watch live you-agent-factory factory-ops metrics for a Factory Session: engine activity, work-token lifecycle, optional resource usage, and the local operator dashboard.

Read Session Status

Factory metrics are scoped to one live Factory Session. Call GET /factory-sessions/{session_id}/status for a deeper runtime health read than session list or factory query. Replace {session_id} with a live session id from session list—often ~default on single-session hosts. The response carries the factory-ops metrics fields operators watch during a live run: engine activity and runtime status, work-token lifecycle buckets, and optional resource usage when pools are declared.

Example status read against the default local host:

curl -s "http://localhost:7437/factory-sessions/~default/status"

Read these fields in the status response:

FieldMeaning
factoryStateFactory lifecycle phase — for example IDLE, RUNNING, PAUSED, COMPLETED, or FAILED.
runtimeStatusWhether the engine is actively processing — IDLE, ACTIVE, or FINISHED.
categoriesToken counts by lifecycle bucket: initial, processing, terminal, and failed.
totalTokensTotal tokens across categories.
resourcesOptional resource usage entries when the factory declares resources.
A factory can be RUNNING while runtimeStatus is IDLE when the host is up but no work is in flight. Read factoryState, runtimeStatus, and categories together when deciding whether to submit more work or wait for completion.

Open The Operator Dashboard

When the service was started via you or you run without quiet mode, open the local operator dashboard as the browser surface for the same live metrics. The dashboard shows live session selection, work position, and factory activity alongside CLI and API status reads.

Default local dashboard URL (service started without quiet mode):

http://localhost:7437/dashboard/ui
Use the same host and port as the API unless the process that bound the listener used a different server or port setting.

Watch Metrics Over Time

A one-shot status read is a snapshot. Time-ordered watching means reading the same factory measures across successive ticks so trends become visible. The teaching chart below plots work-token lifecycle buckets — processing, terminal, and failed — over successive status ticks, the same categories totals operators watch while a Factory Session runs.

Work-token lifecycle buckets over successive status ticks

X axis: Status tick. Y axis: Token count.

Across successive status ticks, processing drains while terminal grows and a small failed bucket accumulates — the same categories totals an operator watches when reading session status over time.

Limits And Assumptions

Metrics covers factory metrics exposure for you-agent-factory operators watching a live Factory Session. It is not a sync of packaged CLI docs, not a logs or diagnostic-event reference, not OpenAPI or API reference ownership, not SPC control-chart depth, not bottleneck diagnosis, and not Model Atlas or benchmark metrics.

Tags

On this page

Read Session StatusOpen The Operator DashboardWatch Metrics Over TimeLimits And AssumptionsTags