Skip to content

Activity

Activity is the user-facing history of work Stallari has run. It shows what started, what completed, what failed, what tools were used, which model route was selected, and what follow-up state was recorded.

Use Activity when you need to answer practical questions:

  • Did the requested work run?
  • Which provider and model handled it?
  • Did it call a plugin or only use local context?
  • Did it create a note, task, notification, or review item?
  • Why did it fail?

The live view focuses on work currently in progress. It shows active jobs, elapsed time, status, and the most recent completed items.

Operations remains the primary place for present-tense work. Activity adds durable detail once a run exists.

Timeline groups work over time so you can compare duration and overlap. It is useful for spotting long-running jobs, repeated failures, and periods where multiple jobs competed for the same local resources.

Common groupings include skill, host, provider, and status.

Analytics summarizes spend, token use, reliability, and duration over a selected time window. Use it to identify:

  • expensive skills,
  • failing plugin paths,
  • model routes that need adjustment,
  • and jobs that should be split or simplified.

Cost values depend on provider pricing configuration. Missing pricing leaves cost fields empty rather than guessing.

Browse is the detail library. It provides search, filters, grouped lists, and a detail pane.

Detail fields may include:

FieldMeaning
Run IDStable identifier for the recorded execution.
Skill or graphWhat Stallari attempted to run.
TriggerManual, schedule, file watch, API, or plugin-triggered start.
Provider and modelThe selected model route.
HostThe Mac or peer that executed the work.
MetricsDuration, cost, tool count, and token counts when available.
Tool callsPlugin calls and summarized results.
ErrorsFailure reason and recovery hints.
Review outputAny human review item created from the run.

Sensitive tool output is reduced before it is stored. Stallari preserves enough detail to debug the run while avoiding long-lived copies of private payloads where possible.

Check Status first. The helper may be stopped, a provider may be missing, or the relevant plugin may need credentials.

Remote activity depends on peer configuration and the local network path. Check Status for peer health and plugin readiness before changing Settings.

Cost reporting requires a provider with known pricing. If the provider is local or custom, Stallari may show tokens and duration without a money value.

  • Agency Model — Activity is the durable trail of typed primitives, tool calls, and authorised side effects
  • Legibility And Continuity — why a recorded run is inspectable through first-party tooling rather than vendor opacity