Command Center
Command Center is the desktop operator surface for Farmslot.
It exists because supervising many agents from one serial chat or a pile of terminal windows does not scale. The operator needs a persistent cockpit for fleet state, run state, evidence, terminals, decisions, and publication gates.
Core jobs
Fleet overview
Slot health, readiness, machine state, and dispatchability.
Slot health, readiness, machine state, and dispatchability.
Run supervision
Run families, lanes, active work, progress, and status transitions.
Run families, lanes, active work, progress, and status transitions.
Terminal streaming
Inspectable worker panes and live output when intervention is needed.
Inspectable worker panes and live output when intervention is needed.
Evidence review
Artifacts, screenshots, traces, diffs, validation outputs, and recipe proof.
Artifacts, screenshots, traces, diffs, validation outputs, and recipe proof.
Decision gates
Human approvals before publication, recovery, or high-impact next steps.
Human approvals before publication, recovery, or high-impact next steps.
Eval/replay
Reference/candidate comparisons for prompts, templates, runners, and harness changes.
Reference/candidate comparisons for prompts, templates, runners, and harness changes.
Relationship to the gateway
Command Center is a client of the gateway. It should not invent a second state model.
Operator workflow
- See which slots and workers are available.
- Dispatch or inspect work from a structured surface.
- Watch live terminals only when needed.
- Review evidence, diffs, recipe traces, and before/after artifacts.
- Approve, reject, recover, or ask for another iteration.
Demo evidence
The illustration above is a docs-safe product mockup. Public demos should use a clean synthetic dataset:
- fleet overview with non-sensitive slot names;
- run detail with a synthetic artifact package;
- evidence workspace with generic before/after media;
- eval comparison with synthetic reference/candidate packages.