{{define "content"}} {{$m := .Data.Mind}}

{{$m.DisplayName}} cognitive replay

Rebuild the mind from a snapshot, feed it the same perception events again, and compare what it produces with what was recorded. Steps are exact when the deterministic code produced the same events and any model calls were served from recordings, approximate when a live model call was needed, and diverged when a deterministic step differed. Replays run in a disabled sandbox mind and can never act.

{{template "mind_tabs" (dict "M" $m "Active" "replays")}}
{{if .Data.Replays}} {{range .Data.Replays}}{{end}}
StartedStatusEventsStepsExactApprox.DivergedLive model calls
{{ago .CreatedAt}} · {{.CreatedBy}}{{.Status}}{{with .Error}} {{.}}{{end}}{{.FromSeq}}–{{.ToSeq}}{{.Summary.Steps}}{{.Summary.Exact}}{{.Summary.Approximate}}{{.Summary.Diverged}}{{.Summary.LLMLive}}
{{else}}

No replays yet.

{{end}}

Replay

{{if not .Data.Snapshots}}

Take a snapshot first; replay starts from one.

{{else if .Principal.User.Role.AtLeast "operator"}}
{{end}}
{{end}}