Archie runs its own agent loop only. Removes Kit resolution and launch,
the harness runner and adapters, the setup terminal and sandbox view,
harness secrets and OAuth capture (table dropped in migration 0075), the
task-tools contract that only Kit containers reached, and the
sandbox-kit-spec dependency. Agent profiles carry image and tools only.
The egress proxy keeps the model-session path; the sandbox relay code
stays for #1867.
Model usage was recorded per call and never read, so an org could not see
what it spent. Add the read side:
- model_usage carries identity_id. Task calls take it from the task row,
chat and curator calls from the calling principal carried in the usage
record, so an outbox replay keeps its attribution. Existing task rows are
backfilled from tasks.
- StatsService.UsageTotals sums the caller org's rows by UTC day, source,
provider, model and identity. The org comes from the principal, with no
org parameter, so cross-org reads do not exist.
- GET /api/usage prices tokens at the catalog's list rate, netting cache
hits, and reports calls to unpriced models as tokens only. The published
catalog view now carries each model's cost.
- A new org Usage page behind a read_usage grant: admin and owner only,
with a 7, 30 or 90 day range, totals, the daily series, and model, source
and identity breakdowns. The org shell keeps it on the caller's own org.
The loop runner hooks every model call and tool call and reports ordered,
redacted, clipped entries. Each is persisted as an agent_transcript task
event through InsertEvent, left out of the timeline read, and served per
attempt at /api/tasks/{id}/transcript. The step run view shows it with Trace.
Add [access].denial_retention (default 90 days, matching the audit-retention
seed) wired through config defaults and validation; the State Store sweeps
denial records older than the configured age hourly via the new
access.DenialPruner / PruneDenials store query. The dashboard gains
GET /api/access/denials (read_audit-gated, so org admins/owners see their
org's records; the store scopes by the credential's org) and an
/org/denials page backed by a Pinia store.
Schedules leave the schedules control-plane resource for State Store tables
behind a ScheduleService contract. A run is claimed per schedule slot, missed
slots fire once, the next run stays on the schedule's grid, cron is read in the
schedule-settings time zone, and failed runs are recorded and advance. Three
failures in a row raise a schedule.failing alert. A fired run records a
schedule-fired capture. Run history is pruned by age and count.
Removes the chat and spawn-task runners, the dangerous-payload guard and the
old schedules settings page.
An approval, ask or park can be resolved by an operator, a rule, a
classifier or automatically, so the vocabulary stops naming a person:
- step type human.approve -> workflow.approve
- park class needs_human -> needs_decision (ParkNeedsDecision)
- event kinds human_* -> operator_*
- alert kind task.needs_human -> task.needs_decision
- status grouping "needs you" -> "needs attention"
0061 rewrites the stored values, including the approval step type in the
workflow-definitions resource and each task's pinned definition, with the
digest recomputed so resumed tasks still match.
The prd step returns the PRD as structured fields and human.approve takes
them as its document. The plan text every surface reads and the approval
card's layout are both rendered from that one document, so what the human
approves is what the implement workflow is handed. Replaces the separately
generated OpenUI view, which could drift from the plan.
agent.run gains views: the agent may return an OpenUI Lang view of its
answer through an optional view tool, read as steps.<id>.view. human.approve
stores it as the task's plan_view, and the approval card renders it with
the dashboard's own components, falling back to the plan text when there
is no view or it does not parse. The feasibility prd step uses it.
archied, archie-ui and the tool gateway re-read the stored policies on a 30s
ticker, so a forbid written seconds earlier did not stop a run. One
instance-wide revision now moves inside the transaction of every policy write,
and a decision reads it first and rebuilds the engine only when it moved.
A version read that fails, or a policy list that fails after a bump, refuses
with ErrChainUnavailable instead of standing on the last chain, so three call
sites now treat unavailable as availability rather than denial: dispatch leaves
the capture for the next cycle, a task start parks transient, and the tool
gateway reports the tool as unavailable.