Babysitter live: interval rounds over all OpenCode sessions with verdicts, quota, and pool attribution

Work (Drew, 2026-10-27)

Stand up the babysitter as a live scheduled job (openchamber- babysitter, every 20m), implementing #1185's design with #1192's
detector:

Every round: enumerate ALL sessions at the laptop OpenChamber
location (and any future locations), bound or unbound, and give
each a verdict: progressing / stalled / done-unreported / blocked
/ walled (402-class) / idle.
Stall rule (from Pilot: session stall taxonomy — observable signatures of stalled vs healthy long sessions #1192's census): suspect when ≥20 min since the
session's last activity AND no live CI on its bound PR; confirm on
a second suspect round or on expiry-equivalent silence. Nudge at
most twice (message injection where the API shape is verified —
verify by the session's next assistant message), then escalate to
the fleet as a stuck-holder case.
Report per round only what changed: new stalls, nudges sent,
harvests ready (done-unreported sessions whose PR merged or work
completed → the owning queue task gets completed by its lane, the
babysitter flags it), quota snapshot per pool in use, and which
pool each active session is drawing on + whether that choice looks
effective (progress per the session's own checkpoints).
Data out: append round observations to the babysitter log in the
goal workspace (the pool-intelligence and meta-reviewer jobs read
it); unbound sessions are observed, never driven.

Acceptance criteria

The job exists and has completed three rounds with per-session
verdicts recorded.
One real (or synthetic) stall is detected inside two rounds
of the rule's threshold.
Every report names each active session's pool; a session on a
walled pool is flagged the round it happens.

Extends #1185; depends on openchamber#524 for the clean last-activity
field (rounds poll what's available until then). Child of the pool &
routing intelligence epic. Filed by Muse (operator) 2026-10-27.

CODE-442

agent-note:v1
state: reviewed (coverage loop-back)
Coverage pass 2026-11-03, loop-back review: this issue was re-reviewed (drift / overlap / already-handled / stale / blockers / ordering) so every open issue in the four repos is reviewed and in the system. Verdict: still live; in the system as task 7230 (triage, queued, priority 50) in repo-health-code-pilot. Labels audited and synced. If it is already handled at current main, its triage closes it with evidence; if oversized, the triage files the split; blockers belong on this issue.
Found via: coverage loop-back 2026-11-03

🤖 Agent note · picked up

Agent: queue worker (repo-worker-code), lane code-d
Session / ID: task 7230
Found via: queue claim
Task: Fresh triage of issue Babysitter live: interval rounds over all OpenCode sessions with verdicts, quota, and pool attribution #1338 at current state: does the openchamber-babysitter job exist and run, are its acceptance criteria met, and what blockers remain?
Expectation: effort/m triage; done = findings posted on issue Babysitter live: interval rounds over all OpenCode sessions with verdicts, quota, and pool attribution #1338 with a fix path enqueued, the work landed, or the issue closed with evidence.
Links: issue Babysitter live: interval rounds over all OpenCode sessions with verdicts, quota, and pool attribution #1338; design issue OpenChamber babysitter v1: one worker, session rounds, quota + progress report (report-only) #1185; detector issue Pilot: session stall taxonomy — observable signatures of stalled vs healthy long sessions #1192

🤖 Agent note · escalated

Agent: queue worker (repo-worker-code), lane code-d
Session / ID: task 7230 (triage of issue Babysitter live: interval rounds over all OpenCode sessions with verdicts, quota, and pool attribution #1338 at current state, 2026-11-03)
Found via: queue claim

Triage findings — the babysitter was built, ran, and worked; it is currently switched off

What exists (verified fresh this pass):

The job openchamber-babysitter exists (interval 20m, goal-owned) and ran from 2026-10-28 through 2026-10-30: 245 round sections in hidden_files/babysitter-log.md, per-session verdicts (progressing / stalled / done-unreported / blocked / walled) every round.
Its dependencies are all closed: issue OpenChamber babysitter v1: one worker, session rounds, quota + progress report (report-only) #1185 (design) CLOSED, issue Pilot: session stall taxonomy — observable signatures of stalled vs healthy long sessions #1192 (stall taxonomy) CLOSED, openchamber#524 (session-activity last-activity field) CLOSED.
Acceptance 1 (three rounds with verdicts): met many times over.
Acceptance 2 (a stall detected inside two rounds): met — e.g. 2026-10-30: suspects raised at 15:17 confirmed at 15:37; task 3943 went suspect → confirmed across consecutive rounds; confirmed stalls were escalated as stuck-holder cases per the rule.
Acceptance 3 (pool named per active session; walled pool flagged the round it happens): met in the rounds — every round carried a per-pool quota snapshot and named the pool active sessions were drawing on; walled pools (opencode-go monthly 100%, cline-pass monthly 100%, openrouter 100%, antigravity weekly 100%) were flagged round after round.
Nudges: none were ever sent — every confirmed stall had no live bound session to inject into, so the rule's escalation path was used instead. The message-injection half of the design remains unexercised, not disproven.

The blocker: the job is disabled. Last successful run Fri 2026-10-30 17:12 CDT; the saved definition also reads enabled: false. Nothing I found this pass records who disabled it or why (the Oct 30 watchdog entries still show it healthy earlier that day), so I am not treating the disable as a considered retirement of the function — and I am not silently reversing it either, because re-enabling changes live fleet behavior.

Decision for Drew — options:

Re-enable as-is (20m). Restores the only per-session stall detection + harvest flagging layer the fleet has had; its rounds repeatedly caught real evaporated claims the queue lanes could not see. Cost: a 20-minute agent round, which was also a notable token consumer.
Re-enable at a slower cadence (e.g. hourly). Keeps stall/harvest detection with ~1/3 the rounds; stall confirmation takes proportionally longer (two-round rule → up to ~2h).
Retire it and close issue Babysitter live: interval rounds over all OpenCode sessions with verdicts, quota, and pool attribution #1338 as superseded. Only sound if another job now does per-session verdicts — the fleet monitor and watchdog reconcile queue/GitHub state and job health, but neither enumerates OpenChamber sessions or attributes pools per session, so retiring loses that function outright.

Recommendation: option 1 or 2 — the acceptance criteria were met by the running system, and the stalls it caught (evaporated claims at max attempts with open PRs) are exactly the failure class the queue fleet still produces. If token cost was the reason for the disable, option 2 keeps the detection at a third of the price.

Until that decision lands, issue #1338 stays open as user/escalated: the work is done and proven, the switch is Drew's.

🤖 Agent note · escalated (supplement to the triage above)

Agent: queue worker (repo-worker-code), lane code-d
Session / ID: task 7230

Correction and completion of the picture, from the fleet backlog (lane E triage record, 2026-10-31) plus a fresh check just now:

The babysitter did not simply get switched off unexplained: on 2026-10-30 it was merged into the openchamber-conductor job (issue Implement the session conductor: merge openchamber-dispatcher + openchamber-babysitter into openchamber-conductor #2153), which is presumably why openchamber-babysitter was disabled.
However, the successor is also off: openchamber-conductor is disabled too — last successful run Mon 2026-11-02 12:15 CST. So whichever of the two jobs is meant to carry the per-session verdict rounds, neither is running today (2026-11-03).

The decision for Drew is unchanged in shape but narrower: re-enable the conductor (the current home of the babysitter function, per the #2153 merge) — or retire both jobs and close this issue as superseded. My recommendation stands with that amendment: re-enable the conductor, at 20m or hourly; the stall/harvest detection it carries has no other live replacement.

Navigation Menu

Repository navigation

Babysitter live: interval rounds over all OpenCode sessions with verdicts, quota, and pool attribution #1338

Work (Drew, 2026-10-27)

Acceptance criteria

26 remaining items

Triage findings — the babysitter was built, ran, and worked; it is currently switched off

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Babysitter live: interval rounds over all OpenCode sessions with verdicts, quota, and pool attribution #1338

Description

Work (Drew, 2026-10-27)

Acceptance criteria

Activity

developing-today commented on Oct 27, 2026

26 remaining items

developing-today commented on Nov 3, 2026

dezren39 commented on Nov 3, 2026

dezren39 commented on Nov 3, 2026

Triage findings — the babysitter was built, ran, and worked; it is currently switched off

dezren39 commented on Nov 3, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions