Shift batching, creator-style: lane handoff snapshots + supervision mode for queue lanes #2140

Activity

1. developing-today commented on Oct 29, 2026

   

   developing-todayBot

   on Oct 29, 2026 – with developing.today

   Contributor

   More actions

   CODE-1038
2. dezren39

   added

   priority/highImportant and time-sensitiveImportant and time-sensitive

   on Oct 29, 2026
3. dezren39 commented on Oct 29, 2026

   

   dezren39

   on Oct 29, 2026

   MemberAuthor

   More actions

   TOP PRIORITY (Drew, 2026-10-29 01:23 CDT): Batch 3 fleet consolidation — pick this up in the next runs, ahead of routine work. Being enqueued to repo-health-code-pilot at top priority now.
4. dezren39

   added

   queue/queuedHas a queued task in the Supabase agent queueHas a queued task in the Supabase agent queue

   work/assignedQueue worker lifecycle stateQueue worker lifecycle state

   queue/claimedQueue worker lifecycle stateQueue worker lifecycle state

   status/in-progressQueue worker lifecycle stateQueue worker lifecycle state

   and removed

   queue/queuedHas a queued task in the Supabase agent queueHas a queued task in the Supabase agent queue

   on Oct 29, 2026
5. dezren39 commented on Oct 29, 2026

   

   dezren39

   on Oct 29, 2026

   MemberAuthor

   More actions

   🤖 Agent note · picked up

   • Agent: queue worker (repo-worker-code), Lane B shift
   • Session / ID: queue task 3169
   • Found via: queue claim
   • Task: Triage Shift batching, creator-style: lane handoff snapshots + supervision mode for queue lanes #2140 (Batch 3, Drew-approved): verify the shift-overhead premise, design lane handoff snapshots + supervision mode (with its binding gate), route implementation in the same pass, update epic epic: fleet expansion & placement program (2026-10-28 directive, LIST 1–5) #2055.
   • Expectation: m — done looks like a verdict + design on Shift batching, creator-style: lane handoff snapshots + supervision mode for queue lanes #2140, a follow-on packet or linked implementation issue, and an epic epic: fleet expansion & placement program (2026-10-28 directive, LIST 1–5) #2055 update.
   • Links: epic epic: fleet expansion & placement program (2026-10-28 directive, LIST 1–5) #2055 · conductor Implement the session conductor: merge openchamber-dispatcher + openchamber-babysitter into openchamber-conductor #2153
6. dezren39 commented on Oct 29, 2026

   

   dezren39

   on Oct 29, 2026

   MemberAuthor

   More actions

   🤖 Agent note · done

   • Agent: queue worker (repo-worker-code), Lane B shift
   • Session / ID: queue task 3169
   • Found via: queue claim
   • Task: Triage Shift batching, creator-style: lane handoff snapshots + supervision mode for queue lanes #2140: premise verified, snapshot + supervision design decided below, Part 1 routed to Implement lane handoff snapshots for queue shifts (Part 1 of #2140; supervision mode gated on conductor + binding) #2158 (queue task 3183), epic epic: fleet expansion & placement program (2026-10-28 directive, LIST 1–5) #2055 updated.
   • Expectation: m — verdict + design in this comment; implementation is Implement lane handoff snapshots for queue shifts (Part 1 of #2140; supervision mode gated on conductor + binding) #2158's, supervision mode is gated.
   • Links: implementation Implement lane handoff snapshots for queue shifts (Part 1 of #2140; supervision mode gated on conductor + binding) #2158 · conductor Implement the session conductor: merge openchamber-dispatcher + openchamber-babysitter into openchamber-conductor #2153 · epic epic: fleet expansion & placement program (2026-10-28 directive, LIST 1–5) #2055

   Triage verdict — premise CONFIRMED, design decided, Part 1 routed

   Premise check: verified from the inside — this triage was executed by a Lane B shift that itself re-read the full worker instruction, re-derived queue health (including one 12s health timeout + retry), and re-validated every task's GitHub state from zero, exactly the per-start overhead this issue names. The scale figures stand as filed (9-lane family = 52% of cron spend at check-in #5; 6–13M input/run; creator batching precedent 3.93M → ~1.4M per filed issue). One refinement: shifts do not start from nothing — checkpoints and progress survive in the queue itself. What is missing is a lane-level digest of everything around the tasks (infra state, in-flight PR/CI pointers, dead ends already discovered), which is what the snapshot below adds. It complements task checkpoints; it does not replace them.

   Design

   Part 1 — lane handoff snapshots (landable now)

   • Write: at shift exit (after the signal refresh, before lock release) the lane overwrites hidden_files/lane-handoff-<queue>-<lane>.json:
     {generated_at, lane, queue_health:{queued, claimed, parked, blocked}, worked:[{task, issue, outcome, pr, links}], in_flight:[{task, issue, pr, ci_state, resume_condition}], infra_notes:[{note, first_seen}], leads:[task ids worth claiming next and why]}
   • Read: the next shift of the same lane reads it right after the Step 0 gate. Leads, not truth: the snapshot replaces search (what exists, where things stand, what was already ruled out) and never verification — claims, CI state, and labels are still checked fresh before acting, and anything contradicted by a fresh check loses to the fresh check. Entries older than ~2 cadences are context only.
   • Why this shape: it is the fleet's proven digest pattern (watchdog's live-state digest, the monitor rewrite in Slim repo-fleet-monitor: snapshot/SQL-first state fetch + 2h cadence (target ≤3M/run) #2134) applied to the lane boundary; one small file per lane, overwritten, no log archaeology, no new store.
   • Success test: daily check-in reports per-lane input/run for two weeks post-landing vs the 6–13M range; success = median down with tasks/shift held or rising.

   Part 2 — supervision mode (designed, GATED — not routed for implementation)

   • Gate (both required): (a) the session conductor (Implement the session conductor: merge openchamber-dispatcher + openchamber-babysitter into openchamber-conductor #2153) is cut over, and (b) session↔task binding is recorded on queue tasks (a claimed task's progress names its OpenCode session id). Until both hold, lanes keep executing directly; a half-bound supervision mode would recreate the dispatcher/babysitter split this batch just merged.
   • Mode, once gated in: a shift batch-supervises 15–20 dispatched sessions (checkpoint reviews, stall nudges through the conductor's escalation ladder, hand-backs with reasons) and executes directly only what no session holds.
   • Cadence: the 40m → 2h evaluation happens only after the gate, and a cadence change is a rate change — it needs Drew's explicit word at that time, with before/after tokens-per-completed-task from the daily check-in as the evidence pack. Nothing in this triage pre-approves it.

   Routing

   Part 1 implementation filed as #2158 (snapshot steps in all lane shift bodies, one evidenced handoff, two-week measurement; Part 2 gate state recorded, not started) and enqueued as queue task 3183, flagged live-fleet-change. Role note: lanes stay Do in the #2131 map in both parts; supervision changes how Do works, not which role owns it.
7. dezren39

   removed

   status/in-progressQueue worker lifecycle stateQueue worker lifecycle state

   queue/claimedQueue worker lifecycle stateQueue worker lifecycle state

   work/assignedQueue worker lifecycle stateQueue worker lifecycle state

   on Oct 29, 2026
8. github-actions

   assigned

   dezren39

   on Oct 29, 2026
9. dezren39 commented on Oct 30, 2026

   

   dezren39

   on Oct 30, 2026

   MemberAuthor

   More actions

   🤖 Agent note · Part 2 gate state (recorded per #2158 acceptance, queue task 3812, 2026-10-30)

   Gate NOT satisfied on either count — no supervision mode started, no cadence change made.

   • (a) Conductor cut over? No. openchamber-dispatcher and openchamber-babysitter are both still enabled: true in the live cron roster (verified 2026-10-30); the conductor cutover for Implement the session conductor: merge openchamber-dispatcher + openchamber-babysitter into openchamber-conductor #2153 is queued as task 3806 and has not landed.
   • (b) Session↔task binding present? No. Queue task progress records stage/artifact/next only — no OpenCode session id is recorded on a claimed task (observed across this shift's claims: tasks 3311, 3557, 3415, 3559, 3647, 3696, 3733, 3797, 3808, 3812).

   Part 1 (per-lane handoff snapshots) landed today under #2158: snapshot write + read steps are in both shared lane instruction files, and the two-week input/run measurement clock starts 2026-10-30. When both gate conditions above flip, file the Part 2 activation as a fresh follow-on from this issue — the 40m → 2h cadence evaluation additionally needs Drew's explicit word even then.
10. 17 remaining items

    Load more
11. dezren39

    removed

    queue/claimedQueue worker lifecycle stateQueue worker lifecycle state

    work/assignedQueue worker lifecycle stateQueue worker lifecycle state

    on Nov 4, 2026
12. dezren39 commented on Nov 4, 2026

    

    dezren39

    on Nov 4, 2026

    MemberAuthor

    More actions

    🤖 Agent note · handed off

    • Agent: fleet monitor (repo-fleet-monitor)
    • Session / ID: task 7304
    • Found via: monitor reconciliation
    • Task: claim-evaporated repair — the queue claim on task 7304 ended without complete/park/block, its lease expired, and the task returned to queued unclaimed while this issue still carried claimed labels. Labels reset to the truth: queue/queued + status/needs-triage. Checkpoint stage at evaporation: handed_off.
    • Expectation: the next lane to claim task 7304 posts a fresh pickup note.
13. dezren39

    added

    work/assignedQueue worker lifecycle stateQueue worker lifecycle state

    queue/claimedQueue worker lifecycle stateQueue worker lifecycle state

    status/in-progressQueue worker lifecycle stateQueue worker lifecycle state

    and removed

    status/needs-triageNot yet triaged; needs a first lookNot yet triaged; needs a first look

    queue/queuedHas a queued task in the Supabase agent queueHas a queued task in the Supabase agent queue

    on Nov 4, 2026
14. dezren39 commented on Nov 4, 2026

    

    dezren39

    on Nov 4, 2026

    MemberAuthor

    More actions

    🤖 Agent note · handed off

    • Agent: queue worker (repo-worker-code), lane code-j
    • Session / ID: task 7304
    • Found via: queue claim
    • Task: Triage Shift batching, creator-style: lane handoff snapshots + supervision mode for queue lanes #2140: Shift batching, creator-style: lane handoff snapshots + supervision mode for que handed off to the opencode execution lane per task 7304 lane assignment (assigned_lane=opencode, session_id None)
    • Expectation: claimant launches the bound opencode session; that session drives task 7304 to its terminal state
15. dezren39 commented on Nov 4, 2026

    

    dezren39

    on Nov 4, 2026

    MemberAuthor

    More actions

    🤖 Agent note · handed off

    • Agent: queue worker (repo-worker-code), lane code-v
    • Session / ID: task 7304
    • Found via: queue claim
    • Task: task 7304 (triage for issue Shift batching, creator-style: lane handoff snapshots + supervision mode for queue lanes #2140) is assigned to the opencode execution lane per repo-queue lane assignment; handing off to the laptop opencode claimant. Issue Shift batching, creator-style: lane handoff snapshots + supervision mode for queue lanes #2140 revalidated fresh this shift: OPEN, no hard-stop labels (held / user/owned / do-not-merge / do-not-auto-edit). Attempts at cap (task 7304 attempts 12 of 12): if the claimant does not bind before the lease/park window, task 7304 may need its cap raised by the fleet monitor to stay claimable.
    • Expectation: claimant launches the bound opencode session and drives task 7304 to a terminal state.
    • Links: issue Shift batching, creator-style: lane handoff snapshots + supervision mode for queue lanes #2140
16. dezren39

    added

    queue/parkedQueue task is parked (waiting on CI, time, or an event)Queue task is parked (waiting on CI, time, or an event)

    status/parkedDeliberately paused with a recorded resume conditionDeliberately paused with a recorded resume condition

    and removed

    status/in-progressQueue worker lifecycle stateQueue worker lifecycle state

    queue/claimedQueue worker lifecycle stateQueue worker lifecycle state

    on Nov 4, 2026
17. dezren39 commented on Nov 4, 2026

    

    dezren39

    on Nov 4, 2026

    MemberAuthor

    More actions

    🤖 Agent note · handed off

    • Agent: queue worker (repo-worker-code), lane code-v
    • Session / ID: task 7304
    • Found via: queue claim
    • Task: task 7304 (triage for issue Shift batching, creator-style: lane handoff snapshots + supervision mode for queue lanes #2140) is assigned to the opencode execution lane per repo-queue lane assignment; handing off to the laptop opencode claimant. Issue Shift batching, creator-style: lane handoff snapshots + supervision mode for queue lanes #2140 revalidated fresh this shift: OPEN, no hard-stop labels (held / user/owned / do-not-merge / do-not-auto-edit). Second handoff this shift: task 7304 was handed off at 13:14Z (note on issue Shift batching, creator-style: lane handoff snapshots + supervision mode for queue lanes #2140) and returned unbound after the 20-minute park; the opencode claimant did not launch a bound session in that window. Attempts remain at cap (task 7304 attempts 12 of 12) — fleet monitor: please raise the cap or bind the claimant session before task 7304 becomes unclaimable.
    • Expectation: claimant launches the bound opencode session and drives task 7304 to a terminal state.
    • Links: issue Shift batching, creator-style: lane handoff snapshots + supervision mode for queue lanes #2140