MC-4631 — Repair Luci Control Room active pickup loop so dev backlog is dispatched, not only alerted

Home Board MC-4631

Repair Luci Control Room active pickup loop so dev backlog is dispatched, not only alerted

Root cause from 2026-06-03 Telegram escalation: legacy ticket-pickup/needs-input-pickup were correctly disabled after duplicate-spawn regressions, but the replacement Luci Contr...

State Done Next Action Closed Owner Luci Runtime Closed Age 10d ago

← MC-4631

Ticket is done; runtime is closed. · cwd /home/lucienne/workspace/state/control-room-worktrees/mc-4631-orchestration-repair · uptime 10d 1h · last activity 9d 23h ago

Description

MC-4631

Root cause from 2026-06-03 Telegram escalation: legacy ticket-pickup/needs-input-pickup were correctly disabled after duplicate-spawn regressions, but the replacement Luci Control Room watcher is no_agent:true and alert-only. It detects new high-priority items/stuck workers but does not actively dispatch medium/low inbox/todo backlog into runtimes. Acceptance: implement/verify event-driven controller handoff or deterministic single-ticket dispatcher that moves Luci-owned dev tickets from inbox/todo into worker runtimes with MC runtime ledger, respects duplicate-launch guards, stays silent when no work, and provides evidence via a controlled smoke ticket.

Activity

done

Luci is working...

Details — Done · High · Luci ▼

State

Done

Closed

State

Priority

People

Owner (assigned to)

Controller

Timing / Details

Due Date

Snooze

Source telegram-escalation (human)

Project Mission Control

Created 10d ago

Updated 10d ago

Advanced / Operator evidence

Routing owner

Routes via

Operator console

Evidence

Ticket is done; runtime is closed. controller_gate_closed cwd /home/lucienne/workspace/state/control-room-worktrees/mc-4631-orchestration-repair · uptime 10d 1h · last activity 9d 23h ago MC is visibility-only. Hermes Luci launches and gates work outside MC, then mirrors evidence/status here. Raw console: luci-controller · claude-code

Workflow

Start Dev Review + QA ▾

Select phases to include:

Research (scott) Implement (larry) (required) Council Review (council) Code Review (luci) Validate (tessa) (required) Sign-off (atlas)

Agents

Review Gates

Decision

WAT routing: choose an agent, review gate, or decision. Buttons use the live runtime when one is attached.

system 10d ago

RUNTIME TERMINAL STATE (MC-3482 contract) status: warning summary: Live ticket runtime timed out without a harvestable verdict; parked for automatic runtime recovery. root_cause: harvest timeout — pane was silent/wedged and the transcript carried no verdict to sweep safe_retry: Runtime was killed and the ticket was returned to todo for a fresh Luci-owned retry; no human input is needed. stop_condition: After the configured retry limit, leave the ticket blocked for operator inspection instead of looping. human_input_required: no next_actions: - Pickup will start a fresh runtime from MC history. - Inspect the pane log artifact only if the retry fails again. artifacts: - ticket:MC-4631

system 10d ago

api 10d ago

MC-4631 delivered on branch cr/mc-4631-orchestration-repair (commit 96deed3, pushed). scripts/control_room_pickup.py: deterministic no_agent loop that reconciles stranded/finished direct workers (recovers MC-4612 class), promotes actionable inbox->todo, and dispatches exactly one bounded detached worker — replacing the alert-only watcher. 5-model council applied (atomic CAS; tmux-dies-on-exit wedge fix; shlex-quoted cmd + identifier validation; tee -a header survival; reconcile grace window; con.rollback; flock; phase isolation; prompt DoS cap). 40 tests pass. Live proof: dry-run selects exactly one (MC-4632) and skips active/blocked/human-wait; reconcile-only recovered genuinely-stranded MC-4637 (in_progress->todo). Wired via deployed shim ~/.hermes/scripts/luci_ticket_watcher.py; loop activates on merge to master. Legacy pickup stays disabled. Kill-switches CR_PICKUP_DISABLE / CR_PICKUP_NO_DISPATCH.

luci-followup-sweep 10d ago

[follow-up] 27 min past expected check-in (2026-06-03T08:35:46.418039+02:00 UTC).

luci-controller 10d ago

Controller escalated this as the P0 orchestration breakage: active watcher is alert-only, legacy pickup disabled, tickets not moving. Dispatched direct Coder worker to build the active pickup/recovery loop. Runtime: tmux cr-MC-4631, pid 2722013, worktree /home/lucienne/workspace/state/control-room-worktrees/mc-4631-orchestration-repair. Expected check-in: 2026-06-03T08:35:46.418039+02:00 Completion gate: branch+tests+smoke proof; no legacy pickup re-enable.

luci-controller 10d ago

Additional root cause from MC-4637 false needs_input: routing_log shows generic/manual_api transitions, not worker output. 07:53 status todo→in_progress with no runtime/session/tmux; generic PATCH only updates DB and does not launch a worker. 08:05 status in_progress→needs_input with failure_reason 'No live worker or interactive session is attached to this ticket.' No comment/runtime existed. This is a technical no-start state misclassified as a human gate. Fix must reject or auto-normalize in_progress/needs_input without runtime evidence and route eligible Luci tickets through active dispatcher instead.

luci-controller 10d ago

Controller landed MC-4631 active orchestration loop. Evidence: - Rebased branch cr/mc-4631-orchestration-repair onto current master after MC-4612. - Tests: `python3 -m pytest tests/test_control_room_pickup.py tests/test_mc4612_home_rail_close_reason.py tests/test_mc4123_home_v2.py -q` → 79 passed. - Fast-forward merged commit 0a4f12a to master and pushed origin/master. - Dry-run smoke: promotes eligible inbox tickets and selects exactly one dispatch target; skips active/blocked/human-wait; no mutations in dry-run. - luci-dashboard.service restarted; local + Tailscale API smoke returned HTTP 200. Note: luci-persistent remains only a dormant interactive console/keepalive, not the controller path. The active loop is the no_agent watcher + control_room_pickup.py.

Live ▼

No activity yet

←