MC-4494 — Implement Control Room Operating Model v2

Home Board MC-4494

Implement Control Room Operating Model v2 — 7-role rollout

Implement the Control Room Operating Model v2 using the 7-role team process first, then use that process to complete the rest of the rollout. Controller: Luci in Telegram topic...

State Done Next Action Waiting: done Owner Luci Runtime Closed Age 13d ago

← MC-4494

Ticket is done; runtime is closed. · profile claude_opus_1m_high · cwd /home/lucienne/workspace/mission-control · uptime 12d 19h · last activity 12d 14h ago

Description

MC-4494

Implement the Control Room Operating Model v2 using the 7-role team process first, then use that process to complete the rest of the rollout. Controller: Luci in Telegram topic with Elmar. Durable ledger: Mission Control. Operating model: Controller decides; MC records; workers/subagents/runtimes execute; evidence returns to MC before Done. 7-role team, treated as responsibilities/gates rather than mandatory separate workers: 1. Controller — Luci; owns routing, technical decisions, gates, and user-facing synthesis. 2. Planner — turns this rollout into concrete phases, acceptance criteria, dependencies, and implementation order. 3. Designer — board/UX/product-shape gate for Inbox/Todo/Done, labels, comments, Workbench/board flows. 4. Coder — implementation only; no self-approval. 5. Reviewer — independent provider/model/tooling where possible; verifies code/design/test evidence. 6. Tester — browser/user journey validation, especially board and comment flows. 7. Validator — final evidence gate before marking rollout done. Key requirements: - Runtime/model picker is advisory guidance only. Controller may choose any suitable tool, runtime, subagent, subprocess, direct script/API, browser harness, remote host, or future adapter. - Elmar-facing board remains simple: Inbox / Todo / Done. - Internal active phases live as labels/metadata/comments, not extra public columns. - Comments notify the controller; they must not automatically spawn stale/random runtimes. - Audit comments must include role, runtime/capability, scope, outcome, evidence, and next step. - Hermes cron remains deterministic/no_agent for lightweight alerts; MC scheduled tasks remain for complex artifact-producing recurring work. - Use event-driven hooks where practical; cron polling is fallback. - Expected check-in: 2026-05-31 16:33:05 UTC. Acceptance criteria: - 7-role process is represented clearly in MC and docs/UI. - Implementation plan exists and is linked from this ticket. - Board simplification, audit schema, controller hook behavior, capability registry/guidance, and scheduler boundaries are implemented or split into runnable child tickets. - Live MC UI/API behavior is smoke-tested from Tailscale-reachable routes. - Final Done requires reviewer + tester + validator evidence comments.

Activity

done

Luci is working...

Details — Done · High · Luci ▼

State

Done

Waiting: done

State

Priority

People

Owner (assigned to)

Controller

Timing / Details

Due Date

Snooze

Source telegram (human)

Project Mission Control

Created 13d ago

Updated 12d ago

Advanced / Operator evidence

Routing owner

Routes via

Operator console

Evidence

Ticket is done; runtime is closed. websocket_idle_not_worker profile claude_opus_1m_high · cwd /home/lucienne/workspace/mission-control · uptime 12d 19h · last activity 12d 14h ago MC is visibility-only. Hermes Luci launches and gates work outside MC, then mirrors evidence/status here. 1745a8db-b21a-4e43-92d5-e9d96fb8dbe8 · luci · claude/anthropic/opus[1m]

Workflow

MC-4495 · luci done MC-4496 · luci done MC-4497 · luci done MC-4499 · luci done MC-4500 · luci done

Agents

Review Gates

Decision

WAT routing is active. Agent/review buttons inject the next step into the live runtime or dispatch a fresh worker.

luci 13d ago

Controller bootstrap: accepted Elmar's direction to create the 7-role team/process first, then use it to implement the rollout. Role scaffold: - Controller: Luci in Telegram topic; owns routing, capability choice, and final synthesis. - Planner: first active gate; produce implementation plan. - Designer: board/UX gate before UI implementation and screenshot signoff after. - Coder: implementation only. - Reviewer: independent review, preferably different provider/model/tooling. - Tester: browser/user journey gate. - Validator: final evidence gate before Done. Runtime guidance is advisory only. Controller may choose any suitable runtime, tool, subagent, subprocess, direct script/API, browser harness, remote host, or future adapter, and must record evidence/outcome back into MC. Active child: MC-4495 Planner gate. Expected check-in: 2026-05-31 16:33:05 UTC.

Controller decision

planner 13d ago

Planner handoff complete for Control Room v2 rollout. Role: Planner Runtime/capability: Hermes delegate_task planner subagent Scope: Produce implementation plan only; no code/file/ticket mutations by planner. Outcome: completed Plan summary: 1. Phase 0 — Baseline/council preflight for workflow/runtime/comment dispatch behavior. 2. Phase 1 — Add inert 7-role registry and audit comment schema. No behavior change. 3. Phase 2 — Add active phase labels/metadata without status fragmentation or reassignment. 4. Phase 3 — Replace/augment WAT with Control Room role/gate surface. 5. Phase 4 — Make Controller dispatch explicit; role/comment actions notify Controller but do not auto-spawn stale/random runtimes. 6. Phase 5 — Notification model: event-driven comments/status changes to Controller; no_agent watchers as fallback. 7. Phase 6 — Scheduled task separation: Hermes cron for deterministic alerts, MC tasks for complex artifacts. 8. Phase 7 — Board/docs/runbook alignment and UI validation. Likely implementation files: - mission-control/models.py - mission-control/app.py - mission-control/mc_role_guidance.py - mission-control/ticket_runtime.py - mission-control/runtime_registry.py - mission-control/templates/ticket_runtime_workbench.html - mission-control/static/ticket_runtime_workbench.js - board/home templates and JS as needed - tests/test_workflow_pipeline.py - tests/test_workflow_idempotency.py - tests/test_ticket_runtime*.py - tests/test_telegram_bridge.py - tests/test_runtime_profiles.py - new tests for Control Room roles/audit schema/comment-only role actions. First safe implementation slice: - Add canonical 7-role registry and audit comment formatter. - Add tests for role keys, optional gates, old-role capability mappings, and audit schema formatting. - Add read-only helper/API only if useful. - Do not change dispatch, pickup, board, Telegram, or runtime behavior in this slice. Acceptance highlights: - Roles are gates/responsibilities, not mandatory separate workers. - Ticket owner remains stable; active phase is label/metadata/comment trail. - Elmar-facing board remains Inbox / Todo / Done. - Runtime/model guidance remains advisory only; Controller may choose any suitable capability. - Durable MC records include selected capability, evidence, and outcome. Next step: Coder slice opened for the inert foundation. Reviewer/Validator gates required before behavior-changing phases.

controller 13d ago

Planner gate MC-4495 completed. Active child is now MC-4496: inert 7-role registry and audit schema.

controller 13d ago

MC-4496 foundation slice landed on origin/master (c1451168b4348bba0fd39de324e00f39616f37e9). Active child is now MC-4497: baseline/council preflight before behavior-changing Control Room v2 work.

controller 13d ago

MC-4497 preflight completed. Active child is now MC-4499: feature-flag WAT dispatch actions so they notify Controller instead of auto-dispatching.

Controller decision

controller 13d ago

MC-4499 landed on origin/master (a0966da113f137c56508c10e3f4e5236a1c7338f). Active child is now MC-4500: feature-flag human ticket comments to notify Controller instead of auto-routing.

Controller decision

controller 13d ago

Runtime directive: remaining Control Room v2 rollout should prefer claude_opus_1m_high / Opus 1M high effort; Codex direct CLI only with tested ChatGPT-OAuth-compatible model (gpt-5.5 currently smoke-tested OK), not codex-mini.

controller 13d ago

Runtime policy update from Elmar for future 7-role Control Room work: - Real coding work should use the strong pair: Codex `gpt-5.5` with reasoning effort `xhigh` and Claude `opus[1m]` / Opus high effort. - Use them as implementation/review peers where practical: one writes, the other reviews, then Controller verifies. - Controller discretion remains for exceptions (budget, availability, tiny Tier-1 work, deterministic scripts), but this is the preferred default for important coding slices. - Do not use Codex `codex-mini` on this ChatGPT OAuth setup; `gpt-5.5` is the smoke-tested Codex model here.

Controller decision

lucienne 13d ago

MC-4500 coder slice landed on origin/master (7a102cd). Atomic Controller-audit transaction + dedupe hardening. 46 tests passed. Now in_review for reviewer gate. This is the second behavior-changing Control Room v2 slice (comment route). Next slice after MC-4500 gates: continue with remaining Control Room v2 rollout items.

controller 13d ago

Runtime budget policy update from Elmar: - Ignore current Claude usage tracker throttle/limits for this Control Room rollout. - Claude limits only come into effect on 2026-06-15, and only for `claude -p` usage. - Until then, continue using the preferred real-coding pair: Codex gpt-5.5 xhigh + Claude Opus/Opus 1M high, subject to actual provider/runtime availability.

validator 13d ago

MC-4500 landed and closed at 594c30ceae57ba7def9413e0035fe380d7a1ff67. Follow-up MC-4501 opened for omitted author_type asymmetry; not release-blocking. Ready for next Control Room v2 slice.

controller 13d ago

Next rollout slice clarified: MC-4468 promoted to high-priority Control Room v2 board cleanup to remove legacy UX/buttons and align the board with Inbox/Todo/Done mental model.

luci-followup-sweep 13d ago

[follow-up] 28 min past expected check-in (2026-05-31 16:33:05 UTC).

validator 13d ago

MC-4468 board cleanup slice landed and deployed at f1abadd0bc441c088bf2b5523707d95eccac1d00. Next rollout candidates: MC-4501 author_type cleanup or event-driven hook/runtime registry wiring.

controller 13d ago

Ledger hygiene: closed stale websocket idle runtime row. MC-4494 is an umbrella/waiting ticket and currently has no active tmux worker; no missing tmux action required.

controller 13d ago

Ledger hygiene: closed idle websocket runtime row with no tmux_target/pid. This was not an active worker and should not be surfaced as missing tmux.

controller 13d ago

Ledger hygiene: closed 1 remaining idle websocket runtime row(s); no active worker expected on umbrella ticket.

controller 13d ago

Controller night objective accepted: get Control Room v2 to full validated implementation. Remaining gates: 1. MC-4502 — harden CLI worker launch/liveness ledger (active next). 2. MC-4501 — normalize omitted author_type behavior. 3. Event-driven Controller hook path. 4. Runtime/role registry wiring into routing/audit surfaces. 5. Final activation/docs/live validation and closeout. Gate rules: TDD for behavior changes, independent reviewer on every coding slice, browser/mobile validation for UI, no Done without landed commit + tests + live smoke + audit comment.

Controller decision

validator 13d ago

MC-4501 author_type normalization landed at e7d4d4f. MC-4502 runtime/worker hardening remains active.

validator 12d ago

MC-4502 runtime/worker hardening landed at 1bd1c48. Next active slice: MC-4503 event-driven Controller hook path.

Controller decision

validator 12d ago

MC-4503 event-driven Controller hook path landed at 51d5574. Next active slice: MC-4504 runtime/role registry wiring.

Controller decision

validator 12d ago

MC-4504 runtime/role registry advisory wiring landed at 38c33ec. Entering final activation/validation slice.

validator 12d ago

Validator gate complete — final activation/live validation passed. Production stance: - `MC_CONTROL_ROOM_V2_CONTROLLER_NOTIFY` is implemented and remains default-off unless explicitly enabled. - `MC_CONTROL_ROOM_V2_CONTROLLER_HOOKS` is implemented and remains default-off unless explicitly enabled. - Live service environment currently has neither flag set; no noisy polling or auto-dispatch is enabled. Validation evidence: - Focused rollout suite: 297 passed + 37 subtests. `python3 -m pytest tests/test_control_room_roles.py tests/test_control_room_role_registry_wiring.py tests/test_control_room_event_hooks.py tests/test_mc4499_controller_notify.py tests/test_mc4500_comment_controller_notify.py tests/test_mc4501_comment_author_type_defaults.py tests/test_mc4502_runtime_hardening.py tests/test_ticket_runtime.py tests/test_terminal_state_reconciliation_regressions.py tests/test_runtime_profiles.py tests/test_mc_pickup_larry_runtime.py tests/test_mc4318_pending_view.py -q` - Compile: `python3 -m py_compile app.py models.py ticket_runtime.py control_room_events.py control_room_roles.py /home/lucienne/workspace/mc_pickup.py` → passed. - Service deploy: `sudo -n systemctl restart luci-dashboard.service` → active. - Live local smoke: `/api/v1/control-room/roles` → 200, advisory_only=true, roles=7. - Live local audit-format smoke → valid request 200 with `[AUDIT]`; invalid non-string role 400 `role must be a string`. - Live Tailscale smoke: `http://100.118.207.3:3001/api/v1/control-room/roles` → 200; explainer static page → 200. - Runtime ledger check: no active runtime_sessions rows for MC-4494/4495/4496/4497/4499/4500/4501/4502/4503/4504/4505. - Git: master clean and pushed to origin/master at 38c33ec. Delivered slices: - MC-4496 inert seven-role registry + audit schema. - MC-4499 WAT dispatch notify-Controller flag path. - MC-4500 human ticket-comment notify-Controller flag path with atomic audit. - MC-4501 author_type default normalization. - MC-4502 worker launch/liveness ledger hardening and Codex model smoke checks. - MC-4503 default-off event-driven Controller hook path. - MC-4504 advisory role/runtime registry wiring into prompts/API/audit-format surfaces. - Docs + interactive explainer: http://100.118.207.3:3001/static/reports/control-room-interactive-explainer.html

Controller decision

Live ▼

No activity yet

←