Sub-mission: make public-contract onboarding the default mode of change in Nuguland. Governing contract NL-SPEC-A7K4QX2M; rollout NL-PLAN-C4V7TQ2X.
Priorities
Locality: one registration site, feature-local slices.
Proof rigor: zero warnings, isolated nested evidence, reviewed pixels.
Owner judgement on recorded metrics over velocity.
Minimal churn per slice; no big-bang rewrite.
Decision defaults
Concern
Default
Locality vs rigor
Rigor wins
New features
Registry contract from day one
Existing features
Migrate slice-wise, one named fixture at a time
Activation
Startup composition only
Internal state API
Typed properties internally; JSON at IPC/diagnostics edges (completed 2026-11-14)
Lock/auth
Specialized controllers, never the popup lifecycle
Gates
Never weakened to make an iteration green
Non-goals
Plugin ecosystem; user configuration; runtime activation; repository-wide rewrite.
Exit criteria
One registry entry plus feature-local files is the entire wiring for an ordinary feature (uptime re-measure, 2026-11-14: Registry.qml plus its IPC handler in shell.qml).
OSD onboarded by a second agent from public contracts alone (osd-onboarding, with its recorded gaps).
Owner continuation judgement recorded against baseline metrics (2026-11-13: continue, adjusting first).
All gates green throughout: lint, JS suite, ui-test (every trunk merge ran just lint green; every merge changing behaviour also ran the full just ui-test green).
Exit evidence is recorded as met; this is not a phase transition.
Only the human changes currentPhase or chooses a successor.
Theme implementation and isolated round-trip proof do not close its separate live human visual acceptance.
---
id: NL-PHASE-NHO6XZIJ
type: phase
title: Architecture revamp
label: arch-revamp
---
# Architecture revamp
Sub-mission: make public-contract onboarding the default mode of change in Nuguland. Governing contract `NL-SPEC-A7K4QX2M`; rollout `NL-PLAN-C4V7TQ2X`.
## Priorities
1. Locality: one registration site, feature-local slices.
2. Proof rigor: zero warnings, isolated nested evidence, reviewed pixels.
3. Owner judgement on recorded metrics over velocity.
4. Minimal churn per slice; no big-bang rewrite.
## Decision defaults
| Concern | Default |
| --- | --- |
| Locality vs rigor | Rigor wins |
| New features | Registry contract from day one |
| Existing features | Migrate slice-wise, one named fixture at a time |
| Activation | Startup composition only |
| Internal state API | Typed properties internally; JSON at IPC/diagnostics edges (completed 2026-11-14) |
| Lock/auth | Specialized controllers, never the popup lifecycle |
| Gates | Never weakened to make an iteration green |
## Non-goals
Plugin ecosystem; user configuration; runtime activation; repository-wide rewrite.
## Exit criteria
- [x] One registry entry plus feature-local files is the entire wiring for an ordinary feature (uptime re-measure, 2026-11-14: `Registry.qml` plus its IPC handler in `shell.qml`).
- [x] OSD onboarded by a second agent from public contracts alone (osd-onboarding, with its recorded gaps).
- [x] Owner continuation judgement recorded against baseline metrics (2026-11-13: continue, adjusting first).
- [x] All gates green throughout: lint, JS suite, ui-test (every trunk merge ran `just lint` green; every merge changing behaviour also ran the full `just ui-test` green).
Exit evidence is recorded as met; this is not a phase transition.
Only the human changes `currentPhase` or chooses a successor.
Theme implementation and isolated round-trip proof do not close its separate live human visual acceptance.