Skip to content

Roadmap: Vox Unified Workstream

Milestone: v6.5 Status: In Progress (Phases 378–382) Approved Visual Direction:

  1. Default Theme — Imperial Carrara White & Gold (arcane / carrara): Sunlit White/Cream Honed Carrara Marble (#FAF6EE / #F5EFE2 / #EFE6D5), Burnished Gold mouldings (#C59B27), Deep Espresso Ink (#1C140C), Roman Imperial Emerald & Gilded Laurel Wreath Signets (#1B6B3A / #228B4E with #D4AF37 gold border) for active/working status (crimson #881B1B reserved strictly for errors/alerts), seamless non-cropped marble textures, and universal Romanesque serif typography (Cinzel, Palatino) across all 32 surfaces, sidebar buttons, tooltips, and drawers.
  2. User-Selectable Minimalist Dark Theme — Grey-Blue Basalt Slate (slate / void): Clean, minimalist grey-blue slate palette (#0c0e10 base, #15191c surface, #1f262b elevated, #2b343a subtle border, #f1f5f9 crisp text, #38bdf8/#60a5fa steel-blue & emerald #10b981 status indicators, zero marble textures or decorative columns), configurable in Settings → Theme.

Phase 378: Seamless Roman Marble Texture & Concept Art Generation + Fix Glitchy/Repeating/Offset Background Layers (ROMA-V65-TEXTURES-01..04) [WAVE 1]

Section titled “Phase 378: Seamless Roman Marble Texture & Concept Art Generation + Fix Glitchy/Repeating/Offset Background Layers (ROMA-V65-TEXTURES-01..04) [WAVE 1]”

Goal: Generate seamless frame-free honed white Carrara marble (carrara-seamless-stone.jpg), wide horizontal Roman acanthus frieze band (carrara-acanthus-header-band.jpg), Roman Imperial Emerald & Gilded Laurel Signet emblem (roman-emerald-laurel-seal.jpg), and Milestone v6.5 concept art plates (v66_concept_settings_dual_theme_white_and_slate.jpg, v66_concept_missed_surfaces_scientia_approvals_white_gold.jpg); fix @keyframes vox-marble-sheen 3-layer background-position bug so marble textures never slide or tile; replace framed travertine-parchment-tablet.jpg and repeating roman-architectural-frieze-footer.jpg astrolabe seams across index.css, Glass.tsx, Backdrop.tsx, RomanSurfaceHeader.tsx, BreadcrumbBar.tsx, and RomanTooltip.tsx.

Phase 379: Roman Emerald & Gilded Laurel Working Status System + Full Sidebar & Outside-<main> Chrome/Drawer White-Gold Conversion (ROMA-V65-EMERALD-CHROME-01..04) [WAVE 2A]

Section titled “Phase 379: Roman Emerald & Gilded Laurel Working Status System + Full Sidebar & Outside-<main> Chrome/Drawer White-Gold Conversion (ROMA-V65-EMERALD-CHROME-01..04) [WAVE 2A]”

Goal: Replace red working/live waxSealBadge pills (RomanSurfaceHeader.tsx) and brown bg-accent-secondary live dots (Sidebar.tsx) with high-visibility Roman Imperial Emerald & Gilded Laurel Wreath Signets (#1B6B3A / #228B4E emerald cabochon + #D4AF37 burnished gold rim + crisp vector laurel wreath SVG, legible at all resolutions); convert all dark-brown buttons and containers in Sidebar.tsx (collapse/expand/hide buttons, ⌘K badge, expanded Chat Sessions list, active child items, count badges, operator card), SessionSidebarSection.tsx, BackendStatusPill.tsx, BottomStatusBar.tsx, RomanTooltip.tsx, Omnibar.tsx, Dialog.tsx, and all Drawers (ExplainDrawer, DocViewerDrawer, InspectorDrawer, AchievementsDrawer, ResearchEngineDrawer) to sunlit White/Gold Carrara Marble.

Phase 380: Convert All 15+ Missed Surfaces & Purge Remaining Dark-Brown Remnants Across 85 Files (ROMA-V65-SURFACES-01..04) [WAVE 2B]

Section titled “Phase 380: Convert All 15+ Missed Surfaces & Purge Remaining Dark-Brown Remnants Across 85 Files (ROMA-V65-SURFACES-01..04) [WAVE 2B]”

Goal: Convert all remaining hardcoded dark-brown (#0A0805..#2A2011) classes, inline styles, and canvas fills across SettingsView.tsx, ScientiaDashboard.tsx, CommandCatalogForm.tsx, LudusSandbox.tsx, HudPanels.tsx, ImperiumInspector.tsx, worldRenderer.ts, PublicationsView.tsx, MemoryView.tsx, DiscoverySurface.tsx, ActivitySurface.tsx, ApprovalsView.tsx, PoliciesView.tsx, SubAgentsView.tsx, RepositoryView.tsx, BrowserView.tsx, CodeRabbitView.tsx, CommandCardsView.tsx, Mercatus.tsx, ModelsView.tsx, MensTrainingView.tsx, MensCorpusStudioPanel.tsx, GamifyView.tsx, LeaderboardArenaPanel.tsx, and HarnessRiverPanel.tsx to the unified White/Gold Carrara Marble aesthetic, and extend index.css attribute overrides from main to #root when in the default White/Gold theme.

Phase 381: Dual-Theme Architecture — Preserve & Wire User-Selectable Minimalist Grey-Blue Dark Theme (slate / void) in Settings → Theme (ROMA-V65-DUAL-THEME-01..04) [WAVE 2C]

Section titled “Phase 381: Dual-Theme Architecture — Preserve & Wire User-Selectable Minimalist Grey-Blue Dark Theme (slate / void) in Settings → Theme (ROMA-V65-DUAL-THEME-01..04) [WAVE 2C]”

Goal: Update crates/vox-gui/ui/src/lib/theme.ts, SettingsView.tsx, index.css, Glass.tsx, Backdrop.tsx, Sidebar.tsx, and RomanSurfaceHeader.tsx so users can switch seamlessly in Settings → Theme between Imperial Carrara White & Gold (arcane — Default) and Minimalist Grey-Blue Slate (slate / void) (#0c0e10 / #15191c / #1f262b / #2b343a dark grey-blue palette with crisp sans/mono typography, emerald working indicators, and zero marble textures).

Phase 382: Full 28-Surface + Dual-Theme Playwright & Gemini Computer Vision Flywheel Verification, Build & Live vox axis Relaunch (ROMA-V65-CV-VERIFY-01..04) [WAVE 3]

Section titled “Phase 382: Full 28-Surface + Dual-Theme Playwright & Gemini Computer Vision Flywheel Verification, Build & Live vox axis Relaunch (ROMA-V65-CV-VERIFY-01..04) [WAVE 3]”

Goal: Expand crates/vox-gui/ui/e2e/review/cv-flywheel.spec.ts to capture and audit all 28 navigable SURFACE_REGISTRY surfaces + drawers/modals across both the Default White/Gold Carrara Marble theme and the Minimalist Grey-Blue Slate theme, inspect screenshots with Gemini built-in vision, present inline Before/After carousels, rebuild vox-gui + vox (--features gui), and relaunch vox axis.


Roadmap: v6.4 — Full-Canvas White/Gold Carrara Marble Unification, Non-Procedural Monument Simulator, Single-Line Roman Controls & Zero-Monospace Typography (Phases 373–377)

Section titled “Roadmap: v6.4 — Full-Canvas White/Gold Carrara Marble Unification, Non-Procedural Monument Simulator, Single-Line Roman Controls & Zero-Monospace Typography (Phases 373–377)”

Milestone: v6.4 Status: Complete (5/5 Phases Verified — 2026-11-08) Approved Visual Direction: Milestone v6.4 White/Gold Carrara Marble Concept Art (v65_concept_urbs_simulator_white_gold_acropolis_1794158463815.jpg, v65_concept_explorer_tasks_runs_white_gold_1794158471421.jpg, v65_concept_chat_mens_knowledge_white_gold_1794158501800.jpg) — Full-body Sunlit White/Cream Carrara Marble & Ivory Travertine (#FAF6EE / #F5EFE2 / #EFE6D5), Burnished Gold borders (#C59B27), High-Contrast Dark Espresso Ink (#1C140C), Vitruvian Architectural Watermark background (carrara-page-watermark-bg.jpg), Sculpted Roman Frieze footer (roman-architectural-frieze-footer.jpg), Painted Roman Monument Simulator (urbs-white-gold-citadel-bg.jpg + monument-*.jpg), Single-Line Roman Controls, and Zero Courier/Arial/Menlo Monospace Typography.

Phase 373: Concept Art Generation, Circular ORCHESTRATOR Bubble Cutoff Fix & Single-Line Explorer Controls (ROMA-V64-FLOW-EXP-01..04) [COMPLETE]

Section titled “Phase 373: Concept Art Generation, Circular ORCHESTRATOR Bubble Cutoff Fix & Single-Line Explorer Controls (ROMA-V64-FLOW-EXP-01..04) [COMPLETE]”

Goal: Generate Milestone v6.4 concept art, monument portraits, and background/frieze textures; enlarge the central ORCHESTRATOR circular hub medallion (r=66 in AgentFlow.tsx) and widen the ORCHESTRATOR DAG node (width=136 in NeedsYouSurface.tsx) for zero text cutoff; make OPEN FORM, EXPLAIN CAPABILITY, OPEN DOCS strictly single-line (ExplorerTree.tsx, Button.tsx) and = GENERIC FORM a sleek single-line Roman pill (CapabilityBadge.tsx).

Phase 374: Non-Procedural Painted Roman Monument Simulator & Live File/Folder Aqueduct Streams (ROMA-V64-URBS-MONUMENTS-01..04) [COMPLETE]

Section titled “Phase 374: Non-Procedural Painted Roman Monument Simulator & Live File/Folder Aqueduct Streams (ROMA-V64-URBS-MONUMENTS-01..04) [COMPLETE]”

Goal: Remove procedural 2D polygon building/wall drawing from worldRenderer.ts (paintFullBleedRomanLandscape, drawDistrictCurtainWall, drawIsoFileBuilding, paintDistrictWonder, paintLandmarks); wire LudusSandbox.tsx to urbs-white-gold-citadel-bg.jpg, generated painted monument portraits (monument-pantheon-orchestrator.jpg, monument-temple-compiler.jpg, monument-library-knowledge.jpg, monument-colosseum-harness.jpg), real-time active file/folder badges (src/parser.rs, gamify/LudusSandbox.tsx, src/turso.rs, src/qlora.rs, models/select.rs, src/mesh.rs), and animated SVG golden aqueduct streams.

Section titled “Phase 375: Full-Body Sunlit White/Cream Carrara Marble & Burnished Gold Theme + Architectural Watermark & Frieze Footer (ROMA-V64-WHITE-GOLD-01..04) [COMPLETE]”

Goal: Convert Backdrop.tsx, Glass.tsx, and index.css (:root color tokens, main, tables, rows, inputs, selects, buttons, and legacy dark utilities) across all 32 surfaces to the Sunlit White/Cream Carrara Marble & Burnished Gold aesthetic with subtle Vitruvian architectural watermark background and sculpted bottom Roman frieze footer.

Phase 376: Universal Roman Typography Enforcement & Zero Monospace/Courier/Arial Purge (ROMA-V64-TYPO-01..04) [COMPLETE]

Section titled “Phase 376: Universal Roman Typography Enforcement & Zero Monospace/Courier/Arial Purge (ROMA-V64-TYPO-01..04) [COMPLETE]”

Goal: Remap --font-mono, --font-family-mono, .font-mono, pre, code, kbd, samp, .cm-editor, .cm-content in index.css and dockview-vox.css, plus all SVG/Canvas font declarations across AgentFlow.tsx, worldRenderer.ts, riverCanvasRenderer.ts, NeedsYouSurface.tsx, Matrix.tsx, and MensTrainingView.tsx, to 'Palatino', 'Palatino Linotype', 'Book Antiqua', 'EB Garamond', Georgia, serif (tabular-nums) and 'Cinzel', 'Palatino', Georgia, serif.

Phase 377: Multi-Resolution (1280×800, 1440×900, 1920×1080) Built-In Gemini Vision + Playwright CV Flywheel Verification (ROMA-V64-VERIFY-01..04) [COMPLETE]

Section titled “Phase 377: Multi-Resolution (1280×800, 1440×900, 1920×1080) Built-In Gemini Vision + Playwright CV Flywheel Verification (ROMA-V64-VERIFY-01..04) [COMPLETE]”

Directory: .planning/phases/377-v64-white-gold-carrara-marble-unification/ Goal: Verify all 32 SURFACE_REGISTRY surfaces and 3 viewport resolutions (1280×800, 1440×900, 1920×1080) via Playwright + @ngi/platform-ssot CV Flywheel (0 axeViolations, 0 clippedInteractiveCount, 0 overlappingControlPairs, 0 digitalColorLeakCount), inspect screenshots with built-in Gemini vision, eliminate all Vite/CSS/bundler warnings, and present inline Before/After carousels.


Roadmap: v6.3 — Game-UI Romanesque Unification, 92px Sculptural Medallion Sidebar Rail, Sunlit Travertine & Parchment Stone Tablets, Zero-Void/Zero-Truncation Multi-Resolution Layout & Built-In VLM Vision Flywheel (Phases 368–372)

Section titled “Roadmap: v6.3 — Game-UI Romanesque Unification, 92px Sculptural Medallion Sidebar Rail, Sunlit Travertine & Parchment Stone Tablets, Zero-Void/Zero-Truncation Multi-Resolution Layout & Built-In VLM Vision Flywheel (Phases 368–372)”

Milestone: v6.3 Status: Executing with Parallel Subagent Waves (image-generator + self). Updated: 2026-11-08 Approved Visual Direction: Game-UI Romanesque Concept Art (v61_concept_compute_and_market_1794141010387.jpg, v61_concept_workspace_and_commands_1794141011579.jpg, v62_concept_semiotic_urbs_codebase_city_1794147288395.jpg) — 92px Sculptural Bronze Medallion Sidebar Rail, Sunlit Travertine & Parchment Stone Tablets (#EDE3D1/#DFD3BD), Riveted Burnished Brass Console Bezels, Large Semicircular Brass Chronometer Dials, Olive-Laurel Bronze & Crimson Wax Seal Badges (zero digital neon green/cyan/blue), and zero clipping/truncation/empty void across 1280×800, 1440×900, and 1920×1080.

Phase 368: Built-In VLM Vision Audit & Structural/Chromatic Flywheel Upgrade (vlmVisionFlywheel.ts, audits.ts, cv-flywheel.spec.ts) (ROMA-CV-V63-01..04) [WAVE 1]

Section titled “Phase 368: Built-In VLM Vision Audit & Structural/Chromatic Flywheel Upgrade (vlmVisionFlywheel.ts, audits.ts, cv-flywheel.spec.ts) (ROMA-CV-V63-01..04) [WAVE 1]”

Goal: Upgrade the Playwright + VLM Computer Vision Flywheel (vlmVisionFlywheel.ts, audits.ts, cv-flywheel.spec.ts) to detect child clipping inside overflow-hidden containers (clippedChildInOverflowHiddenCount), mid-word text truncation (truncatedTextElementCount), empty viewport void (emptyViewportVoidRatio), digital green/teal/cyan pixel leaks (digitalGreenBluePixelRatio), and sunlit travertine/parchment tablet coverage (travertineParchmentSlabRatio). Depends on: Phase 367 Requirements: ROMA-CV-V63-01, ROMA-CV-V63-02, ROMA-CV-V63-03, ROMA-CV-V63-04

Phase 369: 92px Sculptural Bronze Medallion Game-UI Sidebar Rail (Sidebar.tsx, Layout.tsx, StatusBar.tsx) & Zero-Truncation Status Bar (ROMA-RAIL-CHROME-01..04) [WAVE 2A]

Section titled “Phase 369: 92px Sculptural Bronze Medallion Game-UI Sidebar Rail (Sidebar.tsx, Layout.tsx, StatusBar.tsx) & Zero-Truncation Status Bar (ROMA-RAIL-CHROME-01..04) [WAVE 2A]”

Goal: Transform Sidebar.tsx from a 216px SaaS sidebar list into the Concept Art’s 92px Sculptural 3D Bronze Coin Medallion Game-UI Rail (embossed circular bronze medallions on top, chiseled uppercase Cinzel labels beneath, glowing vertical gold active bar, reclaiming 124px of horizontal screen space); fix StatusBar.tsx truncation (ROUTING Auto → anthr) and replace neon green/cyan dots with Romanesque gold, olive-laurel bronze, and crimson wax seals. Depends on: Phase 368 Requirements: ROMA-RAIL-CHROME-01, ROMA-RAIL-CHROME-02, ROMA-RAIL-CHROME-03, ROMA-RAIL-CHROME-04

Phase 370: Sunlit Travertine & Parchment Stone Tablets, Large Semicircular Brass Chronometer Dials & Zero-Digital-Color Purge (index.css, RomanSurfaceHeader.tsx, All 32 Surfaces) (ROMA-TABLET-GAMEUI-01..04) [WAVE 2B]

Section titled “Phase 370: Sunlit Travertine & Parchment Stone Tablets, Large Semicircular Brass Chronometer Dials & Zero-Digital-Color Purge (index.css, RomanSurfaceHeader.tsx, All 32 Surfaces) (ROMA-TABLET-GAMEUI-01..04) [WAVE 2B]”

Goal: Replace dark #0E0C0A/#14100B IBM-style card/table bodies and digital neon green/teal/cyan pills across index.css and surface components (ModelsView, VoxPopuliView, MercatusView, TerminalSurface, RepositoryView, CommandCatalogSurface, SkillsPluginsView, MemoryView, ScientiaSurface, ExplorerView, SettingsView, Dashboard, HarnessRiverPanel) with Warm Sunlit Roman Travertine & Parchment Stone Tablets (#EDE3D1 / #DFD3BD with #1E150C deep espresso ink), Riveted Brass Console Bezels, large semicircular <RomanBrassGauge> chronometer dials, and Olive-Laurel / Crimson Wax Seal badges. Depends on: Phase 368 Requirements: ROMA-TABLET-GAMEUI-01, ROMA-TABLET-GAMEUI-02, ROMA-TABLET-GAMEUI-03, ROMA-TABLET-GAMEUI-04

Phase 371: Multi-Resolution (1280×800, 1440×900, 1920×1080) Zero-Clipping & Empty-Void Elimination (LudusSandbox.tsx, HarnessRiverPanel.tsx, VoxPopuliView.tsx, MercatusView.tsx, TerminalSurface.tsx) (ROMA-VOID-RES-01..04) [WAVE 2C]

Section titled “Phase 371: Multi-Resolution (1280×800, 1440×900, 1920×1080) Zero-Clipping & Empty-Void Elimination (LudusSandbox.tsx, HarnessRiverPanel.tsx, VoxPopuliView.tsx, MercatusView.tsx, TerminalSurface.tsx) (ROMA-VOID-RES-01..04) [WAVE 2C]”

Goal: Eliminate bottom dock/radar collision and top toolbar wrapping at 1280×800 in LudusSandbox.tsx, eliminate bottom funnel clipping at 1280×800 in HarnessRiverPanel.tsx, and replace empty bottom viewport voids in VoxPopuliView.tsx, MercatusView.tsx, and TerminalSurface.tsx with interactive Roman Mesh Topology, Arbitrage Matrix, and Command Quick-Launch Stone Tablets. Depends on: Phases 369–370 Requirements: ROMA-VOID-RES-01, ROMA-VOID-RES-02, ROMA-VOID-RES-03, ROMA-VOID-RES-04

Phase 372: Multi-Pass Built-In Vision + Playwright CV Flywheel Verification & Inline Before/After Presentation (ROMA-V63-VERIFY-01..04) [WAVE 3]

Section titled “Phase 372: Multi-Pass Built-In Vision + Playwright CV Flywheel Verification & Inline Before/After Presentation (ROMA-V63-VERIFY-01..04) [WAVE 3]”

Goal: Run multi-pass Playwright + VLM CV Flywheel audits across all 32 surfaces and 3 resolutions (1280×800, 1440×900, 1920×1080), inspect every screenshot directly with Gemini built-in vision against the Concept Art, verify pnpm --dir crates/vox-gui/ui build, and present inline Before/After carousels. Depends on: Phase 371 Requirements: ROMA-V63-VERIFY-01, ROMA-V63-VERIFY-02, ROMA-V63-VERIFY-03, ROMA-V63-VERIFY-04


Roadmap: v6.2 — Romanesque Semiotic Codebase City, Aqueduct LLM/Orchestrator Cascade & Gamified Developer Controls (Phases 363–367)

Section titled “Roadmap: v6.2 — Romanesque Semiotic Codebase City, Aqueduct LLM/Orchestrator Cascade & Gamified Developer Controls (Phases 363–367)”

Milestone: v6.2 Status: Executing with Parallel Subagent Waves (image-generator + self). Updated: 2026-11-08 Research & Design SSOTs:

Phase 363: Nov 2026 Semiotic Codebase & Aqueduct Orchestrator Research KB + High-Resolution Concept Art & Simulator Assets (ROMA-KB-CONCEPT-01..04) [WAVE 1]

Section titled “Phase 363: Nov 2026 Semiotic Codebase & Aqueduct Orchestrator Research KB + High-Resolution Concept Art & Simulator Assets (ROMA-KB-CONCEPT-01..04) [WAVE 1]”

Goal: Research external software city visualizers (CodeCharta, CodeCity, M3triCity), multi-agent LLM orchestrator graphs (LangGraph Studio, AutoGen Studio, Rivet, Arize Phoenix), and hydraulic city-builder UX mechanics (Anno 117: Pax Romana, Timberborn, Caesar III), persist the findings in docs/src/architecture/semiotic-codebase-and-aqueduct-orchestrator-simulator-research-2026.md, and generate high-resolution Concept Art plates + simulator/aqueduct raster assets (urbs-semiotic-citadel-bg.jpg, aqueduct-sluice-cascade-bg.jpg). Depends on: Phase 362 Requirements: ROMA-KB-CONCEPT-01, ROMA-KB-CONCEPT-02, ROMA-KB-CONCEPT-03, ROMA-KB-CONCEPT-04

Phase 364: Semiotic Codebase City (L0–L5), White Carrara Marble Temples & Interactive Developer Command Dock (LudusSandbox.tsx, worldRenderer.ts) (ROMA-URBS-SEMIOTIC-01..04) [WAVE 2A]

Section titled “Phase 364: Semiotic Codebase City (L0–L5), White Carrara Marble Temples & Interactive Developer Command Dock (LudusSandbox.tsx, worldRenderer.ts) (ROMA-URBS-SEMIOTIC-01..04) [WAVE 2A]”

Goal: Upgrade worldRenderer.ts and LudusSandbox.tsx with high-contrast white Carrara marble temple/basilica colonnades, gilded domes, semiotic architectural layer badges (L0 Foundation through L5 Surfaces), hydraulic token-stream particles, a Semiotic Layer & Legend Bar ([All | L0..L5]), and a 1-click Developer Action Command Dock ([⚙ Check Crate], [✓ Run Tests], [🔍 Diff Loupe], [📌 Pin Context], [🔒 Quarantine], [⚔ Spawn Agent], [🏆 Claim Yield]). Depends on: Phase 363 Requirements: ROMA-URBS-SEMIOTIC-01, ROMA-URBS-SEMIOTIC-02, ROMA-URBS-SEMIOTIC-03, ROMA-URBS-SEMIOTIC-04

Phase 365: Aqueduct Hydraulic LLM & Orchestrator Cascade, 8-Spoke Source Arches & Interactive Brass Sluice Gates (HarnessRiverPanel.tsx, riverCanvasRenderer.ts) (ROMA-AQUEDUCT-FLOW-01..04) [WAVE 2B]

Section titled “Phase 365: Aqueduct Hydraulic LLM & Orchestrator Cascade, 8-Spoke Source Arches & Interactive Brass Sluice Gates (HarnessRiverPanel.tsx, riverCanvasRenderer.ts) (ROMA-AQUEDUCT-FLOW-01..04) [WAVE 2B]”

Goal: Transform HarnessRiverPanel.tsx and riverCanvasRenderer.ts into a multi-tiered Roman Aqueduct LLM & Orchestrator Cascade backed by aqueduct-sluice-cascade-bg.jpg, featuring the Aquae Caput (Hub + 8-Spoke Source Arcade), 3-Tier Marble Channels (Cursus I/II/III) with live shape+color token particles (❯❯ Cyan Supply, ◆◆ Gold Token Stream, ◉ Emerald Verified, ▲ Crimson Fault), 4 Interactive Brass Sluice Gates (Clausurae: AST Parse, Typecheck, TOESTUB Audit, Test Suite), and 1-click developer orchestrator actions ([Dispatch Task], [Flush Sluice], [Run Check], [Compact Context], [Claim Quest Yield]). Depends on: Phase 363 Requirements: ROMA-AQUEDUCT-FLOW-01, ROMA-AQUEDUCT-FLOW-02, ROMA-AQUEDUCT-FLOW-03, ROMA-AQUEDUCT-FLOW-04

Phase 366: Gamified Developer Loop & Bi-Directional Surface Navigation (ROMA-GAMIFY-DEV-01..04) [WAVE 2C]

Section titled “Phase 366: Gamified Developer Loop & Bi-Directional Surface Navigation (ROMA-GAMIFY-DEV-01..04) [WAVE 2C]”

Goal: Connect developer actions in LudusSandbox and HarnessRiverPanel to real backend/RPC operations and live gamified XP/Denarii quest yield feedback with toast/status confirmation and 1-click surface navigation. Depends on: Phases 364–365 Requirements: ROMA-GAMIFY-DEV-01, ROMA-GAMIFY-DEV-02, ROMA-GAMIFY-DEV-03, ROMA-GAMIFY-DEV-04

Phase 367: Multi-Pass VLM + Playwright Computer Vision Flywheel Convergence & Inline Before/After Presentation (ROMA-V62-FLYWHEEL-01..04) [WAVE 3]

Section titled “Phase 367: Multi-Pass VLM + Playwright Computer Vision Flywheel Convergence & Inline Before/After Presentation (ROMA-V62-FLYWHEEL-01..04) [WAVE 3]”

Goal: Capture true Before baselines, upgrade vlmVisionFlywheel.ts and cv-flywheel.spec.ts to verify semiotic layer badges, aqueduct sluice gates, and developer action docks, run multi-pass Playwright + VLM CV Flywheel audits across all surfaces, build vox-gui, and present Concept Art + Before/After carousels inline. Depends on: Phase 366 Requirements: ROMA-V62-FLYWHEEL-01, ROMA-V62-FLYWHEEL-02, ROMA-V62-FLYWHEEL-03, ROMA-V62-FLYWHEEL-04

Phase 368: Romanesque MENS Corpus & Pipeline Studio GUI + Auto-Sync JSONL Flywheel (MENS-STUDIO-GUI-01..04)

Section titled “Phase 368: Romanesque MENS Corpus & Pipeline Studio GUI + Auto-Sync JSONL Flywheel (MENS-STUDIO-GUI-01..04)”

Directory: .planning/phases/368-mens-corpus-pipeline-studio-gui/ Goal: Eliminate manual JSONL regeneration and surface the entire MENS corpus/mix/reasoning/checkpoint pipeline in vox-gui (MensCorpusStudioPanel.tsx inside MensTrainingView.tsx), matching the Romanesque Concept Art (concept_mens_pipeline_studio_romanesque_1794153153510.jpg) via CV Flywheel convergence:

  1. Auto-materialize missing/stale synthetic and extracted JSONL sources (vox_reasoning_traces.jsonl, docs.jsonl, rust_source.jsonl) and auto-regenerate train_mixed*.jsonl + mens/data/train.jsonl whenever source fingerprints or weights change.
  2. Expose real Tauri IPC commands (mens_corpus_studio_snapshot, mens_corpus_read_jsonl, mens_corpus_save_jsonl_row, mens_corpus_rebuild_mix) backed by vox_corpus and vox_compiler::parser::parse AST verification — zero hardcoded/fake data.
  3. Build the 3-arch Romanesque MensCorpusStudioPanel.tsx (Mix Sources & Auto-Materializer DAG, Live JSONL Record Inspector & Inline AST Editor, Ladder Checkpoints & Collateral Gate Matrix) with live search, task-family filtering, inline <think> & Vox code editing, compiler diagnostics, and one-click/auto mix rebuild.
  4. Verify with Rust unit tests, Vitest component tests, and Playwright CV Flywheel visual convergence against the Romanesque concept art. Depends on: Phase 362 Requirements: MENS-STUDIO-GUI-01, MENS-STUDIO-GUI-02, MENS-STUDIO-GUI-03, MENS-STUDIO-GUI-04 Plans: 1/1 plan complete (368-01-PLAN.md -> 368-01-SUMMARY.md) — Complete

Roadmap: v6.1 — Full-Surface Plate 2 Romanesque Generalization, High-Resolution Diorama Simulator & Structural CV Flywheel (Phases 358–361)

Section titled “Roadmap: v6.1 — Full-Surface Plate 2 Romanesque Generalization, High-Resolution Diorama Simulator & Structural CV Flywheel (Phases 358–361)”

Milestone: v6.1 Status: Executing with Parallel Subagent Waves (pro + flash disjoint file scopes). Updated: 2026-11-08 Approved Visual Direction: Concrete Plate 2 Romanesque Suite (axis_romanesque_concept_2_urbs_ludus_1794129218189.jpg, v61_concept_compute_and_market_1794141010387.jpg, v61_concept_workspace_and_commands_1794141011579.jpg, v61_concept_knowledge_explorer_urbs_refined_1794141055158.jpg) — Universal carved Roman marble bas-relief frieze headers (<RomanSurfaceHeader>), embossed circular bronze medallions (<RomanMedallion>), semicircular antique brass dial gauges (<RomanBrassGauge>), crimson wax-seal status badges, warm travertine stone tablets, readable humanist typography, high-resolution transparent-composited urbs-diorama-bg.jpg simulator canvas, and upgraded CV Flywheel detectors (skeletonCount === 0, digitalColorLeakCount === 0, rawBacktickCount === 0).

Phase 358: Structural CV Flywheel Upgrade (audits.ts, cvFlywheel.ts, cv-flywheel.spec.ts) (ROMA-CV-STRUCT-01..04) [WAVE 1A]

Section titled “Phase 358: Structural CV Flywheel Upgrade (audits.ts, cvFlywheel.ts, cv-flywheel.spec.ts) (ROMA-CV-STRUCT-01..04) [WAVE 1A]”

Goal: Upgrade the automated Playwright CV Flywheel (audits.ts, cvFlywheel.ts, cv-flywheel.spec.ts) so it catches and fails on non-Romanesque regressions across all 32 surfaces: loading skeleton bars (skeletonCount), cold digital slate/cyan/violet leaks (digitalColorLeakCount), raw unrendered markdown backticks (rawBacktickCount), and opaque canvas fills obscuring the Roman diorama in LudusSandbox. Depends on: Phase 357 Requirements: ROMA-CV-STRUCT-01, ROMA-CV-STRUCT-02, ROMA-CV-STRUCT-03, ROMA-CV-STRUCT-04

Phase 359: Concrete Non-Abstract Concept Art Plates & High-Resolution Roman Diorama Simulator (LudusSandbox.tsx, worldRenderer.ts, hydraulicNetwork.ts) (ROMA-SIM-DIORAMA-01..04) [WAVE 1B]

Section titled “Phase 359: Concrete Non-Abstract Concept Art Plates & High-Resolution Roman Diorama Simulator (LudusSandbox.tsx, worldRenderer.ts, hydraulicNetwork.ts) (ROMA-SIM-DIORAMA-01..04) [WAVE 1B]”

Goal: Generate concrete, non-abstract Plate 2 Romanesque concept art plates for COMPUTE & MARKET, WORKSPACE & COMMANDS, and KNOWLEDGE, EXPLORER & URBS SIMULATOR; eliminate the double opaque dark-blue canvas fill (#0a1019 in LudusSandbox.tsx and #040e18 in worldRenderer.ts) that obliterated 90% of urbs-diorama-bg.jpg; and replace neon cyan viaduct/conduit strokes in worldRenderer.ts and hydraulicNetwork.ts with burnished Roman gold (#D4AF37), warm travertine (#F3E5AB), and verdigris aqueduct emerald (#10B981). Depends on: Phase 358 Requirements: ROMA-SIM-DIORAMA-01, ROMA-SIM-DIORAMA-02, ROMA-SIM-DIORAMA-03, ROMA-SIM-DIORAMA-04

Phase 360: Universal Plate 2 Romanesque Generalization Across COMPUTE, MARKET, WORKSPACE, COMMANDS, KNOWLEDGE, REVIEW, EXPLORER & SETTINGS (ROMA-ALL-TABS-01..04) [PARALLEL-SAFE WAVE 2]

Section titled “Phase 360: Universal Plate 2 Romanesque Generalization Across COMPUTE, MARKET, WORKSPACE, COMMANDS, KNOWLEDGE, REVIEW, EXPLORER & SETTINGS (ROMA-ALL-TABS-01..04) [PARALLEL-SAFE WAVE 2]”

Goal: Build <RomanSurfaceHeader> (carved marble bas-relief frieze banner, embossed bronze medallion, semicircular <RomanBrassGauge> KPI dials, crimson wax-seal status pill, and progressive-disclosure <RomanTooltip>) + global Plate 2 CSS table/card/input overrides in index.css, and generalize Plate 2 Romanesque styling across ModelsView, CommandCardsView (Populi & Oratio), Mercatus, RepositoryView, Console, BrowserView, CodeRabbitView, Catalog, CommandCatalogForm, SkillsPluginsView, MemoryView, ScientiaSurface, ResearchView, PublicationsView, VoxGraphStatusPanel, ApprovalsView, PoliciesView, ExplorerView (instant zero-skeleton hydration), and SettingsView. Depends on: Phase 359 Requirements: ROMA-ALL-TABS-01, ROMA-ALL-TABS-02, ROMA-ALL-TABS-03, ROMA-ALL-TABS-04

Phase 361: Multi-Round 32-Surface Before/After Playwright CV Flywheel Verification & Inline Presentation (ROMA-V61-VERIFY-01..04) [WAVE 3]

Section titled “Phase 361: Multi-Round 32-Surface Before/After Playwright CV Flywheel Verification & Inline Presentation (ROMA-V61-VERIFY-01..04) [WAVE 3]”

Goal: Run multi-round Playwright CV Flywheel verification across all 32 surfaces (skeletonCount === 0, digitalColorLeakCount === 0, rawBacktickCount === 0, overlappingControlPairs === 0, axeViolations === 0), capture before/after screenshots across COMPUTE, MARKET, WORKSPACE, COMMANDS, KNOWLEDGE, EXPLORER, and URBS SIMULATOR, rebuild vox-gui, and present inline before/after carousels. Depends on: Phase 360 Requirements: ROMA-V61-VERIFY-01, ROMA-V61-VERIFY-02, ROMA-V61-VERIFY-03, ROMA-V61-VERIFY-04

Phase 362: MENS Regression Root-Cause Fixes (supervised_ratio, <think> Reasoning Synthesis & Cued Answer Grader) (MENS-REGFIX-01..04)

Section titled “Phase 362: MENS Regression Root-Cause Fixes (supervised_ratio, <think> Reasoning Synthesis & Cued Answer Grader) (MENS-REGFIX-01..04)”

Goal: Eliminate the four verified root causes of MENS training/collateral regressions across the Qwen3 ladder:

  1. Switch generate_training_system_prompt() in crates/vox-corpus/src/training/mod.rs from generate_system_prompt() (~2,200 tokens / 8.6 KB mens/config/system_prompt.txt) to builtin_system_prompt() (~350 tokens / 1.48 KB), raising supervised_ratio from 4.5% (< 10% gate failure) to ~20–25% and speeding up QLoRA step time by ~4x.
  2. Add deterministic, compiler-verified, zero-leakage Vox <think> reasoning trace synthesis (crates/vox-corpus/src/corpus/vox_reasoning_synth.rs: 120 rows across literal_debug, state_trace, and invariant_check) and wire mens/data/mix_sources/vox_reasoning_traces.jsonl (weight: 2.0) into mens/config/mix.yaml, mens/config/mix-vox-lang.yaml, and mix_prepare.rs so adapters do not auto-fix literal-execution debug items or ramble past 2048 tokens without closing </think>.
  3. Harden extract_cued_answer in crates/vox-eval/src/qwen_reasoning.rs (scanning cues right-to-left, rejecting English prose stopwords like "The"/"that"/"not", and recovering earlier valid cues when a trace truncates right after "So the answer is ") and sync qwen_reasoning.rs v2 across worktrees.
  4. Use --rank 8 --alpha 16 (grad_accum=8) when retraining 28-layer models (0.6B, 1.7B) so lm_head + projection LoRA updates preserve base reasoning calibration. Depends on: Phase 333, Phase 345 Requirements: MENS-REGFIX-01, MENS-REGFIX-02, MENS-REGFIX-03, MENS-REGFIX-04 Plans: 1 plan (362-01-PLAN.md — Complete 2026-11-08)

Roadmap: v5.9 — Full-Prompt Goal Audit & Universal Plate 2 Romanesque Convergence (Phases 352–356)

Section titled “Roadmap: v5.9 — Full-Prompt Goal Audit & Universal Plate 2 Romanesque Convergence (Phases 352–356)”

Milestone: v5.9 Status: Executing with Parallel Subagent Waves (pro + flash disjoint file scopes). Updated: 2026-11-08 Approved Visual Direction: Concept Art Plate 2 (axis_romanesque_concept_2_urbs_ludus_1794129218189.jpg) — Universal carved Roman travertine & burnished brass design system across all 32 Vox Axis surfaces, Plate 2 exact 5-medallion top frieze & 3-part Stone Telemetry Console (Git Timeline, 3 RomanBrassGauge dials, Architecture Astrolabe Radar) on LudusSandbox, complete elimination of residual sci-fi “Argon AI” cyan/void visuals in riverCanvasRenderer.ts and AgentTrajectoryBraid.tsx, 1-click navigable monuments from both Dashboard and GamifyView, universal progressive-disclosure RomanTooltip & Tooltip across all surfaces, live Chat Orchestrator→Spoke & Hub Model visualizer, and 32-surface × 3-resolution CV Flywheel convergence.

Phase 352: Plate 2 Exact Urbs Stone Telemetry Console, Top-5 Medallion Frieze & 1-Click Navigable Monuments (ROMA-URBS-CONSOLE-01..04) [PARALLEL-SAFE WAVE 1A — PRO]

Section titled “Phase 352: Plate 2 Exact Urbs Stone Telemetry Console, Top-5 Medallion Frieze & 1-Click Navigable Monuments (ROMA-URBS-CONSOLE-01..04) [PARALLEL-SAFE WAVE 1A — PRO]”

Goal: Converge LudusSandbox.tsx, HudPanels.tsx, ViaTemporisScrubber.tsx, GamifyView.tsx, and surfaceComponents.tsx with Concept Art Plate 2 by adding the top-row 5-medallion carved marble frieze (Workspace Map, Agent Orchestrator, CI & Verification, Task Harbor, Token Treasury), upgrading the bottom stone telemetry console so all 3 gauges (Budget, Energy, Speed) render as animated semicircular RomanBrassGauge dials flanked by carved Roman column dividers (I and III), adding the crimson wax seal (wax-seal-medallion.jpg) + laurel wreath to ViaTemporisScrubber.tsx, threading onNavigate into GamifyView.tsx, and making every monument plaque and landmark medallion 1-click navigable. Depends on: Phase 351 Requirements: ROMA-URBS-CONSOLE-01, ROMA-URBS-CONSOLE-02, ROMA-URBS-CONSOLE-03, ROMA-URBS-CONSOLE-04

Phase 353: Sci-Fi “Argon AI” Elimination in HarnessRiverPanel (riverCanvasRenderer.ts, AgentTrajectoryBraid.tsx), AgentFlow & NeedsYouSurface (ROMA-ARGON-PURGE-01..04) [PARALLEL-SAFE WAVE 1B — FLASH]

Section titled “Phase 353: Sci-Fi “Argon AI” Elimination in HarnessRiverPanel (riverCanvasRenderer.ts, AgentTrajectoryBraid.tsx), AgentFlow & NeedsYouSurface (ROMA-ARGON-PURGE-01..04) [PARALLEL-SAFE WAVE 1B — FLASH]”

Goal: Replace the remaining cold sci-fi #070709 black background, glowing cyan singularity orb, and laser scanner line in riverCanvasRenderer.ts and AgentTrajectoryBraid.tsx with warm Roman aqueduct stone, molten gold (#D4AF37), amber-bronze (#D97706), verdigris emerald (#10B981), and a carved travertine & brass keystone banner; reskin AgentFlow.tsx (AGENTS › FLOW) and NeedsYouSurface.tsx to Plate 2 marble frieze headers, embossed bronze/gold node rings, and <RomanTooltip> progressive disclosure. Depends on: Phase 351 Requirements: ROMA-ARGON-PURGE-01, ROMA-ARGON-PURGE-02, ROMA-ARGON-PURGE-03, ROMA-ARGON-PURGE-04

Phase 354: Universal Plate 2 Romanesque Card/Tooltip Primitives (Glass, Kpi, Tooltip), Live Chat Orchestrator→Spoke Visualizer & ExplorerView Hydration (ROMA-UNIVERSAL-01..04) [PARALLEL-SAFE WAVE 1C — FLASH]

Section titled “Phase 354: Universal Plate 2 Romanesque Card/Tooltip Primitives (Glass, Kpi, Tooltip), Live Chat Orchestrator→Spoke Visualizer & ExplorerView Hydration (ROMA-UNIVERSAL-01..04) [PARALLEL-SAFE WAVE 1C — FLASH]”

Goal: Upgrade Glass.tsx, Kpi.tsx, and Tooltip.tsx to Plate 2’s warm carved travertine (#15110C / #1B160F) and burnished brass (#D4AF37) aesthetic so all 32 surfaces share a unified Romanesque skin; enhance ChatExecutionRail.tsx with a prominent Plate 2 Live Orchestrator Hub & Active Spoke visualizer with <RomanTooltip> progressive disclosure; and hydrate get_capability_index in tauriMockRich.ts so ExplorerView renders rich capability cards without loading skeletons. Depends on: Phase 351 Requirements: ROMA-UNIVERSAL-01, ROMA-UNIVERSAL-02, ROMA-UNIVERSAL-03, ROMA-UNIVERSAL-04

Phase 355: Multi-Resolution (1280×800, 1440×900, 1920×1080) 32-Surface CV Flywheel Pass & Full-Goal Audit Matrix (ROMA-AUDIT-CV-01..04) [WAVE 2]

Section titled “Phase 355: Multi-Resolution (1280×800, 1440×900, 1920×1080) 32-Surface CV Flywheel Pass & Full-Goal Audit Matrix (ROMA-AUDIT-CV-01..04) [WAVE 2]”

Goal: Execute multi-resolution Playwright CV Flywheel verification across all 32 surfaces and 3 resolutions (1280×800, 1440×900, 1920×1080), capture updated before/after visual artifacts, verify every prompt requirement end-to-end, rebuild the production bundle, and relaunch vox-gui. Depends on: Phases 352–354 Requirements: ROMA-AUDIT-CV-01, ROMA-AUDIT-CV-02, ROMA-AUDIT-CV-03, ROMA-AUDIT-CV-04

PhaseRequirementsStatus
352 Plate 2 Exact Urbs Stone Telemetry Console, Top-5 Medallion Frieze & 1-Click Navigable MonumentsROMA-URBS-CONSOLE-01..04Complete (2026-11-08)
353 Sci-Fi “Argon AI” Elimination in HarnessRiverPanel, AgentFlow & NeedsYouSurfaceROMA-ARGON-PURGE-01..04Complete (2026-11-08)
354 Universal Plate 2 Primitives (Glass, Kpi, Tooltip), Chat Orchestrator Visualizer & ExplorerViewROMA-UNIVERSAL-01..04Complete (2026-11-08)
355 Multi-Resolution 32-Surface CV Flywheel Pass & Full-Goal Audit MatrixROMA-AUDIT-CV-01..04Complete (2026-11-08)

Roadmap: v5.8 — Plate 2 Romanesque Axis Overhaul, Plain-English Navigable Surfaces, Generated Art Assets, Progressive-Disclosure Tooltips & 32-Surface CV Flywheel (Phases 346–351) — COMPLETE

Section titled “Roadmap: v5.8 — Plate 2 Romanesque Axis Overhaul, Plain-English Navigable Surfaces, Generated Art Assets, Progressive-Disclosure Tooltips & 32-Surface CV Flywheel (Phases 346–351) — COMPLETE”

Milestone: v5.8 Status: Complete (24/24 requirements verified 2026-11-08) — Extended in v5.9 (Phases 352–355). Updated: 2026-11-08 Approved Visual Direction: Concept Art Plate 2 (axis_romanesque_concept_2_urbs_ludus_1794129218189.jpg) — Carved marble frieze headers, embossed bronze/gold medallions, warm travertine tablets, crimson wax seals, brass gauges, and engraved astrolabe radar, with 100% plain-English labels, generated raster art assets (no crude polygons), clickable/navigable map buildings & radar regions, Plate 2-styled Sidebar, and Romanesque progressive-disclosure tooltips explaining what every feature is for and what clicking it does.

Phase 346: Generated Romanesque Art Assets, Plate 2 Marble Frieze Sidebar, Plain-English Semiotics & Progressive-Disclosure Tooltips (ROMA-CHROME-01..04) [PARALLEL-SAFE WAVE 1]

Section titled “Phase 346: Generated Romanesque Art Assets, Plate 2 Marble Frieze Sidebar, Plain-English Semiotics & Progressive-Disclosure Tooltips (ROMA-CHROME-01..04) [PARALLEL-SAFE WAVE 1]”

Goal: Generate real raster art assets (public/assets/romanesque/* — carved marble frieze, engraved astrolabe dial, isometric Roman architectural diorama & monuments, bronze medallion icons, crimson wax seal), reskin Sidebar.tsx and BreadcrumbBar.tsx into Plate 2’s carved marble frieze & bronze medallion style (fixing the top-left [ Sidebar › ] collision), replace emojis with bespoke SVG/art semiotics (Icons.tsx), and ship the Romanesque Progressive-Disclosure Tooltip component (RomanTooltip.tsx) that explains in plain English what every button, knob, and surface is for and what interacting with it does. Depends on: Nothing (Wave 1) Requirements: ROMA-CHROME-01, ROMA-CHROME-02, ROMA-CHROME-03, ROMA-CHROME-04 Success Criteria (what must be TRUE):

  1. Generated Romanesque art assets (frieze-header.jpg, astrolabe-dial.jpg, urbs-diorama-bg.jpg, wax-seal-badge.jpg) live under crates/vox-gui/ui/public/assets/romanesque/ and replace polygon-only placeholders.
  2. Sidebar.tsx and BreadcrumbBar.tsx use Plate 2’s carved marble frieze & bronze medallion styling with plain-English navigation labels, Romanesque progressive-disclosure tooltips on every nav item, and zero overlap when sidebarMode === 'hidden' across 1280×800, 1440×900, and 1920×1080.
  3. RomanTooltip.tsx and SurfaceHelpTooltip.tsx render carved travertine & burnished brass progressive-disclosure tooltips explaining What it is, What it does, and Click action / shortcut in clear English.

Phase 347: Plain-English Interactive Workspace Map (LudusSandbox), Navigable Buildings, Animated Astrolabe Radar & Dashboard Minimap Isolation (ROMA-DASH-URBS-01..04) [PARALLEL-SAFE WAVE 2A]

Section titled “Phase 347: Plain-English Interactive Workspace Map (LudusSandbox), Navigable Buildings, Animated Astrolabe Radar & Dashboard Minimap Isolation (ROMA-DASH-URBS-01..04) [PARALLEL-SAFE WAVE 2A]”

Goal: Wrap Dashboard.tsx’s embedded mini-map in <EmbeddedSurfaceProvider value={true}>, replace all cryptic Latin-only labels in LudusSandbox.tsx, HudPanels.tsx, and ViaTemporisScrubber.tsx with clear English (Workspace Map, Orchestrator, MCP Tools, CI & Gates, Task Queue, Git History, Compiler, Database, MENS Models, Mesh Network, Security, Git Timeline, Priority Alerts, Architecture Radar), make every mapped building/landmark and radar region clickable and navigable to its real Vox Axis surface, replace the 148-polygon radar smear with the generated/animated 7-Region Architecture Astrolabe Radar, and animate brass gauges, radar sweep, and hydraulic flow. Depends on: Phase 346 Requirements: ROMA-DASH-URBS-01, ROMA-DASH-URBS-02, ROMA-DASH-URBS-03, ROMA-DASH-URBS-04 Success Criteria (what must be TRUE):

  1. Dashboard.tsx embeds <LudusSandbox /> via <EmbeddedSurfaceProvider value={true}> with zero HUD pileups and clear plain-English controls + progressive-disclosure tooltips.
  2. Every landmark pill, mapped building (Orchestrator, Compiler, GUI & Surfaces, Database, MENS Models, Mesh Network, Security & Crypto), and radar sector (Foundation, Core, Orchestrator, Runtime, MENS, Mesh & DB, GUI) is interactive: clicking focuses the region and provides a working [Open <Surface> ↗] action that navigates directly to that surface (onNavigate).
  3. The 148-polygon wireframe radar is replaced by the Plate 2 engraved marble & brass Architecture Astrolabe Radar with animated sweep, glowing hotspot braziers, draggable/clickable viewport frustum, and zero clipped background canvas text (AQUA Supply).

Phase 348: Plate 2 Romanesque Reskin of GamifyView (Arena & Quests) & HarnessRiverPanel (Pipeline Cascade & Task Stream) with Generated Art & Animations (ROMA-LUDUS-RIVER-01..04) [PARALLEL-SAFE WAVE 2B]

Section titled “Phase 348: Plate 2 Romanesque Reskin of GamifyView (Arena & Quests) & HarnessRiverPanel (Pipeline Cascade & Task Stream) with Generated Art & Animations (ROMA-LUDUS-RIVER-01..04) [PARALLEL-SAFE WAVE 2B]”

Goal: Reskin GamifyView.tsx, LudusHud.tsx, LeaderboardArenaPanel.tsx, and HarnessRiverPanel.tsx (WaveformTierStrip.tsx, riverCanvasRenderer.ts, VerificationFunnelFooter.tsx) into Plate 2’s carved marble, bronze medallion, brass gauge, and crimson wax-seal aesthetic with 100% plain-English terminology (Workspace Map, Arena & Quests, Engineering Leaderboard, Fast Local Tier, Balanced Hybrid Tier, Deep Architecture Tier, Live Task Stream, Verification Funnel: Generated → Compiled → Tested → Gated), generated art backgrounds, smooth animations, and progressive-disclosure tooltips on every control. Depends on: Phase 346 Requirements: ROMA-LUDUS-RIVER-01, ROMA-LUDUS-RIVER-02, ROMA-LUDUS-RIVER-03, ROMA-LUDUS-RIVER-04 Success Criteria (what must be TRUE):

  1. GamifyView.tsx, LudusHud.tsx, and LeaderboardArenaPanel.tsx render Plate 2 travertine tablets, laurel wreaths, crimson wax-seal verification badges, and plain-English tabs/controls with interactive tooltips.
  2. HarnessRiverPanel.tsx replaces sci-fi argon neon with Plate 2 brass pressure gauges, animated golden/amber/emerald laminar task streams through a bronze astrolabe interceptor gate, and stepped travertine verification basins.

Phase 349: Interactive Status Bar Popover Dropdowns, Live Repo / Local Version Tile & Configure ▾ Persistence (ROMA-STATUSBAR-01..04) [PARALLEL-SAFE WAVE 2C]

Section titled “Phase 349: Interactive Status Bar Popover Dropdowns, Live Repo / Local Version Tile & Configure ▾ Persistence (ROMA-STATUSBAR-01..04) [PARALLEL-SAFE WAVE 2C]”

Goal: Upgrade BottomStatusBar.tsx and useHudTiles.ts to Plate 2’s burnished brass & travertine style, add the live Repo / Version (REPO vox@main · 98ea5d2 · v0.6.0) tile backed by get_vcs_status / get_app_version, and give every status bar tile (Repo, Engine, Spend, Mesh, Routing, Alerts, Research) both a rich interactive popover dropdown (with plain-English progressive-disclosure explanations and live controls) and one-click navigation to its surface, all configurable in Configure ▾. Depends on: Phase 346 Requirements: ROMA-STATUSBAR-01, ROMA-STATUSBAR-02, ROMA-STATUSBAR-03, ROMA-STATUSBAR-04 Success Criteria (what must be TRUE):

  1. BottomStatusBar.tsx displays REPO vox@<branch> · <short_sha> · v<version> and opens a rich Repository popover (Branch, Commit, Working Tree Status, [Open Git Surface ↗], [Refresh Status]).
  2. Every status bar tile (Repo, Engine, Spend, Mesh, Routing, Alerts, Research) has both an interactive popover dropdown and a direct [Open <Surface> ↗] navigation button, plus progressive-disclosure tooltips and Configure ▾ persistence.

Phase 350: Progressive Disclosure, Button Clustering & Multi-Resolution (1280×800, 1440×900, 1920×1080) Overflow Elimination Across All 32 Surfaces (ROMA-DISCLOSURE-01..04) [WAVE 3]

Section titled “Phase 350: Progressive Disclosure, Button Clustering & Multi-Resolution (1280×800, 1440×900, 1920×1080) Overflow Elimination Across All 32 Surfaces (ROMA-DISCLOSURE-01..04) [WAVE 3]”

Goal: Audit and refine all 32 registered surfaces (Dashboard, Chat, MENS Training, Workspace Map, Harness River, Operations Hub, Tasks, Mesh, Sub-Agents, Model Router, Cost Analytics, Scientia, Repository, Settings, etc.) to apply Plate 2 button theming, progressive disclosure ([Details ▾] drawers and Romanesque tooltips), and responsive overflow fixes at 1280×800, 1440×900, and 1920×1080. Depends on: Phases 346–349 Requirements: ROMA-DISCLOSURE-01, ROMA-DISCLOSURE-02, ROMA-DISCLOSURE-03, ROMA-DISCLOSURE-04 Success Criteria (what must be TRUE):

  1. All 32 surfaces use plain-English labels, Plate 2 Romanesque button/card tokens, and progressive-disclosure tooltips/drawers without crowded toolbars or text truncation.

Phase 351: Multi-Pass 32-Surface × 3-Resolution Before/After Playwright CV Flywheel & Live Vox Axis Verification (ROMA-CV-FLYWHEEL-01..04) [WAVE 4]

Section titled “Phase 351: Multi-Pass 32-Surface × 3-Resolution Before/After Playwright CV Flywheel & Live Vox Axis Verification (ROMA-CV-FLYWHEEL-01..04) [WAVE 4]”

Goal: Run multi-pass Playwright CV Flywheel verification across all 32 surfaces and 3 resolutions (1280×800, 1440×900, 1920×1080) against Concept Art Plate 2 until zero mechanical or accessibility defects remain, capture before/after visual comparisons, and rebuild/relaunch the live vox-gui desktop binary. Depends on: Phase 350 Requirements: ROMA-CV-FLYWHEEL-01, ROMA-CV-FLYWHEEL-02, ROMA-CV-FLYWHEEL-03, ROMA-CV-FLYWHEEL-04 Success Criteria (what must be TRUE):

  1. Multi-resolution Playwright CV Flywheel passes with overlappingControlPairs === 0, clippedTextCount === 0, horizontalOverflowPx === 0, and axeViolations === 0.
  2. Live vox-gui binary is rebuilt and running with all v5.8 Plate 2 improvements.
PhaseRequirementsStatus
346 Top-Left Collision Fix, Romanesque Tokens & Engraved SVG SemioticsROMA-CHROME-01..04Ready
347 Dashboard Embedded Minimap, Frieze Drawers & 7-Regio Forma Urbis RadarROMA-DASH-URBS-01..04Ready
348 Romanesque Reskin of GamifyView & AQVAEDUCTUS Harness RiverROMA-LUDUS-RIVER-01..04Ready
349 Interactive Status Bar Popover Dropdowns & Repo / Local Version TileROMA-STATUSBAR-01..04Ready
350 Progressive Disclosure, Button Clustering & 32-Surface Overflow EliminationROMA-DISCLOSURE-01..04Ready
351 Multi-Pass 32-Surface × 3-Resolution Before/After CV FlywheelROMA-CV-FLYWHEEL-01..04Ready

Roadmap: v5.7 — Vox ↔ Python Bridge & Tolerant CLI Ergonomics

Section titled “Roadmap: v5.7 — Vox ↔ Python Bridge & Tolerant CLI Ergonomics”

Milestone: v5.7 Status: Phase 324 verified 2026-11-03 (5/5). phase.complete did not write this file: workstream mode has no active workstream, and v5.7 lives on this root roadmap. Updated: 2026-11-01 Design SSOT: docs/src/architecture/vox-python-interop-design-2026-11-01.md; docs/src/architecture/cli-semantic-tolerance-research-2026.md Paper-code rule (all phases): a phase exits only on an end-to-end vox run golden through the real worker (or, for compiler-only phases, a compiler golden through the real pipeline) — never on unit tests or the fake worker alone. Feature-matrix cells flip to Implemented only with a running ladder entry.

Phase 324: Effect Gate & Determinism (EFF-01..03)

Section titled “Phase 324: Effect Gate & Determinism (EFF-01..03)”

Goal: The effect and capability gate that every later Python call depends on exists and cannot be silently bypassed. Depends on: Nothing (first phase of v5.7) Requirements: EFF-01, EFF-02, EFF-03 Success Criteria (what must be TRUE):

  1. @uses(nonexistent) and uses nonexistent both produce a compile error; @uses(mcp(tool)) appears in HIR as Mcp(tool) (golden).
  2. @uses(python_unverified) survives into HIR and HirCapability; a materializer-shaped call without the effect fails typecheck (golden).
  3. vox run --caps without py refuses a Python-capability call at runtime with a caps error.
  4. A workflow calling a Python-capability fn is rejected by capability (not name); the same call in an activity compiles.

Plans: 2/2 complete (324-01, 324-02). Verification passed 2026-11-03.

Phase 325: Bridge Runtime & Environment (BR-01..04, GOV-01)

Section titled “Phase 325: Bridge Runtime & Environment (BR-01..04, GOV-01)”

Goal: vox-python-bridge spawns a real uv-managed worker, speaks framed Arrow IPC, survives crashes/timeouts, and lands with governance rows. Depends on: Phase 324 Requirements: BR-01, BR-02, BR-03, BR-04, GOV-01 Success Criteria (what must be TRUE):

  1. vox py sync on a [python] project creates the env and records the uv.lock hash in vox.lock; vox py doctor reports status; a project without [python] gets py-env: and spawns no process.
  2. A one-op golden runs end-to-end under vox run through the real worker; a worker op that prints to stdout still returns a correct result.
  3. A deliberately hung op hits timeout_s, the worker restarts, and a pre-restart handle returns py: stale handle; the host process survives a worker crash.
  4. An over-cap fetch returns py: fetch exceeds cap; sandboxed spawn goes through vox-skill-runtime, unsandboxed warns.
  5. vox-arch-check and vox ci crate-edges pass with the new crate; AGENTS.md exemption and stale-citation fixes are in the diff.

Plans: 2/2 complete (325-01, 325-02)

Phase 326: std.array on numpy (ARR-01..04, K-01)

Section titled “Phase 326: std.array on numpy (ARR-01..04, K-01)”

Goal: Vox code computes with numpy through a closed, typed, standards-named vocabulary and pure lazy plans. Depends on: Phase 325 Requirements: ARR-01, ARR-02, ARR-03, ARR-04, K-01 Success Criteria (what must be TRUE):

  1. examples/golden/py_array_stats.vox runs under vox run with asserted output, using one round-trip per materialization.
  2. Mixed int/float without .astype, std without correction=, and sort without nulls= are compile errors.
  3. NaN, ±Inf and i64::MIN/MAX round-trip through to_list; uint64 overflow errors.
  4. vox ci python-bridge-parity fails when an allowlisted op is missing its signature, handler or conformance test (negative test).
  5. Ladder entry + complexity-budget row exist for the golden; LEXER_KEYWORDS unchanged; K2 ceiling enforced.

Plans: 8/8 executed. Verified 2026-11-04 (8/8 must-haves). Code review left 3 critical findings open.

Phase 327: std.frame on polars (FRM-01..03)

Section titled “Phase 327: std.frame on polars (FRM-01..03)”

Goal: Vox code runs lazy dataframe pipelines on polars and fetches typed Records with exact type mapping and host-side path control. Depends on: Phase 326 Requirements: FRM-01, FRM-02, FRM-03 Success Criteria (what must be TRUE):

  1. examples/golden/py_frame_report.vox (scan_csv → filter → group_by/agg → sort → collect) passes with one round-trip per collect.
  2. A fixture with decimal, tz-aware timestamp, date, duration and nullable columns fetches as Decimal, UTC DateTime, Date, Duration, Option[T]; a naive timestamp errors with the fix-it.
  3. scan_csv on a path outside --caps fs:ro= is refused before any worker request (asserted via fake-worker request log).

Plans: TBD

Phase 328: std.plot & std.tensor (PLT-01, TNS-01..02)

Section titled “Phase 328: std.plot & std.tensor (PLT-01, TNS-01..02)”

Goal: Vox code renders matplotlib charts and runs exported PyTorch models without data or tensors leaving the worker unnecessarily. Depends on: Phase 327 Requirements: PLT-01, TNS-01, TNS-02 Success Criteria (what must be TRUE):

  1. A plot golden writes a non-empty PNG with expected dimensions; the request log shows no frame fetch to host.
  2. examples/golden/py_tensor_infer.vox loads a checked-in tiny .pt2 and returns expected logits on CPU; a pickle .pt is refused.
  3. Two seeded runs produce identical numpy/torch outputs; the activity journal entry contains library versions and lock hash.

Plans: TBD

Phase 329: Tier 2 Escape Hatch (PYX-01..04)

Section titled “Phase 329: Tier 2 Escape Hatch (PYX-01..04)”

Goal: Vox code reaches the long tail (scipy, scikit-learn, pillow, pandas, any bound package) through a typed, gated, counted escape hatch. Depends on: Phase 328 Requirements: PYX-01, PYX-02, PYX-03, PYX-04 Success Criteria (what must be TRUE):

  1. examples/golden/py_escape_psd.vox (import py sig from "scipy.signal") passes; an unlisted name, a stale sidecar, and PyObject.field are compile errors.
  2. Curated sidecars + goldens for scipy, scikit-learn, pillow and pandas pass in the opt-in Python CI job.
  3. Each Tier 2 call site yields vox/python/unverified-call and a python.escape_call event; syntax-K support metrics show py_tier1_calls / py_tier2_calls.
  4. Attempting to bind pickle or subprocess fails with a Denied-policy error.

Plans: TBD

Phase 330: DB & Backend Interop (DBI-01..03)

Section titled “Phase 330: DB & Backend Interop (DBI-01..03)”

Goal: Vox apps share databases owned by Drizzle/SQLAlchemy/Alembic and serve Python clients, using existing introspection and OpenAPI surfaces. Depends on: Phase 324 (independent of 325–329; may run in parallel) Requirements: DBI-01, DBI-02, DBI-03 Success Criteria (what must be TRUE):

  1. vox db introspect --emit vox on a sqlite fixture emits table decls that vox db verify accepts against the same DB (round-trip test; postgres/mysql behind existing features).
  2. With [db] schema_owner = "external", vox db migrate refuses to apply and --dry-run prints the plan.
  3. The how-to’s Python client example is generated from a golden server app’s OpenAPI and checked in CI docs lint.

Plans: 1/1 complete (330-01)

Phase 331: Tolerant CLI Ergonomics & Precomputed Semantic Routing (CLI-TOL-01..04)

Section titled “Phase 331: Tolerant CLI Ergonomics & Precomputed Semantic Routing (CLI-TOL-01..04)”

Goal: The vox CLI tolerates human memory gaps (unambiguous prefix completions, global flag hoisting, verb↔noun token permutations, unique leaf elision, and precomputed compositional synonyms) in <1.5 ms with zero runtime LLM calls, while enforcing side_effect_class safety gates and returning structured 1-turn self-repair redirects for LLM agents and non-TTY callers. Depends on: Independent of Phases 324–330 (may run in parallel) Requirements: CLI-TOL-01, CLI-TOL-02, CLI-TOL-03, CLI-TOL-04 Success Criteria (what must be TRUE):

  1. VoxCliRoot enables infer_subcommands(true) and infer_long_args(true), and all GlobalOpts flags (--json, --quiet, --verbose, --color) carry global = true so vox sec doc and vox secrets list --json parse natively at 0 ms overhead.
  2. When VoxCliRoot::try_parse_from fails, argv_normalize resolves order-agnostic verb↔noun permutations (vox list secrets → vox secrets list), unique leaf elision (vox command-compliance → vox ci command-compliance), and wrong-subcommand flags (vox build --release → suggests vox compile --release) in <1.5 ms over CommandCatalog and CapabilityIndex.
  3. Execution is gated by side_effect_class and TTY/agent detection: read_only permutations/synonyms auto-execute on interactive TTYs with a dim stderr canonical echo, mutating commands prompt [Y/n] on TTYs, and non-TTY/CI/LLM callers (--json, CI=true, VOX_MCP=1) never hang on prompts or silently auto-execute fuzzy guesses—returning exit code 2 with structured aci.error_recovery canonical redirects.
  4. vox ci cli-synonyms --write compiles compositional verb_aliases, noun_aliases, and foreign_idioms from contracts/operations/catalog.v1.yaml (using vox_actor_runtime::llm + gemini-3.5-flash-lite structured output and gemini-embedding-2 / local_embed_text cosine margin $\ge 0.08$ verification) into contracts/cli/semantic-aliases.v1.json, and vox ci command-compliance fails on any alias collision or antonym banlist violation (add $\leftrightarrow$ remove, start $\leftrightarrow$ stop, lock $\leftrightarrow$ unlock, set $\leftrightarrow$ get).

Plans: 3 plans

Plans:

  • 331-01-PLAN.md — Prefix unfolding (infer_subcommands/infer_long_args), global flag hoisting, and <1.5 ms deterministic argv_normalize::classify_recovery (CLI-TOL-01, CLI-TOL-02)
  • 331-02-PLAN.md — Dual-persona (InteractiveHuman vs NonInteractiveOrAgent) & side_effect_class safety gate + structured aci.error_recovery diagnostics (CLI-TOL-03)
  • 331-03-PLAN.md — Precomputed semantic synonym compiler (vox ci cli-synonyms), contracts/cli/semantic-aliases.v1.json lockfile, and antonym/collision firewall (CLI-TOL-04)

Phase 333: Gated training for every model size

Section titled “Phase 333: Gated training for every model size”

Goal: Every published Qwen3 size finishes the same release review, then a merged complete model is stored and distributed, and only then does its Hugging Face card call it ready. 0.6B leaves the smoke recipe. 4B is already a full shared-recipe run. Cards stay in plain language and describe the trained add-on until the complete model is the file being distributed. Depends on: Independent of Phases 324–331 Requirements: REL-01, REL-02, REL-03, REL-04 Success Criteria (what must be TRUE):

  1. 0.6B, 1.7B, 4B, and 8B each have a published run on the shared recipe (rank 16, alpha 32, gradient accumulation 8). The 0.6B smoke preset (qwen3_dev_cpu, rank 8, accumulation 1) is not the published 0.6B adapter. The published 4B run stays classified as a full shared-recipe run that was uploaded before the checklist passed. A resumed run reports token and step counts for the whole run, not only the last process.
  2. Each of those four sizes has a passing collateral check and a release receipt with overall_passed: true. 14B and 32B stay without a distributed Vox model until a run clears that same review.
  3. After that receipt, vox mens merge-qlora writes a complete checkpoint from the local Qwen base plus the trained adapter, and that checkpoint is the file stored and distributed. The local cache already holds original Qwen bases for 0.6B, 1.7B, 4B, 8B, and 14B. 32B is not downloaded. 14B has no trained adapter yet.
  4. The Hugging Face card and the size index change in the same step as that distribution. Until then they say the page is the trained add-on and the review is unfinished. They do not use the word “ungated”. A reader can download the distributed file from the Files tab or with one vox mens hub download command.

Plans: 0 plans

Plans:

  • TBD (run /gsd-plan-phase 333 to break down)
PhaseRequirementsStatus
324 Effect Gate & DeterminismEFF-01..03Verified 2026-11-04
325 Bridge Runtime & EnvironmentBR-01..04, GOV-01Verified 2026-11-04
326 std.array on numpyARR-01..04, K-01Verified 2026-11-04
327 std.frame on polarsFRM-01..03Not started
328 std.plot & std.tensorPLT-01, TNS-01..02Not started
329 Tier 2 Escape HatchPYX-01..04Not started
330 DB & Backend InteropDBI-01..03Verified 2026-11-04
331 Tolerant CLI Ergonomics & Precomputed Semantic RoutingCLI-TOL-01..04Planned (3 plans)
333 Gated training for every model sizeREL-01..04Not planned

Phase 342: Docs Authoring Parity and Starlight 0.42 Upgrade

Section titled “Phase 342: Docs Authoring Parity and Starlight 0.42 Upgrade”

Directory: phases/342-docs-authoring-parity-and-starlight-0-42-upgrade/ Goal: Close the verified remainder of the AsciiDoc-parity blueprint for voxlang.org: upgrade docs-astro to Starlight ≥0.42 / Astro ≥7.2.10 (the gate for the ecosystem plugins), render the 24 GitHub-alert docs as real asides, and add build-time internal-link validation — without new contracts, new .mdx churn, or a parallel status system. Requirements: DOCS-PARITY-01..04 Depends on: none (independent of v5.6 lanes; coordinate banners with docs-freshness Phase 32) SSOT: docs/src/architecture/docs-site-asciidoc-parity-findings-2026.md Plans: 4/4 plans complete — ✅ verified 2026-11-06 (7/7, 342-VERIFICATION.md, 09c272e99). Resolved stack: astro 7.4.5, starlight 0.42.9 (mature-only install, no release-age exceptions). Playwright 20 / unit 3 failures are the frozen pre-upgrade baseline, unchanged.

  • DOCS-PARITY-01 (✅ committed 9829418e9): remark-vox-include.mjs supports tag::x[]/end::x[], # leaders, comma multi-select, N..M line ranges, exact names, dedent; unit tests in docs-astro/tests/unit/remark-vox-include.test.mjs; llms.txt version derived from root Cargo.toml + namespace disambiguation. Build green (906 pages).
  • DOCS-PARITY-02: Starlight ^0.38 → ≥0.42, Astro ^6 → ≥7.2.10; fix overrides (Banner, Sidebar, PageSidebar, TableOfContents, SiteTitle), routeData.ts, and the @astrojs/starlight/components Vite alias shims; pnpm build + tests/unit + Playwright reader-layout/trust-signals stay green.
  • DOCS-PARITY-03: starlight-github-alerts renders the 24 > [!NOTE|TIP|IMPORTANT|WARNING|CAUTION] docs as asides (identical on GitHub and the site).
  • DOCS-PARITY-04: starlight-links-validator in the build; fix the two current remark-doc-links failures (t3-vox-wave4-gap-closure-and-live-pipeline-audit-2026.md out-of-root link, missing api/vox-cargo-shim.generated.md from l5-binaries.generated.md); feeds nightly Docs Link Checker (#578).

Out of scope (rejected after verification — see SSOT): reality-audit-driven banners (inventory has no status field; banners already ledger-driven, owned by docs-freshness Phase 32), stability-matrix.v1.json + Content Layer collections (invented contract; generated-.md pattern is canonical), remark-vox-attributes tokens / ::include{} directives / // !hide / examples/golden/anti/ titles (zero call sites), starlight-markdown-blocks custom asides.

Plans:

  • 342-01-PLAN.md — Fix both remark-doc-links failures at source (t3 page links; hub generator links only existing API pages, test-first) and freeze the pre-upgrade test baseline (wave 1)
  • 342-02-PLAN.md — Package-legitimacy gate, then astro 7 / Starlight 0.42 upgrade (unified() processor, alias via import.meta.resolve, llms-txt 0.12 patch), dependabot unpin, agent-safe Playwright (wave 2)
  • 342-03-PLAN.md — starlight-github-alerts renders every alert as an aside; un-nest the two list-nested alerts for GitHub parity (wave 3)
  • 342-04-PLAN.md — starlight-links-validator on default fatal options with a unit-tested exclude predicate, mutation probe, findings SSOT update (wave 4)

Phase 343: Docs System SSOT and Automated Drift Control

Section titled “Phase 343: Docs System SSOT and Automated Drift Control”

Goal: Every kind of documentation divergence is either prevented (generated from a contract), detected (a gate that fails), or auto-corrected (regenerated by the ssot-drift / ssot-autoregen loop). No documented gap stays manual and nobody notices. The docs-system map docs/src/architecture/docs-system-ssot.md (landed 2026-11-06) is the SSOT; its §6 “Known gaps” list is this phase’s backlog. Each gap closed removes its bullet in the same PR. Requirements: DOCS-DRIFT-01..06 Depends on: Phase 342 SSOT: docs/src/architecture/docs-system-ssot.md Plans: 9 plans

  • DOCS-DRIFT-01 (prevent, regen-list parity): derive scripts/ssot-regen.vox’s regenerator list from run_ssot_drift’s stage table (or add a parity check that fails on mismatch). Removes the “keep in sync” hand list.
  • DOCS-DRIFT-02 (detect, overlay parity): a vox ci check that extracts normative claims (forbidden/required frontmatter keys, retired symbols, banned commands) from AGENTS.md and fails when .cursor/rules/*.mdc, GEMINI.md, CLAUDE.md or .github/copilot-instructions.md contradict them. First fix: the last_updated contradiction in documentation-policy.mdc.
  • DOCS-DRIFT-03 (detect, prose vs code): scan docs and agent files for vox <cmd> [--flag] mentions and “planned per …” claims. Fail when a mentioned command or flag is absent from contracts/cli/command-registry.yaml, or when a “planned” surface already exists there. First fixes: AGENTS.md vox audit retirement / vox audit effort.
  • DOCS-DRIFT-04 (detect, dead guards and orphans): make guards that silently no-op when their input file is missing fail loudly (check_docs_ssot orphan inventory). Add an orphan-page report: a docs/src page no other page or nav entry links to, with zero reader-facing orphans (no baseline).
  • DOCS-DRIFT-05 (auto-correct, scheduled loops): a scheduled job runs vox audit docs --update-ledger and the external link checker. It opens one draft PR with ledger refreshes and proposed .lycheeignore / link fixes, instead of a nightly-failure issue (#578). Decide whether to revive or retire the dormant docs-reality-audit program and the self-heal bot’s draft-PR stage.
  • DOCS-DRIFT-06 (meta-gate): a test asserts that every gate listed in docs-system-ssot §4 is wired where the table says (pre-push tier / docs-quality.yml / nightly). It also asserts §6 lists only gaps with no matching gate. The map can then never silently disagree with the machinery.

Plans: Wave 1

  • 343-01-PLAN.md — ssot-drift stage table + ssot_regen_parity (ssot-regen.vox and lefthook retry list checked against it); decision on the lefthook retry list (wave 1)

Wave 2 (blocked on Wave 1 completion)

  • 343-02-PLAN.md — overlay parity: AGENTS.md claims YAML checked against CLAUDE/GEMINI/copilot/.cursor overlays; last_updated and other contradictions fixed; decision on the build-environment cargo path (wave 2)

Wave 3 (blocked on Wave 2 completion)

  • 343-03-PLAN.md — CLI-claims stage (clap tree + registry): agent files gated, reader-facing pages measured; decision fix-vs-baseline for reader-facing defects (wave 3)

Wave 4 (blocked on Wave 3 completion)

  • 343-04-PLAN.md — fix every reader-facing CLI-claim defect and enforce the reader-facing scope (wave 4)

Wave 5 (blocked on Wave 4 completion)

  • 343-05-PLAN.md — dead guards removed or made fatal (check_docs_ssot, check-links, retired-symbol overlays); decision on expired link-allowlist entries (wave 5)

Wave 6 (blocked on Wave 5 completion)

  • 343-06-PLAN.md — corrected inbound-link graph + docs_orphan_pages stage; decision on orphan linking vs shrink-only baseline (wave 6)

Wave 7 (blocked on Wave 6 completion)

  • 343-07-PLAN.md — idempotent ledger refresh + Docs Drift Refresh draft-PR workflow; decisions on .lycheeignore scope and reality-audit/self-heal (wave 7)

Wave 8 (blocked on Wave 7 completion)

  • 343-08-PLAN.md — lychee triage script into the bot PR, reality-audit/self-heal decision applied, docs describe what runs (wave 8)

Wave 9 (blocked on Wave 8 completion)

  • 343-09-PLAN.md — docs_system_parity meta-gate: §4 wiring proven against the machinery, §6 lists only open gaps (wave 9)

Phase 345: Kiln Synthetic Topic Trees & Vox Axis MENS Studio (MENS-STUDIO-01..04)

Section titled “Phase 345: Kiln Synthetic Topic Trees & Vox Axis MENS Studio (MENS-STUDIO-01..04)”

Directory: .planning/phases/345-kiln-inspired-synthetic-topic-trees-vox-axis-mens-all-hubs-a/ Goal: Port Kiln AI’s high-leverage synthetic dataset mechanisms natively into Rust (SpokeTopicTreeV1 recursive topic-tree expansion, Teacher→Student CoT distillation, and automatic Oracle-Fail→Teacher-Repair→DPO pair synthesis in vox-corpus without Python kiln_ai dependencies) and upgrade Vox Axis’s MensTrainingView into the Option A “Ladder × Spokes Crucible Matrix” MENS Training Studio with real Rust Tauri backend commands (mens_ladder_status, mens_save_custom_spoke, mens_save_crucible_config, mens_live_telemetry, explain_routing), Clavis HF_TOKEN auto-sync (1 auto 2 confirm), live Orchestrator→Spoke routing visualization in Chat and MENS, and multi-resolution CV Before/After Flywheel verification (1280×800, 1440×900, 1920×1080). Requirements: MENS-STUDIO-01, MENS-STUDIO-02, MENS-STUDIO-03, MENS-STUDIO-04 Depends on: Independent of Phases 342–344 SSOT: docs/src/architecture/kiln-synthetic-data-and-mens-axis-gui-research-2026.md UI-SPEC: .planning/phases/345-kiln-inspired-synthetic-topic-trees-vox-axis-mens-all-hubs-a/345-UI-SPEC.md Plans: 2 plans

Plans:

  • 345-01: Option A MENS Crucible Studio UI, Clavis HF_TOKEN Settings deep-link, Chat Orchestrator→Spoke badge, and 8-pass multi-resolution Playwright CV Flywheel (visualAfterScore: 0.9352, axeViolations: 0, overlappingControlPairs: 0) (completed 2026-11-08)
  • 345-02: Full-stack Rust Tauri backend plumbing (mens_ladder_status reading mens/config/gpu-specs.yaml & mens/runs/size_train/*.json, mens_save_custom_spoke, mens_save_crucible_config, mens_live_telemetry JSONL time-series spline hydration, explain_routing daemon integration, and CLI flag parity for train/eval-local/eval-gate/serve/hub) (completed 2026-11-08)

Roadmap: v5.6 — Reasoning Where the Lane Needs It

Section titled “Roadmap: v5.6 — Reasoning Where the Lane Needs It”

Milestone: v5.6 Status: Phases 313–323 complete (25/25 requirements implemented and verified). Updated: 2026-11-01

Goal: Codegen requests and codegen training rows use the non-thinking prefix, and hub and agent requests use the bare thinking-mode prefix. Depends on: Nothing (first phase of v5.6) Requirements: LANE-01, LANE-02 Success Criteria (what must be TRUE):

  1. A codegen serve prompt contains /no_think and ends with the empty think block.
  2. A hub or agent serve prompt ends at a bare assistant turn and does not contain /no_think.
  3. A codegen direct-answer training row keeps the empty think block outside the loss, and a row that already has a non-empty think body is not rewritten with /no_think.

Plans: 4 plans

Plans:

Wave 1

  • 313-01: Tracer: raw /generate carries lane through to the plugin payload, and an unknown lane is a 400 before any worker call (completed 2026-10-26)

Wave 2 (blocked on Wave 1 completion)

  • 313-02: CUDA and Metal assemble_prompt wrap raw prompts by lane through one shared builder (completed 2026-10-26)
  • 313-03: Chat completions accept lane and flatten the last user turn plus the assistant opener (completed 2026-10-26)

Wave 3 (blocked on Wave 2 completion)

  • 313-04: Lock the existing codegen training fusion with tests only (completed 2026-10-26)

Phase 314: Held-out Reasoning Set (BENCH-02, BENCH-03)

Section titled “Phase 314: Held-out Reasoning Set (BENCH-02, BENCH-03)”

Goal: The same GSM8K, MATH-500, and Vox items exist for a before/after, and a missing model cannot write a passing score. Depends on: Phase 313 Requirements: BENCH-02, BENCH-03 Success Criteria (what must be TRUE):

  1. A checked-in Vox slice has a checkable answer on every item and does not reuse the fixed Plan / Syntax / Implementation template.
  2. The GSM8K and MATH-500 slices used for the comparison are the checked-in files, not a new sample.
  3. A reasoning eval that cannot load the model writes no passing score and records how many items were scored.

Plans: 2 plans

Plans:

  • 314-01 (Wave 1): Check in the gsm8k_think, math500_think, and vox_think 32-item slices with a pinned loader and an interpreter oracle for Vox golds (completed 2026-10-26)
  • 314-02 (Wave 2, after 314-01): Fail closed when the model does not load or any inference error occurs; write not_scored with n: 0, and commit the scorer (completed 2026-10-26)

Goal: Serve refuses an adapter whose collateral report is missing or worse on accuracy or think-initiation, and the report labels the benches as slices. Depends on: Phase 314 Requirements: GATE-01, GATE-02 Success Criteria (what must be TRUE):

  1. Serve refuses to bind an adapter when the collateral report is missing.
  2. Serve refuses when accuracy or think-initiation on the reasoning keys drops past the configured threshold.
  3. The report states that the benches are slices and records the item count, so the 5% bar is not read as a full GSM8K or MATH-500 result.

Plans: 2 plans

Plans:

  • 315-01 (Wave 1): Strict per-slice evaluator in vox-eval; eval-collateral-damage requires all six reasoning keys, removes a stale report, and labels the slices with n (completed 2026-10-26)
  • 315-02 (Wave 2, after 315-01): Serve re-validates the on-disk report and the per-slice reasoning reports and refuses a missing, stale, edited, or unscored one (completed 2026-10-26)

Phase 316: Before/After Comparison (BENCH-01)

Section titled “Phase 316: Before/After Comparison (BENCH-01)”

Goal: A pinned Qwen3 base of any ladder size and its adapter are scored on the same slices when the weights exist, and no score is invented when they do not. Depends on: Phase 315 Requirements: BENCH-01 Success Criteria (what must be TRUE):

  1. When the requested pinned Qwen3 snapshot (default Qwen/Qwen3-8B) and an adapter trained on it are on disk, one report holds accuracy and think-initiation for the GSM8K slice, the MATH-500 slice, and the Vox slice on both sides.
  2. When the snapshot is absent, the run records that fact and does not write numeric scores.
  3. Both sides use the checked-in items.

Plans: 1/1 plans executed

Plans:

  • 316-01-PLAN.md
  • 316-01: Run the before/after on the checked-in slices, or record that the snapshot is absent (completed 2026-10-28)

Goal: Every Qwen3 dense size is a pinned, licensed rung that the trainer and the harness can name. Depends on: Phase 316 Requirements: SIZE-01 Success Criteria (what must be TRUE):

  1. The ladder has Qwen3-0.6B, 1.7B, 4B, 8B, 14B, and 32B, each pinned to a real Hugging Face commit and carrying a license.
  2. A rung whose memory floor has not been measured on hardware says so.

Plans: 1/1 plans executed

Plans:

  • 317-01-PLAN.md — Pin 1.7B and 4B on the qwen3_code ladder and name every dense size from the overlay

Goal: One selector reads current free memory, titrates to the largest Qwen3 size that fits, and picks the hub and its spokes together. Training, local serve, and chat all use it. Depends on: Phase 317 Requirements: SIZE-02 Success Criteria (what must be TRUE):

  1. The selector reads current free memory (Metal live budget, CUDA free VRAM), not total memory, and picks the largest ladder size whose predicted footprint fits.
  2. If free memory drops between the pick and the load, it steps down a size and says why, instead of loading and running out of memory.
  3. Every spoke’s base and preset follow the picked size. An explicit size overrides the pick, and a size that does not fit is refused rather than silently shrunk.
  4. The inference auto-selector and the training base pick both use this selector. The old fixed tiers that point at run directories that do not exist are gone.

Plans: 3/3 plans executed

Plans:

  • 318-01-PLAN.md — Measured QLoRA picker, pre-load step-down, and free-memory adapter (completed 2026-10-29)
  • 318-02-PLAN.md — One size for every spoke, and a re-read immediately before training load (completed 2026-10-29)
  • 318-03-PLAN.md — Chat and the GUI host call the selector; directory tiers are removed (completed 2026-10-29)

Goal: Across a mesh, the selector picks the largest size that fits on any node’s current free memory and routes the work there. Depends on: Phase 318 Requirements: MESH-01 Success Criteria (what must be TRUE):

  1. Node heartbeats carry current free accelerator memory, not only total VRAM.
  2. The mesh pick chooses the largest size that fits on some node, routes the job to that node, and titrates down when no node fits the larger size.
  3. A stale or missing node reading is not treated as free memory, and the local fallback still works with no mesh.

Plans: 3/3 plans executed

Plans:

  • 319-01-PLAN.md — Largest measured QLoRA floor, routed to a node that holds it (completed 2026-10-29)
  • 319-02-PLAN.md — Stale window, explicit-size refusal, and heartbeat free memory (completed 2026-10-29)
  • 319-03-PLAN.md — Local fallback when no node has a fresh reading (completed 2026-10-29)

Goal: Each Qwen3 size that fits this Mac is trained with the lane prefix, and each size that does not is recorded as not trained. Depends on: Phase 319 Requirements: SIZE-03 Success Criteria (what must be TRUE):

  1. Each size that fits produces a Metal QLoRA adapter whose manifest names its pinned base.
  2. A size that does not fit has a record with the measured memory and no adapter.

Plans: 2/2 plans executed

Plans:

  • 320-01-PLAN.md — One size_train.v1 record per qwen3_code QLoRA size, from an injected free-memory reading (completed 2026-11-01)
  • 320-02-PLAN.md — Register vox mens size-train and take that reading once (completed 2026-11-01)

Phase 321: Harness Before/After (HARN-01, BENCH-04)

Section titled “Phase 321: Harness Before/After (HARN-01, BENCH-04)”

Goal: The harness serves a chosen size, and one report shows before and after reasoning and intelligence scores for every trained size. Depends on: Phase 320 Requirements: HARN-01, BENCH-04 Success Criteria (what must be TRUE):

  1. The harness serves a chosen size’s base or adapter through the existing serve path, behind the collateral gate.
  2. For every trained size, the report holds base and adapter reasoning accuracy and think-initiation on the three slices, plus an intelligence score on a codegen or agent bench, with n recorded.
  3. At least two sizes have real trained scores. Sizes without them are listed as not trained, with no numbers.

Plans: 1/1 plans executed

Plans:

  • 321-01: Chosen-size serve resolution behind collateral gate (HARN-01) and multi-size harness_before_after.v1 reasoning + intelligence report (BENCH-04) (completed 2026-11-01)
PhaseNamePlansStatus
Phase 313Lane Prefix4/4Complete
Phase 314Held-out Reasoning Set2/2Complete
Phase 315Serve Gate2/2Complete
Phase 316Before/After Comparison1/1Complete 2026-10-28
Phase 317Size Ladder1/1Complete 2026-10-28
Phase 318Live VRAM Selector3/3Complete 2026-10-29
Phase 319Mesh-Wide Selection3/3Complete 2026-10-29
Phase 320Train the Sizes2/2Complete 2026-11-01
Phase 321Harness Before/After1/1Complete 2026-11-01
Phase 322Calibrated turn decisions + ChatARC1/1Complete 2026-11-01
Phase 323High-Value Starlight MDX Conversions + Ingest Parity2/2Complete 2026-11-01

Phase 322: Calibrated turn decisions and dynamic action catalog (DEC-01..04, CAT-01..02)

Section titled “Phase 322: Calibrated turn decisions and dynamic action catalog (DEC-01..04, CAT-01..02)”

Directory: phases/322-calibrated-turn-decisions-and-dynamic-action-catalog/ Goal: Fully plumb System One / RLCD-style calibrated decisions into the live model-route and agent-loop control points (no facade). ChatARC composes the catalogs the harness already discovers (skills, MCP, clap commands) and chooses among shortlists. Complements models::select, clutch/risk axes, SIZE/MESH selectors, and MENS spokes — does not replace them or pretend allow_free_in_performance_mode is PreferFree. Depends on: Nothing in 317–321 (parallel). Must not fork SIZE-02 / MESH-01. Primary insertion is llm_bridge/model_route_policy/resolve.rs, plus repair escalation and stop/post-tool hooks. Requirements: DEC-01, DEC-02, DEC-03, DEC-04, CAT-01, CAT-02 Success Criteria (what must be TRUE):

  1. Turn-start pack feeds build_selection_request without overriding clutch-derived axes or allow_free when a clutch profile is set; flag off matches today’s request bytes-for-bytes on fixtures.
  2. allow_free_in_performance_mode is documented and tested as scorer eligibility only — free_floor slug order stays in bootstrap_inference / FREE_FLOOR.
  3. Post-tool low confidence reuses apply_repair_escalation_to_config; stop/verify stay on evaluate_stop_hook / evaluate_post_tool_batch_hook.
  4. ChatARC shortlist Choice is plumbed into an existing activation path and stays correct when a new skill, MCP tool, or clap command appears; selective-prediction metrics published; hosted Jev behind local-only privacy.

Plans: 1/1 plans executed

Research / prior art: docs/src/architecture/jev-system-one-harness-integration-research-2026.md (§§11–16); cursor plan labeled_router_plus_jev folded here (do not execute as a standalone facade).

Plans:

  • 322-01: Pre-turn TurnDecisionPack (DEC-01), post-tool confidence escalation (DEC-02), flag-off byte-identity tests (DEC-03), selective-prediction & privacy gate (DEC-04), and ChatARC top-K shortlist + calibrated Choice (CAT-01..02) (completed 2026-11-01)

Phase 323: High-Value Starlight MDX Conversions and Ingest Parity (MDX-INGEST-01..03, MDX-CONV-01..03)

Section titled “Phase 323: High-Value Starlight MDX Conversions and Ingest Parity (MDX-INGEST-01..03, MDX-CONV-01..03)”

Directory: phases/323-high-value-starlight-mdx-conversions-and-ingest-parity/ Goal: Make .mdx a first-class, zero-regression citizen across vox-kb-facets, vox-search (Library index, kb_ids, RAG ingest_markdown_tree, chunk_document), vox-doc-pipeline, and MENS training extraction; then convert high-value tutorials, onboarding guides, and flagship language reference pages to Starlight MDX (<Steps>, <Tabs>, <CardGrid>, <VoxPlayground>) while keeping prose-heavy architecture/ADR/research docs in .md. Depends on: Nothing (parallel documentation and search track) Requirements: MDX-INGEST-01, MDX-INGEST-02, MDX-INGEST-03, MDX-CONV-01, MDX-CONV-02, MDX-CONV-03 Success Criteria (what must be TRUE):

  1. vox-kb-facets (inspect_summary, is_summary_marker_line, strip_summary_blocks) accepts both HTML (<!-- kb:summary -->) and MDX JSX ({/* kb:summary */}) markers and ignores top-level ESM import lines between # H1 and the summary block.
  2. vox-search::kb_ids::is_kb_corpus_path, doc_title, and ingest::ingest_markdown_tree index both .md and .mdx under docs/src/; library::chunk::first_paragraph and MENS corpus extraction strip ESM import/export statements and JSX layout wrapper tags (<Tabs>, <TabItem>, <Steps>, <CardGrid>, {/* ... */}) so Tantivy cards and training chunks never leak JSX boilerplate.
  3. vox-doc-pipeline validates .mdx files (frontmatter, facets, vox doctest fences, internal .md/.mdx links) and upgrades docs/src/index.mdx to clean Starlight components (<CardGrid>, <LinkButton>) without inline-style HTML soup.
  4. High-value tutorials (docs/src/tutorials/**) and flagship reference/getting-started pages are converted to .mdx using Starlight <Steps>, <Tabs> (showing Vox source alongside generated SQL/Rust/TSX), and embedded <VoxPlayground>, passing pnpm --dir docs-astro build, vox-doc-pipeline --lint-only, and cargo test -p vox-search -p vox-kb-facets -p vox-doc-pipeline.

Plans: 2/2 plans executed

Plans:

  • 323-01: Dual HTML/MDX summary markers and L0 MDX boilerplate stripping across vox-kb-facets, vox-search, vox-doc-pipeline, and vox-corpus (MDX-INGEST-01..03) (completed 2026-11-01)
  • 323-02: Upgrade docs/src/index.mdx, add .md ↔ .mdx link fallback, and ship interactive Starlight MDX tutorial tut-fullstack-walkthrough.mdx (MDX-CONV-01..03) (completed 2026-11-01)

Roadmap: v5.5 — T3 CI/TOESTUB Deep Enforcement (Phases 307–312) — SHIPPED

Section titled “Roadmap: v5.5 — T3 CI/TOESTUB Deep Enforcement (Phases 307–312) — SHIPPED”

Milestone: v5.5 Status: Planning complete — ready for /gsd-plan-phase 307 Updated: 2026-10-26T01:59:06Z

Phase 307: PR Vouch & Contributor Trust Gate (T3-TRUST-01..04)

Section titled “Phase 307: PR Vouch & Contributor Trust Gate (T3-TRUST-01..04)”

Directory: phases/307-pr-vouch-and-contributor-trust/ Goal: Add .github/VOUCHED.td (or contracts/ci/vouched.v1.yaml) SSOT for trusted/denounced GitHub logins, documented in AGE…

  • 307-01: Implement T3-TRUST requirements

Phase 308: Preview/Privileged Workflow Trust-Split Policy (T3-PREVIEW-01..04)

Section titled “Phase 308: Preview/Privileged Workflow Trust-Split Policy (T3-PREVIEW-01..04)”

Directory: phases/308-preview-trust-split-policy/ Goal: Author docs/src/ci/privileged-preview-trust-split.md capturing t3code relay pattern: deploy may run PR code only under…

  • 308-01: Implement T3-PREVIEW requirements

Phase 309: Live Transfer-Budget Measurement & Job-Matrix Change Detector (T3-LIVE-CI-01..04)

Section titled “Phase 309: Live Transfer-Budget Measurement & Job-Matrix Change Detector (T3-LIVE-CI-01..04)”

Directory: phases/309-live-transfer-budget-and-job-matrix/ Goal: Replace fixture-only transfer-budget-report CI path with artifact emission from real vox-orchestrator-mcp transfer-b…

  • 309-01: Implement T3-LIVE-CI requirements

Phase 310: TOESTUB Test-Discipline Detectors (Wall-Clock, Real-Wait, Loop, Frame) (T3-TEST-DET-01..04)

Section titled “Phase 310: TOESTUB Test-Discipline Detectors (Wall-Clock, Real-Wait, Loop, Frame) (T3-TEST-DET-01..04)”

Directory: phases/310-test-discipline-detectors/ Goal: Extend test_clock_and_skip (or new detector) to flag wall-clock reads in tests: Instant::now, SystemTime::now, `Da…

  • 310-01: Implement T3-TEST-DET requirements

Phase 311: Host-Runtime Boundary, Schema Hoist, Knip Dual & DOM-Env Discipline (T3-HOST-SCHEMA-01..04)

Section titled “Phase 311: Host-Runtime Boundary, Schema Hoist, Knip Dual & DOM-Env Discipline (T3-HOST-SCHEMA-01..04)”

Directory: phases/311-host-runtime-schema-knip-dom/ Goal: New detector banning direct host platform/arch probes outside allowlisted host crates (std::env::consts::{OS,ARCH}, `p…

  • 311-01: Implement T3-HOST-SCHEMA requirements

Phase 312: Sidecar Fingerprint Hash Compare, Surface Graph Regen & v5.5 Scorecard (T3-FP-GRAPH-01..04)

Section titled “Phase 312: Sidecar Fingerprint Hash Compare, Surface Graph Regen & v5.5 Scorecard (T3-FP-GRAPH-01..04)”

Directory: phases/312-fingerprint-hash-graph-scorecard/ Goal: Upgrade gui-sidecar-fingerprint.yml from path-diff labeling to content hash of native/sidecar inputs (Cargo.lock subse…

  • 312-01: Implement T3-FP-GRAPH requirements
PhaseNamePlansStatus
Phase 307PR Vouch & Contributor Trust Gate1/1Complete
Phase 308Preview/Privileged Workflow Trust-Split Policy1/1Complete
Phase 309Live Transfer-Budget Measurement & Job-Matrix Change Detector1/1Complete (Partials)
Phase 310TOESTUB Test-Discipline Detectors (Wall-Clock, Real-Wait, Loop, Frame)1/1Complete
Phase 311Host-Runtime Boundary, Schema Hoist, Knip Dual & DOM-Env Discipline1/1Complete (Partials)
Phase 312Sidecar Fingerprint Hash Compare, Surface Graph Regen & v5.5 Scorecard1/1Complete (Partials)

  • ✅ v5.3 (Phases 295–300, 24/24 requirements — Complete 2026-10-25; CV residual): Photorealistic Edge-Softening, Continuous Monumental Aqueduct-to-Sewer Architecture, Sub-Surface Ground Integration & CV Flywheel Score Maximization — Requirements | Architecture Spec
  • ✅ v5.2 (Phases 289–294, 24/24 requirements — Complete 2026-10-25): Smooth Physically-Grounded Lighting, High-Detail Vitruvian Architecture & Grounds, Integrated Roman Hydraulic/Sewer Grid (Aquaeductus, Castellum Divisorium, Fistulae & Cloaca Maxima) & Animated Shape+Color Activity Conduits — Requirements | Architecture Spec
  • ✅ v5.1 (Phases 283–288, 24/24 requirements — Complete 2026-10-25): T3 Ecosystem Wave 5 Deep Review & End-to-End Parity Closure: System Prompt $variable & Worktree Branch Fix, Full context_sanitizer & ToolRefusalDetector, Sequenced ResumableTurnStreamRegistry, Full ThreadRunQueue & ModelSpec Cost Accounting, Async Worktree Lease & Monotonic Sparse Checkpoints, Real MultiVariantWorktreePool, Multi-Provider Consensus Synthesizer, Depth-4 External Session Discovery, Unified Tauri/MCP message_id & oxlint-plugin-t3code Zero-Violation Audit Parity — Requirements | Architecture Audit
  • ✅ v5.0 (Phases 277–282, 24/24 requirements — Complete 2026-10-25): Photorealistic Vitruvian Material, Depth-Shading, Atmospheric Lighting & FX Engine (pbrMaterials.ts, Projected Cast Shadows, Vertical AO, Specular Chamfer Bevels, Bathymetric Sea + Surf Foam, Crepuscular God-Rays, Additive GI Pools, Particle FX & Diegetic Cliff-Apron Plaques) — Requirements | Architecture Spec
  • ✅ v4.9 (Phases 271–276, 24/24 requirements — Complete 2026-10-25): Vitruvian Roman Tactical Developer Controller: Zero-Hop Diff/Diagnostic Loupe (Speculum Codicis), Multi-Agent Write-Set Collision Cones (Concursus Legionum), Clickable Viaduct Contract Inspector (Pons Foederis), Hard Agent Scope Quarantine Fencing (Clausura Ferrea), Milestone Road Turn Time-Scrubber (Via Temporis) & Split-Dock Workbench Integration (Tabula Praetoria ⌘M) — Requirements | Architecture Spec
  • ✅ v4.8 (Phases 265–270, 24/24 requirements — Complete 2026-10-25): T3 Wave 4 Live Pipeline End-to-End Closure: Real LLM Consensus, Stateful Session/Guest/Checkpoint Persistence, Live Agent-Loop Guards & Tauri/MCP Envelope Unwrapping — Requirements | Architecture Audit
  • ✅ v4.7 (Phases 259–264, 24/24 requirements — Shipped 2026-10-25): Universal Repo-Agnostic Codebase Civilization Simulator & Attention-Weighted Developer Control Surface — Archive | Requirements | Audit
  • ✅ v4.6 (Phases 253–258, 24/24 requirements — Shipped 2026-10-25): T3 Wave 3 Live Surface & Daemon RPC Wiring, Router/DPO Quality-Prior Flywheel & Workspace-Wide TOESTUB Remediation — Archive | Requirements | Audit
  • ✅ v4.5 (Phases 242–252, 44/44 requirements — Shipped 2026-10-25): T3 Ecosystem Wave 2 Advancements, Consensus Engine, Git Plumbing Checkpoints & TOESTUB Static Enforcement — Archive | Requirements
  • ✅ v4.4 (Phases 236–241, 24/24 requirements — Shipped 2026-10-25): Concept-Art Pixel & Structural Convergence, Live Animated Roman Civilization Simulator (VRBS ROMA), Interactive DAG/Diff Workbench & Before/After Visual Proof — Archive | Requirements
  • ✅ v4.3 (Phases 230–235, 24/24 requirements — Shipped 2026-10-25): Roman Civilization Codebase Simulator (VRBS ROMA), Strategic Minimap Radar & Cross-Surface Readability/Occlusion Polish — Requirements (no roadmap archive was written)

Phase 265: tool:<name> Fence Salvage Fix, Live @rule:<name> Mention Resolution & $variable System Prompt Interpolation (T3-SALVAGE-RULES-01..04)

Section titled “Phase 265: tool:<name> Fence Salvage Fix, Live @rule:<name> Mention Resolution & $variable System Prompt Interpolation (T3-SALVAGE-RULES-01..04)”

Goal: Eliminate the tool:<name> stream-suppression vs. salvage mismatch in salvage_and_stream.rs, wire resolve_user_rule_mentions (@rule:<name> + Always rules) into resolve_mentions before file-path matching, and wire interpolate_system_prompt_variables into live turn preparation in preamble_and_retrieval.rs. Requirements: T3-SALVAGE-RULES-01, T3-SALVAGE-RULES-02, T3-SALVAGE-RULES-03, T3-SALVAGE-RULES-04 Plans:

  • 265-01: Fix tool:<name> fence salvage in salvage_and_stream.rs, wire @rule:<name> resolution in mentions.rs, wire $variable interpolation in preamble_and_retrieval.rs, and add integration tests (phases/265-salvage-rules-and-prompt-vars/265-01-PLAN.md)

Phase 266: Live context_sanitizer History Stripping, vox_chat_compact_up_to, BatteryCostLedger Accounting, VoxIdAllocatorV2 Message IDs & Real ResponseQualityAnalyzer Telemetry (T3-HIST-BATTERY-01..04)

Section titled “Phase 266: Live context_sanitizer History Stripping, vox_chat_compact_up_to, BatteryCostLedger Accounting, VoxIdAllocatorV2 Message IDs & Real ResponseQualityAnalyzer Telemetry (T3-HIST-BATTERY-01..04)”

Goal: Wire sanitize_assistant_for_history and escape_prompt_injection_tokens into transcript_entry_to_llm_message (conversation.rs), implement vox_chat_compact_up_to (history.rs), and wire live BatteryCostLedger (check_preflight_balance + record_turn_usage), VoxIdAllocatorV2 deterministic message IDs, and real ResponseQualityAnalyzer scores into message.rs and persist_and_envelope.rs. Requirements: T3-HIST-BATTERY-01, T3-HIST-BATTERY-02, T3-HIST-BATTERY-03, T3-HIST-BATTERY-04 Plans:

  • 266-01: Wire context_sanitizer in conversation.rs, implement vox_chat_compact_up_to in history.rs, and wire BatteryCostLedger, VoxIdAllocatorV2, and ResponseQualityAnalyzer in message.rs & persist_and_envelope.rs (phases/266-history-sanitizer-and-battery-telemetry/266-01-PLAN.md)

Phase 267: Live agent_loop Integration: CommandOutputHub \r Normalization, 20 KiB DOM Shedding, Pre-Write Sparse Git Checkpoints, ResumableTurnStreamRegistry & Mid-Turn Steer Draining (T3-LOOP-CHECKPOINT-01..04)

Section titled “Phase 267: Live agent_loop Integration: CommandOutputHub \r Normalization, 20 KiB DOM Shedding, Pre-Write Sparse Git Checkpoints, ResumableTurnStreamRegistry & Mid-Turn Steer Draining (T3-LOOP-CHECKPOINT-01..04)”

Goal: Wire normalize_terminal_output (\r folding), bound_snapshot_metadata (20 KiB DOM shedding), automatic pre-write sparse Git checkpoints (should_sparse_checkpoint), ResumableTurnStreamRegistry (vox_chat_stream_resume), and mid-turn ThreadRunQueue (vox_chat_queue_steer) into agent_loop.rs and salvage_and_stream.rs. Requirements: T3-LOOP-CHECKPOINT-01, T3-LOOP-CHECKPOINT-02, T3-LOOP-CHECKPOINT-03, T3-LOOP-CHECKPOINT-04 Plans:

  • 267-01: Wire \r normalization, 20 KiB DOM shedding, pre-write sparse Git checkpoints, ResumableTurnStreamRegistry, and mid-turn ThreadRunQueue steer draining into agent_loop.rs & salvage_and_stream.rs (phases/267-agent-loop-checkpoints-and-resumable-streams/267-01-PLAN.md)

Phase 268: Stateful t3_wave3_tools.rs: Concurrent join_all LLM Consensus Fan-Out, Live Transcript Hydration/Persistence for vox_context_handoff & vox_session_* & Persistent GuestShareService (T3-STATEFUL-TOOLS-01..04)

Section titled “Phase 268: Stateful t3_wave3_tools.rs: Concurrent join_all LLM Consensus Fan-Out, Live Transcript Hydration/Persistence for vox_context_handoff & vox_session_* & Persistent GuestShareService (T3-STATEFUL-TOOLS-01..04)”

Goal: Upgrade t3_wave3_tools.rs so vox_consensus_run dispatches concurrent join_all LLM calls across models, vox_context_handoff and vox_session_* hydrate and persist real transcripts in ServerState (chat_history:{session_id} + real ~/.claude/~/.codex/~/.cursor files), and vox_guest_share_* uses a persistent GuestShareService on ServerState. Requirements: T3-STATEFUL-TOOLS-01, T3-STATEFUL-TOOLS-02, T3-STATEFUL-TOOLS-03, T3-STATEFUL-TOOLS-04 Plans:

  • 268-01: Upgrade t3_wave3_tools.rs to dispatch concurrent join_all LLM consensus legs, hydrate/persist live ServerState transcripts in vox_context_handoff & vox_session_*, and persist GuestShareService on ServerState (phases/268-stateful-consensus-handoff-and-session-tools/268-01-PLAN.md)

Phase 269: vox_chat_regenerate & vox_chat_edit_message_in_place Backend MCP Tools, DelegationWorktree Lease/Env Wiring, submit_webhook_events Dedupe/Rate Gate & Compact vox_orchestrator_status (T3-REGEN-WORKTREE-01..04)

Section titled “Phase 269: vox_chat_regenerate & vox_chat_edit_message_in_place Backend MCP Tools, DelegationWorktree Lease/Env Wiring, submit_webhook_events Dedupe/Rate Gate & Compact vox_orchestrator_status (T3-REGEN-WORKTREE-01..04)”

Goal: Implement vox_chat_regenerate and vox_chat_edit_message_in_place in history.rs, wire WorkspaceLeaseMap + provision_untracked_env_files in DelegationWorktree::create, wire WebhookDedupeAndRateGate in submit_webhook_events, and wire compact in vox_orchestrator_status + transfer_budget.rs. Requirements: T3-REGEN-WORKTREE-01, T3-REGEN-WORKTREE-02, T3-REGEN-WORKTREE-03, T3-REGEN-WORKTREE-04 Plans:

  • 269-01: Implement vox_chat_regenerate & vox_chat_edit_message_in_place, wire WorkspaceLeaseMap & .env provisioning in agy_worktree.rs, wire WebhookDedupeAndRateGate in webhook_intake.rs, and wire compact in vox_orchestrator_status & transfer_budget.rs (phases/269-regenerate-worktree-lease-and-webhook-gate/269-01-PLAN.md)

Phase 270: t3Wave3Rpc.ts Nested Tauri Envelope Unwrapping, Live ChatTranscript.tsx In-Place Edit & Posture-Aware Regeneration & Full-Stack Verification (T3-GUI-ENVELOPE-01..04)

Section titled “Phase 270: t3Wave3Rpc.ts Nested Tauri Envelope Unwrapping, Live ChatTranscript.tsx In-Place Edit & Posture-Aware Regeneration & Full-Stack Verification (T3-GUI-ENVELOPE-01..04)”

Goal: Fix unwrapMcpData in t3Wave3Rpc.ts to unwrap nested Tauri { is_error, result: { success, data } } and stringified MCP envelopes, expose typed editChatMessageInPlace / regenerateChatTurn / compactChatUpTo / resumeChatStream / steerChatQueue client helpers, wire in-place edit and posture regeneration in ChatTranscript.tsx, and verify all Rust, Vitest, TypeScript, and audit gates. Requirements: T3-GUI-ENVELOPE-01, T3-GUI-ENVELOPE-02, T3-GUI-ENVELOPE-03, T3-GUI-ENVELOPE-04 Plans:

  • 270-01: Fix unwrapMcpData nested Tauri/MCP envelope unwrapping in t3Wave3Rpc.ts, wire in-place edit & regenerate in ChatTranscript.tsx, and run full-stack verification (phases/270-tauri-envelope-unwrapping-and-transcript-actions/270-01-PLAN.md)

PhaseTitlePlansStatus
Phase 265tool:<name> Fence Salvage Fix, Live @rule:<name> Mention Resolution & $variable System Prompt Interpolation1/1Complete (2026-10-25)
Phase 266Live context_sanitizer History Stripping, vox_chat_compact_up_to, BatteryCostLedger Accounting, VoxIdAllocatorV2 Message IDs & Real ResponseQualityAnalyzer Telemetry1/1Complete (2026-10-25)
Phase 267Live agent_loop Integration: CommandOutputHub \r Normalization, 20 KiB DOM Shedding, Pre-Write Sparse Git Checkpoints, ResumableTurnStreamRegistry & Mid-Turn Steer Draining1/1Complete (2026-10-25)
Phase 268Stateful t3_wave3_tools.rs: Concurrent join_all LLM Consensus Fan-Out, Live Transcript Hydration/Persistence for vox_context_handoff & vox_session_* & Persistent GuestShareService1/1Complete (2026-10-25)
Phase 269vox_chat_regenerate & vox_chat_edit_message_in_place Backend MCP Tools, DelegationWorktree Lease/Env Wiring, submit_webhook_events Dedupe/Rate Gate & Compact vox_orchestrator_status1/1Complete (2026-10-25)
Phase 270t3Wave3Rpc.ts Nested Tauri Envelope Unwrapping, Live ChatTranscript.tsx In-Place Edit & Posture-Aware Regeneration & Full-Stack Verification1/1Complete (2026-10-25)

Phases (v4.9 — Vitruvian Roman Tactical Developer Controller)

Section titled “Phases (v4.9 — Vitruvian Roman Tactical Developer Controller)”

Phase 271: Zero-Hop Inline Diff & Compiler Diagnostic Loupe (Speculum Codicis) (URBS-LOUPE-01..04)

Section titled “Phase 271: Zero-Hop Inline Diff & Compiler Diagnostic Loupe (Speculum Codicis) (URBS-LOUPE-01..04)”

Goal: Eliminate context-switching between VRBS and external diff/terminal tabs by adding workspace_file_diff_and_diagnostics in workspace_town.rs and rendering SpeculumLoupePopover.tsx directly above any clicked or hovered Treadwheel Crane (+LoC / -LoC) or red error building (N ERR) with real unified Git diff hunks, file:line:col compiler diagnostics, and 1-click [✓ Stage File], [↺ Revert File], and [⚒ Fix Exact Error] actions. Requirements: URBS-LOUPE-01, URBS-LOUPE-02, URBS-LOUPE-03, URBS-LOUPE-04 Plans:

  • 271-01: Implement workspace_file_diff_and_diagnostics + workspace_stage_file / workspace_revert_file in workspace_town.rs, build SpeculumLoupePopover.tsx, wire into LudusSandbox.tsx, and add Rust/Vitest tests (phases/271-inline-diff-and-diagnostic-loupe/271-01-PLAN.md)

Phase 272: Multi-Agent Write-Set Collision & Blast-Radius Cone Radar (Concursus Legionum) (URBS-COLLISION-01..04)

Section titled “Phase 272: Multi-Agent Write-Set Collision & Blast-Radius Cone Radar (Concursus Legionum) (URBS-COLLISION-01..04)”

Goal: Prevent concurrent multi-agent merge conflicts and semantic clobbering by computing each active agent’s write-set + 1-hop blast-radius cone (collisionRadar.ts), rendering intersecting Roman surveyor groma cones and a 3D Hammered-Bronze Crossed-Swords Collision Standard (⚠ WRITE-SET COLLISION) in worldRenderer.ts, and wiring 1-click [⇄ Serialize Agents] and [🌿 Split Worktree] actions. Requirements: URBS-COLLISION-01, URBS-COLLISION-02, URBS-COLLISION-03, URBS-COLLISION-04 Plans:

  • 272-01: Build collisionRadar.ts, render surveyor groma cones & Crossed-Swords Collision Standard in worldRenderer.ts, wire [⇄ Serialize Agents] & [🌿 Split Worktree] in LudusSandbox.tsx / ImperiumInspector.tsx, and add Vitest tests (phases/272-multi-agent-collision-cone-radar/272-01-PLAN.md)

Phase 273: Clickable Viaduct Cross-Module Contract & Layer-Inversion Inspector (Pons Foederis) (URBS-VIADUCT-01..04)

Section titled “Phase 273: Clickable Viaduct Cross-Module Contract & Layer-Inversion Inspector (Pons Foederis) (URBS-VIADUCT-01..04)”

Goal: Turn stone viaducts between districts into clickable first-class architectural contract boundaries by scanning cross-module dependencies and layers.toml layer assignments in workspace_town.rs (TownDependencyEdge), adding viaduct hit-testing in layout.ts, rendering a Cracked Red Arch Keystone (⚠ LAYER VIOLATION) in worldRenderer.ts on illegal upward dependencies, and adding the PONS CONTRACTVS inspector panel in ImperiumInspector.tsx. Requirements: URBS-VIADUCT-01, URBS-VIADUCT-02, URBS-VIADUCT-03, URBS-VIADUCT-04 Plans:

  • 273-01: Add TownDependencyEdge + layers.toml violation scanner in workspace_town.rs, viaduct hit-testing in layout.ts, cracked keystone rendering in worldRenderer.ts, and PONS CONTRACTVS inspector panel in ImperiumInspector.tsx (phases/273-clickable-viaduct-contract-inspector/273-01-PLAN.md)

Phase 274: Hard Agent Scope Quarantine Fencing (Clausura Ferrea) (URBS-QUARANTINE-01..04)

Section titled “Phase 274: Hard Agent Scope Quarantine Fencing (Clausura Ferrea) (URBS-QUARANTINE-01..04)”

Goal: Give developers a physical spatial guardrail that prevents agents from modifying off-limits modules by storing quarantined_paths on ServerState, exposing vox_quarantine_set / vox_quarantine_list MCP & Tauri RPCs, enforcing pre-execution write rejection in agent_loop.rs, and rendering a 3D Wrought-Iron Perimeter Portcullis & Bronze Padlock (🔒 QUARANTINED: READ-ONLY) toggled via Alt+Click on any district. Requirements: URBS-QUARANTINE-01, URBS-QUARANTINE-02, URBS-QUARANTINE-03, URBS-QUARANTINE-04 Plans:

  • 274-01: Add quarantined_paths to ServerState, expose vox_quarantine_set / vox_quarantine_list, enforce write rejection in agent_loop.rs, and render 3D perimeter portcullis fencing in worldRenderer.ts (phases/274-hard-agent-quarantine-fencing/274-01-PLAN.md)

Phase 275: Interactive Milestone Road Turn Time-Scrubber (Via Temporis Visual git bisect) (URBS-SCRUBBER-01..04)

Section titled “Phase 275: Interactive Milestone Road Turn Time-Scrubber (Via Temporis Visual git bisect) (URBS-SCRUBBER-01..04)”

Goal: Turn the Stone Milestone Road (I, II, III, IV) at PORTA GIT into an interactive visual git bisect timeline bar (ViaTemporisScrubber.tsx) with a draggable Bronze Sundial Gnomon Scrubber that previews per-checkpoint building diffs (+LoC / -LoC), test shield state flips (PASS → FAIL), and token/BU spend, with 1-click [↺ Rollback Worktree to Milestone] wired to vcsCheckpointRestore. Requirements: URBS-SCRUBBER-01, URBS-SCRUBBER-02, URBS-SCRUBBER-03, URBS-SCRUBBER-04 Plans:

  • 275-01: Add vox_vcs_checkpoint_diff_summary RPC, build ViaTemporisScrubber.tsx sundial timeline bar, wire historical building diff preview & 1-click rollback in LudusSandbox.tsx, and add Vitest tests (phases/275-milestone-road-time-scrubber/275-01-PLAN.md)

Phase 276: Split-Dock Tactical Map Workbench Integration (Tabula Praetoria ⌘M) (URBS-DOCK-01..04)

Section titled “Phase 276: Split-Dock Tactical Map Workbench Integration (Tabula Praetoria ⌘M) (URBS-DOCK-01..04)”

Goal: Promote VRBS from Settings → Gamify to a 1-click split-dock Tactical Map ([🏛 Tactical Map ⌘M]) inside DockWorkspaceShell.tsx and ChatSurface.tsx, wiring bi-directional district context filtering between the map and Chat/Review plus real-time tool-call mechanism animations (read_file → Screw Pump, write_file → Treadwheel Crane, test → Shield Wall, approval → Portcullis Gate). Requirements: URBS-DOCK-01, URBS-DOCK-02, URBS-DOCK-03, URBS-DOCK-04 Plans:

  • 276-01: Wire [🏛 Tactical Map ⌘M] split-dock panel in DockWorkspaceShell.tsx & ChatSurface.tsx, bi-directional district filter in useAppStore.ts, live tool-call mechanism pulses in LudusSandbox.tsx, and Playwright CV Flywheel verification (phases/276-split-dock-tactical-map-sync/276-01-PLAN.md)

PhaseTitlePlansStatus
Phase 271Zero-Hop Inline Diff & Compiler Diagnostic Loupe (Speculum Codicis)1/1Complete (2026-10-25)
Phase 272Multi-Agent Write-Set Collision & Blast-Radius Cone Radar (Concursus Legionum)1/1Complete (2026-10-25)
Phase 273Clickable Viaduct Cross-Module Contract & Layer-Inversion Inspector (Pons Foederis)1/1Complete (2026-10-25)
Phase 274Hard Agent Scope Quarantine Fencing (Clausura Ferrea)1/1Complete (2026-10-25)
Phase 275Interactive Milestone Road Turn Time-Scrubber (Via Temporis Visual git bisect)1/1Complete (2026-10-25)
Phase 276Split-Dock Tactical Map Workbench Integration (Tabula Praetoria ⌘M)1/1Complete (2026-10-25)

Phases (v5.0 — Photorealistic Vitruvian Material, Depth-Shading, Atmospheric Lighting & FX Engine)

Section titled “Phases (v5.0 — Photorealistic Vitruvian Material, Depth-Shading, Atmospheric Lighting & FX Engine)”

Phase 277: Procedural Natural Surface Material Synthesis (pbrMaterials.ts) (URBS-MAT-01..04)

Section titled “Phase 277: Procedural Natural Surface Material Synthesis (pbrMaterials.ts) (URBS-MAT-01..04)”

Goal: Eliminate flat solid-color polygon fills by creating a deterministic offscreen procedural material synthesis atlas (pbrMaterials.ts) producing tileable CanvasPatterns and metallic gradients for Pitted Roman Travertine Ashlar, Scalloped Terracotta Roof Tiles, Opus Incertum Flagstone Plazas, Stratified Basalt Cliffs, and Anisotropic Hammered Bronze. Requirements: URBS-MAT-01, URBS-MAT-02, URBS-MAT-03, URBS-MAT-04 Plans:

  • 277-01: Create pbrMaterials.ts procedural natural surface material atlas + unit tests (phases/277-procedural-natural-materials/277-01-PLAN.md)

Phase 278: True Isometric Directional Cast Shadows, Contact Ambient Occlusion & Specular Chamfer Bevels (URBS-DEPTH-01..04)

Section titled “Phase 278: True Isometric Directional Cast Shadows, Contact Ambient Occlusion & Specular Chamfer Bevels (URBS-DEPTH-01..04)”

Goal: Ground every 3D structure with physical depth by adding projected directional twilight cast shadows (drawIsoCastShadow), vertical height-shaded face gradients (contact AO at base → golden-hour rim light at cornice) overlaid with travertine ashlar texture, and physical specular chamfer ridge highlights replacing flat black cartoon outlines. Requirements: URBS-DEPTH-01, URBS-DEPTH-02, URBS-DEPTH-03, URBS-DEPTH-04 Plans:

  • 278-01: Implement projected directional cast shadows, vertical AO/rim gradients, and specular chamfer bevels in worldRenderer.ts (phases/278-depth-shading-and-cast-shadows/278-01-PLAN.md)

Phase 279: Sculpted Coastal Bathymetry, Surf Foam Ribbons, Craggy Cliff Geology & Cobblestone Terraces (URBS-TERRAIN-01..04)

Section titled “Phase 279: Sculpted Coastal Bathymetry, Surf Foam Ribbons, Craggy Cliff Geology & Cobblestone Terraces (URBS-TERRAIN-01..04)”

Goal: Transform the flat background and terraces into a natural Mediterranean peninsula with multi-stop bathymetric sea depth gradients, coastal surf foam ribbons, stratified basalt cliff geology with vertical fissures and moss ledges, and cobblestone flagstone plazas. Requirements: URBS-TERRAIN-01, URBS-TERRAIN-02, URBS-TERRAIN-03, URBS-TERRAIN-04 Plans:

  • 279-01: Upgrade paintFullBleedRomanLandscape with bathymetric sea depth, coastal surf foam, stratified cliff textures, and cobblestone terrace paving (phases/279-bathymetry-cliffs-and-terrace-geology/279-01-PLAN.md)

Phase 280: Golden-Hour & Twilight Dual-Key Atmospheric Lighting, Volumetric God-Rays, Elevation Mist & Additive GI Pools (lightingAndFx.ts) (URBS-LIGHT-01..04)

Section titled “Phase 280: Golden-Hour & Twilight Dual-Key Atmospheric Lighting, Volumetric God-Rays, Elevation Mist & Additive GI Pools (lightingAndFx.ts) (URBS-LIGHT-01..04)”

Goal: Deliver cinematic atmospheric lighting by creating lightingAndFx.ts with inter-terrace atmospheric elevation mist, diagonal golden-hour crepuscular god-rays, additive point-light GI bounce pools around braziers/forges/windows/aqueducts, and a cinematic vignette & warm highlight bloom pass. Requirements: URBS-LIGHT-01, URBS-LIGHT-02, URBS-LIGHT-03, URBS-LIGHT-04 Plans:

  • 280-01: Create lightingAndFx.ts atmospheric mist, volumetric god-rays, additive GI light pools, and cinematic post-processing (phases/280-atmospheric-lighting-and-god-rays/280-01-PLAN.md)

Phase 281: Live Particle FX Engine (Forge Embers, Volumetric Smoke Plumes, Waterfall Spray & Mason Chisel Sparks) (URBS-FX-01..04)

Section titled “Phase 281: Live Particle FX Engine (Forge Embers, Volumetric Smoke Plumes, Waterfall Spray & Mason Chisel Sparks) (URBS-FX-01..04)”

Goal: Bring the 3D mechanisms to life with deterministic real-time particle emitters in lightingAndFx.ts and drawAnimatedWorldEffects: rising glowing forge/error embers, drifting volumetric smoke plumes, waterfall mist spray, harbor water caustics, and mason chisel sparks on active Treadwheel Cranes. Requirements: URBS-FX-01, URBS-FX-02, URBS-FX-03, URBS-FX-04 Plans:

  • 281-01: Implement live forge ember, smoke plume, waterfall spray, water caustic, and mason chisel spark particle emitters (phases/281-live-particle-fx-engine/281-01-PLAN.md)

Phase 282: Diegetic Cliff-Apron Architrave Plaques, Architecture-First HUD Decluttering & Playwright CV Flywheel Convergence (URBS-DIEGETIC-01..04)

Section titled “Phase 282: Diegetic Cliff-Apron Architrave Plaques, Architecture-First HUD Decluttering & Playwright CV Flywheel Convergence (URBS-DIEGETIC-01..04)”

Goal: Uncover 85%+ of the 3D architectural diorama by mounting district plaques as chiseled bronze/travertine inscriptions on the front cliff retaining wall apron, progressive-disclosing secondary file slabs on hover/selection, upgrading sprites.ts with photorealistic gradients/shadows, and verifying via the Playwright CV Flywheel. Requirements: URBS-DIEGETIC-01, URBS-DIEGETIC-02, URBS-DIEGETIC-03, URBS-DIEGETIC-04 Plans:

  • 282-01: Implement diegetic cliff-apron architrave plaques, HUD progressive disclosure, photorealistic sprite upgrades, and Playwright CV Flywheel verification (phases/282-diegetic-plaques-and-cv-convergence/282-01-PLAN.md)

PhaseTitlePlansStatus
Phase 277Procedural Natural Surface Material Synthesis (pbrMaterials.ts)1/1Complete (2026-10-25)
Phase 278True Isometric Directional Cast Shadows, Contact Ambient Occlusion & Specular Chamfer Bevels1/1Complete (2026-10-25)
Phase 279Sculpted Coastal Bathymetry, Surf Foam Ribbons, Craggy Cliff Geology & Cobblestone Terraces1/1Complete (2026-10-25)
Phase 280Golden-Hour & Twilight Dual-Key Atmospheric Lighting, Volumetric God-Rays, Elevation Mist & Additive GI Pools1/1Complete (2026-10-25)
Phase 281Live Particle FX Engine (Forge Embers, Volumetric Smoke Plumes, Waterfall Spray & Mason Chisel Sparks)1/1Complete (2026-10-25)
Phase 282Diegetic Cliff-Apron Architrave Plaques, Architecture-First HUD Decluttering & Playwright CV Flywheel Convergence1/1Complete (2026-10-25)

Phases (v5.1 — T3 Ecosystem Wave 5 Deep Review & End-to-End Parity Closure)

Section titled “Phases (v5.1 — T3 Ecosystem Wave 5 Deep Review & End-to-End Parity Closure)”

Phase 283: System Prompt $variable Target Fix, Sanitizer Tag Completeness, ToolRefusalDetector & Lazy MCP Meta-Tools (T3-PROMPT-LOOP-01..04)

Section titled “Phase 283: System Prompt $variable Target Fix, Sanitizer Tag Completeness, ToolRefusalDetector & Lazy MCP Meta-Tools (T3-PROMPT-LOOP-01..04)”

Goal: Apply interpolate_system_prompt_variables to system_prompt with full ISO-8601 UTC $current_datetime and worktree .git file gitdir: branch resolution, expand escape_user_prompt_special_tokens (<tool_call>, <tool_response>, <think>, active-turn user prompt) + compact_historical_web_search (<WebFetch>) + thought_signature extraction, wire pi-t3chat ToolRefusalDetector ([SYSTEM TOOL CORRECTION]) and NotT3Chat search circuit breaker (total_searches >= 3) in agent_loop.rs, and wire LazyMcpGroupRegistry (vox_list_mcp_servers, vox_list_mcp_tools, vox_call_mcp_lazy) + sanitize_mcp_tool_search_env. Requirements: T3-PROMPT-LOOP-01, T3-PROMPT-LOOP-02, T3-PROMPT-LOOP-03, T3-PROMPT-LOOP-04 Plans:

  • 283-01: Fix system_prompt $variable interpolation & worktree branch detection, expand context_sanitizer & thought_signature persistence, wire ToolRefusalDetector & search circuit breaker in agent_loop.rs, and wire LazyMcpGroupRegistry (phases/283-prompt-vars-sanitizer-and-loop-guards/283-01-PLAN.md)

Phase 284: Sequenced ResumableTurnStreamRegistry, Full ThreadRunQueue Lifecycle, Real ModelSpec Cost Ledger & SessionPrivacyPosture Gate (T3-STREAM-QUEUE-01..04)

Section titled “Phase 284: Sequenced ResumableTurnStreamRegistry, Full ThreadRunQueue Lifecycle, Real ModelSpec Cost Ledger & SessionPrivacyPosture Gate (T3-STREAM-QUEUE-01..04)”

Goal: Replace ServerState::resumable_streams tuple map with ResumableTurnStreamRegistry (SequencedTurnChunk, multi-iteration accumulation, vox_chat_stream_resume + vox_chat_stream_cancel), wire ThreadRunQueue (pop_next_follow_up, background work settlement, 429 LimitRecoveryState), wire post-resolution preflight check + ModelRegistry::compute_turn_cost_delta + 4-hour/monthly quota windows + vox_battery_status, and enforce SessionPrivacyPosture::Ephemeral disk-persistence gating. Requirements: T3-STREAM-QUEUE-01, T3-STREAM-QUEUE-02, T3-STREAM-QUEUE-03, T3-STREAM-QUEUE-04 Plans:

  • 284-01: Wire ResumableTurnStreamRegistry, full ThreadRunQueue follow-up/settlement/recovery, ModelRegistry::compute_turn_cost_delta + vox_battery_status, and SessionPrivacyPosture persistence gate (phases/284-resumable-stream-queue-and-cost-ledger/284-01-PLAN.md)

Phase 285: Worktree Async Lease Mutex Fix, Persistent Webhook Gate, Monotonic Sparse Git Checkpoints, Real MultiVariantWorktreePool & Compact Snapshot Fix (T3-WORKTREE-CKPT-01..04)

Section titled “Phase 285: Worktree Async Lease Mutex Fix, Persistent Webhook Gate, Monotonic Sparse Git Checkpoints, Real MultiVariantWorktreePool & Compact Snapshot Fix (T3-WORKTREE-CKPT-01..04)”

Goal: Hold a tokio::sync::MutexGuard across .await in DelegationWorktree::create and persist WebhookDedupeAndRateGate on ServerState, fix monotonic per-session sparse checkpoint turn counters + Windows NTFS temp_index sanitization + -p HEAD parent commits + clean untracked restore + diff_checkpoints, materialize real linked worktrees in MultiVariantWorktreePool (vox_muse_pool_spawn / vox_muse_pool_adopt), and fix orchestrator_status_compact double-envelope + parent_agent_id subagent filtering + LiveBodyDeduper. Requirements: T3-WORKTREE-CKPT-01, T3-WORKTREE-CKPT-02, T3-WORKTREE-CKPT-03, T3-WORKTREE-CKPT-04 Plans:

  • 285-01: Fix async worktree lease lock & persistent webhook gate, monotonic sparse Git checkpoints & clean restore, real MultiVariantWorktreePool worktrees, and unwrapped orchestrator_status_compact (phases/285-worktree-lease-checkpoints-and-muse-pool/285-01-PLAN.md)

Phase 286: UTF-8 Boundary Safety, Multi-Provider Consensus Synthesizer, Recursive Session Discovery & Real Guest Share / Diagnostics (T3-TOOLS-HARDENING-01..04)

Section titled “Phase 286: UTF-8 Boundary Safety, Multi-Provider Consensus Synthesizer, Recursive Session Discovery & Real Guest Share / Diagnostics (T3-TOOLS-HARDENING-01..04)”

Goal: Fix UTF-8 char-boundary truncation panics in context_handoff.rs & mcp_tool_access.rs + real multi-turn digest in HandoffStrategy::Summarized, check all configured LLM providers + run synthesize_consensus_response in handle_consensus_run, recursively scan depth-4 external sessions (~/.claude/projects, ~/.codex/sessions, ~/.cursor/chats + vox_session_list_external) + expose vox_guest_share_redeem / vox_guest_share_submit, and enforce server-side access caps in authorize_wave3_call + persistent PullRequestReviewGuideService + real diagnostics in handle_file_diff_and_diagnostics. Requirements: T3-TOOLS-HARDENING-01, T3-TOOLS-HARDENING-02, T3-TOOLS-HARDENING-03, T3-TOOLS-HARDENING-04 Plans:

  • 286-01: Fix UTF-8 boundary panics & summarized handoff digest, multi-provider consensus synthesizer, depth-4 external session discovery & guest share submit/redeem, and server-side access / real diagnostics (phases/286-utf8-safety-consensus-and-session-discovery/286-01-PLAN.md)

Phase 287: Unified message_id & Telemetry Across Tauri/MCP, Live Transcript Actions, Mid-Turn Steer Composer & streamWatchdog.ts (T3-GUI-PIPELINE-01..04)

Section titled “Phase 287: Unified message_id & Telemetry Across Tauri/MCP, Live Transcript Actions, Mid-Turn Steer Composer & streamWatchdog.ts (T3-GUI-PIPELINE-01..04)”

Goal: Preserve backend_message_id (usr:.../ast:...) + battery/quality/posture telemetry across vox-gui (chat.rs/chat_turn.rs) and history.rs (with turn-index fallback & SQLite sync), wire live user+assistant ✎ Edit, executing ↻ Regenerate (execute: true), ✂ Compact Here (compactChatUpTo), /compact slash command, and real branch comparison in ChatTranscript.tsx, fix the Enter-key task-abort bug and wire [⚡ Steer Turn] / [+ Queue Next] (steerChatQueue) in Loquela.tsx, and implement streamWatchdog.ts + resumeChatStream + full ChatExecutionRail / SessionInteropModal / ConsensusGridView / GuestShare wiring. Requirements: T3-GUI-PIPELINE-01, T3-GUI-PIPELINE-02, T3-GUI-PIPELINE-03, T3-GUI-PIPELINE-04 Plans:

  • 287-01: Unify message_id & telemetry across Tauri/MCP, wire live edit/regenerate/compact & mid-turn steer composer, and implement streamWatchdog.ts + full modal/rail RPC wiring (phases/287-unified-message-id-and-live-gui-actions/287-01-PLAN.md)

Phase 288: vox-code-audit T3 Detector Catalog Registration, oxlint-plugin-t3code Rule Parity & 43-Violation Workspace Remediation (T3-AUDIT-PARITY-01..04)

Section titled “Phase 288: vox-code-audit T3 Detector Catalog Registration, oxlint-plugin-t3code Rule Parity & 43-Violation Workspace Remediation (T3-AUDIT-PARITY-01..04)”

Goal: Register all T3 detector rule IDs in diagnostics/catalog.rs (ALL_KNOWN_IDS) with diagnostic_id: Some(...), upgrade unreasoned_suppression.rs, test_clock_and_skip.rs, mcp_access_and_error_safety.rs, and ui_perf_hazards.rs to full oxlint-plugin-t3code parity, refactor raw Command::new calls in t3_wave3_tools.rs, and remediate all 43 live workspace violations across vox-gui, vox-cli, and vox-orchestrator-mcp. Requirements: T3-AUDIT-PARITY-01, T3-AUDIT-PARITY-02, T3-AUDIT-PARITY-03, T3-AUDIT-PARITY-04 Plans:

  • 288-01: Register T3 detectors in diagnostics/catalog.rs, upgrade all 4 detectors to oxlint-plugin-t3code parity, and remediate all 43 workspace violations (phases/288-code-audit-parity-and-workspace-remediation/288-01-PLAN.md)

PhaseTitlePlansStatus
Phase 283System Prompt $variable Target Fix, Sanitizer Tag Completeness, ToolRefusalDetector & Lazy MCP Meta-Tools1/1Complete (2026-10-25)
Phase 284Sequenced ResumableTurnStreamRegistry, Full ThreadRunQueue Lifecycle, Real ModelSpec Cost Ledger & SessionPrivacyPosture Gate1/1Complete (2026-10-25)
Phase 285Worktree Async Lease Mutex Fix, Persistent Webhook Gate, Monotonic Sparse Git Checkpoints, Real MultiVariantWorktreePool & Compact Snapshot Fix1/1Complete (2026-10-25)
Phase 286UTF-8 Boundary Safety, Multi-Provider Consensus Synthesizer, Recursive Session Discovery & Real Guest Share / Diagnostics1/1Complete (2026-10-25)
Phase 287Unified message_id & Telemetry Across Tauri/MCP, Live Transcript Actions, Mid-Turn Steer Composer & streamWatchdog.ts1/1Complete (2026-10-25)
Phase 288vox-code-audit T3 Detector Catalog Registration, oxlint-plugin-t3code Rule Parity & 43-Violation Workspace Remediation1/1Complete (2026-10-25)

Phases (v5.2 — Smooth Physically-Grounded Lighting, High-Detail Vitruvian Architecture & Grounds, Integrated Roman Hydraulic/Sewer Grid & Animated Shape+Color Activity Conduits)

Section titled “Phases (v5.2 — Smooth Physically-Grounded Lighting, High-Detail Vitruvian Architecture & Grounds, Integrated Roman Hydraulic/Sewer Grid & Animated Shape+Color Activity Conduits)”

Phase 289: Smooth Ground-Plane-Recessed Lighting, 2D-Feathered Elliptical Sun-Shafts & Periodic Seamless Sea Synthesis (URBS-SMOOTH-LIGHT-01..04)

Section titled “Phase 289: Smooth Ground-Plane-Recessed Lighting, 2D-Feathered Elliptical Sun-Shafts & Periodic Seamless Sea Synthesis (URBS-SMOOTH-LIGHT-01..04)”

Goal: Eliminate alpha-overlay lighting artifacts and repeating water tile seams by moving warm brazier/forge/worksite bounce illumination onto the terrace ground plane before 3D buildings are painted, replacing hard-edged polygon god-rays with 2D-feathered elliptical radial gradients, replacing the hard ctx.arc district pulse with soft torch/window rim shimmer, and synthesizing 2π-periodic seamless water waves in pbrMaterials.ts. Requirements: URBS-SMOOTH-LIGHT-01, URBS-SMOOTH-LIGHT-02, URBS-SMOOTH-LIGHT-03, URBS-SMOOTH-LIGHT-04 Plans:

  • 289-01: Implement pre-building ground-plane illumination, 2D-feathered elliptical sun-shafts, soft torch/window rim shimmer, and periodic seamless sea water synthesis (phases/289-smooth-ground-lighting-and-seamless-sea/289-01-PLAN.md)

Phase 290: High-Detail Terrace Grounds: Orthogonal Cardo/Decumanus Streets, Opus Sectile Marble Plazas, Vaulted Substructio Cliff Arcades & Balustrades (URBS-GROUNDS-01..04)

Section titled “Phase 290: High-Detail Terrace Grounds: Orthogonal Cardo/Decumanus Streets, Opus Sectile Marble Plazas, Vaulted Substructio Cliff Arcades & Balustrades (URBS-GROUNDS-01..04)”

Goal: Transform flat district terraces and plain cliff retaining walls into richly articulated Roman urban grounds featuring recessed basalt Cardo/Decumanus streets with raised travertine curbstones and stepping-stone crosswalks, central Opus Sectile geometric marble forum plazas around District Wonders, vaulted Substructio blind arcades and buttress pilasters on cliff walls, and perimeter stone balustrades with boundary obelisks (Termini). Requirements: URBS-GROUNDS-01, URBS-GROUNDS-02, URBS-GROUNDS-03, URBS-GROUNDS-04 Plans:

  • 290-01: Upgrade terrace grounds and cliff walls in worldRenderer.ts with Cardo/Decumanus streets, Opus Sectile plazas, Substructio cliff arcades, and stone balustrades (phases/290-high-detail-terrace-grounds-and-substructio/290-01-PLAN.md)

Phase 291: High-Detail Vitruvian Building Architecture (T0–T3): Rusticated Ground Arcades (Fornices), Belt Cornices, Atrium Impluvia & Acroteria Roofs (URBS-ARCH-DETAIL-01..04)

Section titled “Phase 291: High-Detail Vitruvian Building Architecture (T0–T3): Rusticated Ground Arcades (Fornices), Belt Cornices, Atrium Impluvia & Acroteria Roofs (URBS-ARCH-DETAIL-01..04)”

Goal: Enrich all T0–T3 file buildings in worldRenderer.ts with ground-floor rusticated stone arcades (Fornices) and warm shopfront lamplight, projecting inter-story Marcapiano belt cornices and corner quoins, recessed atrium Impluvium rainwater basins in T1 Domus courtyards, and sculpted rooftop Acroteria finials and brick hypocaust chimneys. Requirements: URBS-ARCH-DETAIL-01, URBS-ARCH-DETAIL-02, URBS-ARCH-DETAIL-03, URBS-ARCH-DETAIL-04 Plans:

  • 291-01: Upgrade drawIsoBox, drawIsoPitchedRoof, and drawIsoFileBuilding with rusticated ground arcades, belt cornices, atrium impluvia, and acroteria roofs (phases/291-high-detail-vitruvian-building-architecture/291-01-PLAN.md)

Phase 292: Integrated Roman Hydraulic Plumbing & Sewer Network (hydraulicNetwork.ts): Aquaeductus Trunk, Castellum Divisorium, Fistulae Supply Pipes & Cloaca Maxima Sewer Trenches/Outfalls (URBS-HYDRAULIC-01..04)

Section titled “Phase 292: Integrated Roman Hydraulic Plumbing & Sewer Network (hydraulicNetwork.ts): Aquaeductus Trunk, Castellum Divisorium, Fistulae Supply Pipes & Cloaca Maxima Sewer Trenches/Outfalls (URBS-HYDRAULIC-01..04)”

Goal: Create hydraulicNetwork.ts integrating the AQVAE MCP aqueduct directly into an upper-terrace Castellum Divisorium header tank that feeds a SimCity-style Roman hydraulic plumbing (Fistulae twin-rail bronze pipes & sluice valves) and Cloaca Maxima arched stone sewer trench network across all district terraces, terminating in cliff-base Cloaca sea outfalls and stepped cascades (Cataracta). Requirements: URBS-HYDRAULIC-01, URBS-HYDRAULIC-02, URBS-HYDRAULIC-03, URBS-HYDRAULIC-04 Plans:

  • 292-01: Create hydraulicNetwork.ts implementing integrated Aquaeductus trunk, Castellum Divisorium, Fistulae supply pipes, Cloaca Maxima sewer trenches, and sea outfalls (phases/292-integrated-roman-hydraulic-and-sewer-network/292-01-PLAN.md)

Phase 293: Multi-Channel Color + Geometric Shape Animated Activity Conduits (AQUA PURA Chevrons, IGNIS OPERIS Diamonds, CLOACA FAULT Triangles & VIA VICTORIAE Seals) (URBS-CONDUIT-ANIM-01..04)

Section titled “Phase 293: Multi-Channel Color + Geometric Shape Animated Activity Conduits (AQUA PURA Chevrons, IGNIS OPERIS Diamonds, CLOACA FAULT Triangles & VIA VICTORIAE Seals) (URBS-CONDUIT-ANIM-01..04)”

Goal: Animate live activities in progress along the Roman hydraulic and sewer conduits using 4 double-coded Color + Geometric Shape channels: Crystalline Cyan Chevron Wave-Crests (❯❯) for AQUA PURA MCP/provider supply, Molten Gold Diamond Ingots (◆◆) for IGNIS OPERIS active agent work-in-progress, Crimson Warning Triangles (▲) & eddies (◎) for CLOACA FAULT error backpressure, and Emerald Circular Laurel Medallions (◉) for VIA VICTORIAE verified CI/Git flows. Requirements: URBS-CONDUIT-ANIM-01, URBS-CONDUIT-ANIM-02, URBS-CONDUIT-ANIM-03, URBS-CONDUIT-ANIM-04 Plans:

  • 293-01: Implement drawAnimatedHydraulicActivityFlows in hydraulicNetwork.ts and wire all 4 shape+color activity conduits into WorldRenderer (phases/293-animated-shape-color-activity-conduits/293-01-PLAN.md)

Phase 294: Diegetic Hydraulic Sluice-Gate Callouts, Semiotic Consistency Audit, Unit Test Suite & Playwright CV Flywheel Convergence (URBS-SEMIOTICS-CV-01..04)

Section titled “Phase 294: Diegetic Hydraulic Sluice-Gate Callouts, Semiotic Consistency Audit, Unit Test Suite & Playwright CV Flywheel Convergence (URBS-SEMIOTICS-CV-01..04)”

Goal: Replace ungrounded screen-space straight dependency lines and the floating yellow arrow with diegetic Roman Hydraulic Sluice-Gate (Cataracta) callouts and isometric conduit routing, add a shape+color hydraulic flow legend, write comprehensive unit tests (hydraulicNetwork.test.ts, pbrMaterials.test.ts, lightingAndFx.test.ts), and verify convergence via the Playwright @ngi/platform-ssot CV Flywheel. Requirements: URBS-SEMIOTICS-CV-01, URBS-SEMIOTICS-CV-02, URBS-SEMIOTICS-CV-03, URBS-SEMIOTICS-CV-04 Plans:

  • 294-01: Implement diegetic hydraulic sluice-gate callouts, shape+color conduit legend, hydraulicNetwork.test.ts unit suite, and Playwright CV Flywheel verification (phases/294-semiotic-consistency-and-cv-convergence/294-01-PLAN.md)

PhaseTitlePlansStatus
Phase 289Smooth Ground-Plane-Recessed Lighting, 2D-Feathered Elliptical Sun-Shafts & Periodic Seamless Sea Synthesis1/1Complete
Phase 290High-Detail Terrace Grounds: Orthogonal Cardo/Decumanus Streets, Opus Sectile Marble Plazas, Vaulted Substructio Cliff Arcades & Balustrades1/1Complete
Phase 291High-Detail Vitruvian Building Architecture (T0–T3): Rusticated Ground Arcades (Fornices), Belt Cornices, Atrium Impluvia & Acroteria Roofs1/1Complete
Phase 292Integrated Roman Hydraulic Plumbing & Sewer Network (hydraulicNetwork.ts): Aquaeductus Trunk, Castellum Divisorium, Fistulae Supply Pipes & Cloaca Maxima Sewer Trenches/Outfalls1/1Complete
Phase 293Multi-Channel Color + Geometric Shape Animated Activity Conduits (AQUA PURA Chevrons, IGNIS OPERIS Diamonds, CLOACA FAULT Triangles & VIA VICTORIAE Seals)1/1Complete
Phase 294Diegetic Hydraulic Sluice-Gate Callouts, Semiotic Consistency Audit, Unit Test Suite & Playwright CV Flywheel Convergence1/1Complete

Phases (v5.4 — T3 CI/TOESTUB Enforcement Deep-Extraction & Harness Policy Hardening)

Section titled “Phases (v5.4 — T3 CI/TOESTUB Enforcement Deep-Extraction & Harness Policy Hardening)”

Phase 301: Change-Gated CI Path Map & vox ci check-changed (T3-CI-CHANGED-01..04)

Section titled “Phase 301: Change-Gated CI Path Map & vox ci check-changed (T3-CI-CHANGED-01..04)”

Goal: Port t3code check:changed + INERT/EXTRA_OWNERS ownership mapping into a Vox contract and CLI so local/CI runs only the gates owned by the diff. Requirements: T3-CI-CHANGED-01, T3-CI-CHANGED-02, T3-CI-CHANGED-03, T3-CI-CHANGED-04 Success criteria:

  1. Owner map contract validates under schema and lists crates/UI/workflows/contracts.
  2. vox ci check-changed (or vox run scripts/check-changed.vox) exits 0 on inert planning-only diffs without invoking cargo.
  3. PR jobs skip heavy work for inert path sets while required contexts remain honest.
  4. Docs in runner-contract.md show the local command and ownership rules.

Plans:

  • 301-01: Define owner map + implement check-changed + CI wiring + docs/tests

Phase 302: Transfer-Budget Artifacts, PR Comments & Weighted Test Sharding (T3-CI-BUDGET-01..04)

Section titled “Phase 302: Transfer-Budget Artifacts, PR Comments & Weighted Test Sharding (T3-CI-BUDGET-01..04)”

Goal: Promote transfer-budget coverage into CI artifacts + trusted PR comments, and balance shards with duration weights. Requirements: T3-CI-BUDGET-01, T3-CI-BUDGET-02, T3-CI-BUDGET-03, T3-CI-BUDGET-04 Success criteria:

  1. CI uploads a schema-valid budget artifact on PRs that touch orchestrator-mcp/budget code.
  2. Comment publisher runs from default-branch code only and posts delta vs main.
  3. Weighted shard file measurably reduces max-shard wall time vs hash-only partitions.
  4. Ratchet report lands under contracts/reports/.

Plans:

  • 302-01: Budget artifact + trusted comment workflow + weighted sharding + ratchet

Phase 303: Knip Dual Gate, PR Size Labels & Evidence-Asset Ban (T3-CI-HYGIENE-01..04)

Section titled “Phase 303: Knip Dual Gate, PR Size Labels & Evidence-Asset Ban (T3-CI-HYGIENE-01..04)”

Goal: Catch unused TS exports, label PR blast radius by non-test LOC, and ban committed evidence spam. Requirements: T3-CI-HYGIENE-01, T3-CI-HYGIENE-02, T3-CI-HYGIENE-03, T3-CI-HYGIENE-04 Success criteria:

  1. Knip (or equivalent) fails CI on new unused exports in vox-gui/ui outside allowlist.
  2. PRs receive exactly one size:* label from the XS–XXL set based on non-test LOC.
  3. Committing a disallowed evidence binary fails the hygiene gate.
  4. Mutation tests prove each gate fails closed; AGENTS/CONTRIBUTING updated.

Plans:

  • 303-01: Knip config + PR size workflow + evidence ban + docs/mutation tests

Phase 304: Agent-Ops Kill-Safety, Hit-Every-Surface Checklist & Worktree Isolation (T3-AGENT-OPS-01..04)

Section titled “Phase 304: Agent-Ops Kill-Safety, Hit-Every-Surface Checklist & Worktree Isolation (T3-AGENT-OPS-01..04)”

Goal: Import t3code agent-ops culture: kill-safety, multi-surface completion checklist, isolated worktree homes, suppression human review. Requirements: T3-AGENT-OPS-01, T3-AGENT-OPS-02, T3-AGENT-OPS-03, T3-AGENT-OPS-04 Success criteria:

  1. AGENTS.md / governance names the kill-safety triad with Vox paths.
  2. Checklist is referenced from PR template or vox doctor advisory for multi-surface diffs.
  3. Worktree setup marker + isolated home documented and scripted via .vox.
  4. Suppression-only PRs cannot merge without human-review label/CODEOWNERS.

Plans:

  • 304-01: Policy docs + checklist advisory + worktree isolation + suppression review gate

Phase 305: TOESTUB Thin-Transport, Suppression Review Gate & Snapshot Interning (T3-TOESTUB-ENF-01..04)

Section titled “Phase 305: TOESTUB Thin-Transport, Suppression Review Gate & Snapshot Interning (T3-TOESTUB-ENF-01..04)”

Goal: Close the oxlint→TOESTUB gap for transport bypasses, suppression reasons, snapshot interning, and remaining high-value rules. Requirements: T3-TOESTUB-ENF-01, T3-TOESTUB-ENF-02, T3-TOESTUB-ENF-03, T3-TOESTUB-ENF-04 Success criteria:

  1. New detector flags raw RPC/client bypass outside allowlisted files.
  2. Unreasoned suppressions in a PR diff fail CI.
  3. Snapshot/message-ID interning is live-path covered by tests.
  4. Remaining oxlint rules are either detectors or explicit deferred ledger rows.

Plans:

  • 305-01: Detectors + suppression ratchet + interning + oxlint mapping ledger

Phase 306: Tauri Fingerprint Advisory, GUI Cache Layers, Surface Graph Refresh & Scorecard (T3-CI-CLOSE-01..04)

Section titled “Phase 306: Tauri Fingerprint Advisory, GUI Cache Layers, Surface Graph Refresh & Scorecard (T3-CI-CLOSE-01..04)”

Goal: Close the milestone with fingerprint labeling, legal GUI caches, a fresh T3 surface graph, and a CI-enforcement scorecard. Requirements: T3-CI-CLOSE-01, T3-CI-CLOSE-02, T3-CI-CLOSE-03, T3-CI-CLOSE-04 Success criteria:

  1. Native/sidecar-affecting PRs get an advisory label; check stays green.
  2. GUI jobs restore layered pnpm/vite caches without PR cache saves.
  3. Surface graph regenerates from current clone SHAs; inventory script compiles.
  4. Scorecard + architecture note list every gem as Shipped/Partial/Deferred with anchors.

Plans:

  • 306-01: Fingerprint workflow + cache layers + graph regen + scorecard/docs

PhaseTitlePlansStatus
Phase 301Change-Gated CI Path Map & vox ci check-changed1/1Complete
Phase 302Transfer-Budget Artifacts, PR Comments & Weighted Test Sharding1/1Complete
Phase 303Knip Dual Gate, PR Size Labels & Evidence-Asset Ban1/1Complete
Phase 304Agent-Ops Kill-Safety, Hit-Every-Surface Checklist & Worktree Isolation1/1Complete
Phase 305TOESTUB Thin-Transport, Suppression Review Gate & Snapshot Interning1/1Complete
Phase 306Tauri Fingerprint Advisory, GUI Cache Layers, Surface Graph Refresh & Scorecard1/1Complete

Phases (v5.3 — Photorealistic Edge-Softening, Continuous Monumental Aqueduct-to-Sewer Architecture, Sub-Surface Ground Integration & CV Flywheel Score Maximization)

Section titled “Phases (v5.3 — Photorealistic Edge-Softening, Continuous Monumental Aqueduct-to-Sewer Architecture, Sub-Surface Ground Integration & CV Flywheel Score Maximization)”

Phase 295: Feathered Coastal Bathymetry, Smooth Macro-Shaded PBR Materials & Laplacian Noise Suppression (URBS-EDGE-PBR-01..04)

Section titled “Phase 295: Feathered Coastal Bathymetry, Smooth Macro-Shaded PBR Materials & Laplacian Noise Suppression (URBS-EDGE-PBR-01..04)”

Goal: Eliminate hard diamond coastal cutouts, 1px Laplacian PBR pixel chatter (noiseFloor: 10.7), and #1a1613 cartoon ink outlines by introducing feathered elliptical radial coastal shoals, smooth bilinear/radial macro-shaded PBR stone/terracotta/water tiles, full-canvas diagonal ocean swells, and soft tonal masonry/roof edge bevels. Requirements: URBS-EDGE-PBR-01, URBS-EDGE-PBR-02, URBS-EDGE-PBR-03, URBS-EDGE-PBR-04 Plans:

  • 295-01: Implement feathered coastal bathymetry, low-Laplacian macro-shaded PBR tiles, diagonal ocean swells, and soft tonal edge bevels (phases/295-feathered-bathymetry-and-smooth-pbr/295-01-PLAN.md)

Phase 296: True Isometric Molded Curtain Walls, Seamless Via Silicea Street Grid & Coffered Opus Sectile Forums (URBS-TERRACE-CRAFT-01..04)

Section titled “Phase 296: True Isometric Molded Curtain Walls, Seamless Via Silicea Street Grid & Coffered Opus Sectile Forums (URBS-TERRACE-CRAFT-01..04)”

Goal: Replace jagged axis-aligned 2D fillRect wall merlons and overlapping street alpha rectangles with true 2:1 isometric molded travertine parapet capstones (Corona), 3D pyramidal-roofed corner bastions, a unified intersection-free Cardo/Decumanus street cross-plaza (Via Silicea), a recessed multi-ring Opus Sectile Forum with 4 corner Corinthian columns, and an overhanging cliff-edge travertine cornice (Geison). Requirements: URBS-TERRACE-CRAFT-01, URBS-TERRACE-CRAFT-02, URBS-TERRACE-CRAFT-03, URBS-TERRACE-CRAFT-04 Plans:

  • 296-01: Upgrade curtain walls, Cardo/Decumanus street intersections, Opus Sectile forum peristyles, and cliff cornices (phases/296-isometric-walls-and-seamless-streets/296-01-PLAN.md)

Phase 297: Cylindrical & Coffered 3D Architectural Realism for Wonders & T0–T3 Insulae (URBS-ARCH-REALISM-01..04)

Section titled “Phase 297: Cylindrical & Coffered 3D Architectural Realism for Wonders & T0–T3 Insulae (URBS-ARCH-REALISM-01..04)”

Goal: Upgrade Tabularium and Pantheon rotundas with multi-stop cylindrical normal shading and coffered half-dome ribs (Lacunaria), replace flat 2D window squares with isometric-skewed recessed stone window reveals and warm interior lamplight, add 2-stepped stylobates and sculpted pediment tympanums to T2/T3 buildings, and upgrade peripheral landmarks (AQVAE, CASTRVM, PORTVS, PORTA) to 3D isometric masonry. Requirements: URBS-ARCH-REALISM-01, URBS-ARCH-REALISM-02, URBS-ARCH-REALISM-03, URBS-ARCH-REALISM-04 Plans:

  • 297-01: Upgrade cylindrical rotunda wonders, isometric window reveals, pediment tympanums, and 3D peripheral landmarks (phases/297-cylindrical-wonders-and-recessed-windows/297-01-PLAN.md)

Phase 298: Continuous 3D Monumental Aqua Claudia Viaduct & Ground-Integrated Beveled Hydraulic/Sewer Grid (URBS-PLUMBING-3D-01..04)

Section titled “Phase 298: Continuous 3D Monumental Aqua Claudia Viaduct & Ground-Integrated Beveled Hydraulic/Sewer Grid (URBS-PLUMBING-3D-01..04)”

Goal: Connect AQVAE MCP via a continuous 3D multi-arch stone aqueduct viaduct (Aqua Claudia) directly into the Acropolis Castellum Divisorium, upgrade drawConduitTrack with rounded isometric elbow fillets (quadraticCurveTo) and 3D cylindrical pipe shading, route all dependency and blast-radius conduits on the terrace ground / viaduct layer beneath 3D buildings (Layer 0.85), and render stepped stone Cataracta cascades between terrace elevations. Requirements: URBS-PLUMBING-3D-01, URBS-PLUMBING-3D-02, URBS-PLUMBING-3D-03, URBS-PLUMBING-3D-04 Plans:

  • 298-01: Build continuous 3D Aqua Claudia aqueduct bridge, rounded pipe elbow fillets, sub-building ground conduit routing, and stepped Cataracta cascades (phases/298-continuous-aqueduct-and-beveled-plumbing/298-01-PLAN.md)

Phase 299: Illuminated Halo-Backed Shape+Color Activity Particles & Interactive Conduit Semiotics (URBS-FLOW-SEMIOTICS-01..04)

Section titled “Phase 299: Illuminated Halo-Backed Shape+Color Activity Particles & Interactive Conduit Semiotics (URBS-FLOW-SEMIOTICS-01..04)”

Goal: Add soft radial glow halos and tapered comet trails to all 4 geometric particle channels (❯ Chevron, ◆ Diamond, ▲ Triangle, ◉ Laurel Seal), scale conduit gauge and particle cadence by live activity load, render glowing bronze manifold rings (Epistomia) at district street ports, and upgrade the top-left Semiotic Legend Pill with vector geometric icons and live channel counts. Requirements: URBS-FLOW-SEMIOTICS-01, URBS-FLOW-SEMIOTICS-02, URBS-FLOW-SEMIOTICS-03, URBS-FLOW-SEMIOTICS-04 Plans:

  • 299-01: Add radial halos and comet tails to activity particles, dynamic load scaling, perimeter manifold rings, and vector-icon legend (phases/299-halo-particles-and-conduit-semiotics/299-01-PLAN.md)

Phase 300: Comprehensive Unit Tests, Semiotic Critique Audit & Playwright CV Flywheel Score Maximization (URBS-CV-MAX-01..04)

Section titled “Phase 300: Comprehensive Unit Tests, Semiotic Critique Audit & Playwright CV Flywheel Score Maximization (URBS-CV-MAX-01..04)”

Goal: Expand unit test suites (pbrMaterials.test.ts, lightingAndFx.test.ts, hydraulicNetwork.test.ts, worldRenderer.test.ts), achieve positive CV noise reduction (noiseFloor < 5.1) and higher visualAfterScore (>= 0.942) on urbs-civilization in e2e/review/cv-flywheel.spec.ts, verify 0 regressions across all 32 GUI surfaces, and update the visual blueprint carousel. Requirements: URBS-CV-MAX-01, URBS-CV-MAX-02, URBS-CV-MAX-03, URBS-CV-MAX-04 Plans:

  • 300-01: Expand unit tests, run Playwright CV Flywheel score maximization, and update visual blueprint (phases/300-cv-flywheel-score-maximization/300-01-PLAN.md)

PhaseTitlePlansStatus
Phase 295Feathered Coastal Bathymetry, Smooth Macro-Shaded PBR Materials & Laplacian Noise Suppression1/1Complete
Phase 296True Isometric Molded Curtain Walls, Seamless Via Silicea Street Grid & Coffered Opus Sectile Forums1/1Complete
Phase 297Cylindrical & Coffered 3D Architectural Realism for Wonders & T0–T3 Insulae1/1Complete
Phase 298Continuous 3D Monumental Aqua Claudia Viaduct & Ground-Integrated Beveled Hydraulic/Sewer Grid1/1Complete
Phase 299Illuminated Halo-Backed Shape+Color Activity Particles & Interactive Conduit Semiotics1/1Complete
Phase 300Comprehensive Unit Tests, Semiotic Critique Audit & Playwright CV Flywheel Score Maximization1/1Complete