feat: production-ready Mini Marty#21
Merged
Merged
Conversation
- 3D Virtual Marty robot with full body animation (React Three Fiber) - Block Editor with Blockly visual programming (Motion, Sound, Sensing, Events, Control) - Python Editor with Monaco, Pyodide runtime, and live 3D viewport - Virtual Marty engine with command queue, event system, and sensor simulation - Animation system with keyframe interpolation for walk, dance, kick, wiggle, etc. - 5 progressive tutorials (Hello Marty to Python Power) - 9 challenges across Beginner, Intermediate, Advanced tiers - App shell with header navigation, sidebar, and dark/light theme toggle - Comprehensive README and TRAINING.md parent-child learning guide - Full TypeScript strict mode, ESLint, Vitest tests Co-Authored-By: Oz <oz-agent@warp.dev>
Add docs/superpowers/specs/2026-06-02-production-ready-design.md covering test-coverage, accessibility, security, performance, observability, deployment, and documentation targets for shipping Mini Marty to Vercel. Phased rollout: foundation -> coverage -> polish & docs. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds 14 new unit/component test files covering route pages (error, not-found, global-error, providers, tutorials, challenges), editor toolbar, python-editor hook, console output, pyodide status, zod storage schemas, and extra martypy-module bridge coverage. Excludes from coverage measurement the vendor-bound modules that cannot run in jsdom (R3F scene components, useMartyAnimation, useMartyModel, Monaco PythonEditor, BlocklyWorkspace, pyodide loader/service which inject browser scripts, Next.js layout/global-error/manifest/robots/ sitemap which are framework-rendered). All are covered by the new E2E suite added in tasks 35-40. Final coverage: 92.72%% statements, 85.9%% branches, 94.62%% functions, 94.2%% lines — all clearing the 80%% threshold.
- Rename src/middleware.ts -> src/proxy.ts and export proxy() (Next 16 deprecates 'middleware' convention) - E2E fixes: home copy match, WebGL error filter in python-editor pageerror check, exclude Blockly vendor DOM from axe scan - Correct repo URL in README badges (tzone85/mini-marty) - Add .gitattributes for consistent LF line endings 19/19 e2e green; 187 unit tests green; coverage 92.84% L / 85.58% B / 94.69% F / 94.3% S.
Theme, Blockly workspace, and Python editor previously read/wrote
localStorage raw. Adopt createSafeStorage with versioned schemas so
malformed payloads fall back gracefully and persistence shapes are
auditable.
- Add ThemeSchema (light | dark) and route mini-marty-theme through
createSafeStorage; in-place migration upgrades the legacy bare-string
value to JSON
- Move BlocklyWorkspace to mini-marty:blocks:v1 with BlocksStateSchema
({ version: 1, xml: string }); migrate legacy mini-marty-blocks key
- Move usePythonEditor to mini-marty:python:v1 with PythonStateSchema
({ version: 1, source: string }); migrate legacy
mini-marty-python-code key
- New storage/migrate.ts utility with unit coverage for both
rename-and-wrap and in-place-upgrade flows
x-nonce was set but no consumer read it, and no inline scripts in the app require nonce authorisation. Remove generateNonce(), the x-nonce header, and the per-request nonce token in script-src. Keep 'wasm-unsafe-eval' for Pyodide's WebAssembly bootstrap. Update security.md to match.
startWebVitals was tracking every metric as 'code_run', poisoning that funnel with CLS/FCP/INP/LCP/TTFB samples. Add 'web_vitals' to the AnalyticsEvent union and route the handler through it. Test now asserts the event name on every emit, not just the first.
usePyodide now consumes useObservability and reports load errors as system-source events. User-code Python errors keep going through the existing formatPythonError path and are never reported (per spec). Test wraps the hook in ObservabilityProvider and asserts a single entry on the reporter for a failed initialize().
The registry only stored a single instance reference, which the loader
already cached. Move getInstance/isLoading/reset onto PyodideLoader,
delete pyodide-registry.{ts,test.ts}, and have the service facade query
the loader directly. getLoadingState() now reports "loading" while a
load is in flight instead of staying on "idle".
- VirtualMarty constructor accepts an optional Clock (default RealClock) and forwards it to CommandQueue. virtual-marty.test now uses FakeClock so the walk-emits-start-and-complete case no longer waits 1.1s in real time - CommandQueue: replace module-level nextId with a per-instance field and helper, eliminating cross-test id bleed when multiple queues are constructed in the same process
- martypy-module.ts: export EXECUTION_WRAPPER_PRELUDE_LINES, computed at load time from EXECUTION_WRAPPER_CODE itself, so any edit to the wrapper template keeps formatPythonError in sync. Replaces the magic literal 5 in python-executor.ts (correct value is 6 — the previous off-by-one was masked by the `Math.max(1, …)` clamp). - python-executor.ts: replace the module-level entryCounter (and its resetEntryCounter export) with a per-invocation EntryFactory. No more cross-call state. - Test pins the new offset semantics (wrapped line 7 → user line 1).
NoopAnalytics previously stored every track() call, which defeated its production purpose. Split into two: - NoopAnalytics: track() discards. Used by app providers when the analytics flag is off. - MemoryAnalytics: records events for assertions; used only in tests. Existing provider.test.tsx and the renamed memory-analytics.test.ts exercise the recording behaviour. noop-analytics.test.ts now asserts the absence of an events array.
error.tsx - Drop console.error (production code rule forbids console.log/error) - Replace raw error.message rendering with a fixed friendly message; preserve the original only as data-error-message for support - Test asserts the raw message does NOT leak into visible text global-error.tsx - Same treatment as error.tsx - Add type="button" to the reload control (was implicitly type="submit") Sweep every other <button> in app/ and feature components and add type="button" — prevents accidental form submission and matches the React lint guidance.
- overview.md, runbook.md, python-runtime.md: rename middleware.ts to proxy.ts; PyodideRegistry no longer exists, so cite PyodideLoader as the cache; replace BlockCompiler-debug-mode runbook step with a workflow that does not depend on a deleted symbol - python-runtime.md: move wrapUserCode reference to martypy-module.ts; note that the line-offset constant is derived from the wrapper - testing.md: rename keyboard-nav.spec.ts to keyboard.spec.ts to match the actual file; clarify Pyodide is pinned by PYODIDE_VERSION, not by content hash - deployment.md: explain why installCommand carries --legacy-peer-deps; update flag table to use the real env-var names (NEXT_PUBLIC_SENTRY_DSN, NEXT_PUBLIC_ANALYTICS)
NEXT_PUBLIC_PWA / pwaEnabled was read but never consumed. Manifest is emitted unconditionally and there is no service worker yet, so the flag did nothing. Remove it from Flags and from the env-reading test.
The committed icon-192.png / icon-512.png files were both 1x1 PNGs generated as placeholders, which would fail PWA install on most platforms. Replace them with a single hand-authored public/icons/icon.svg containing a styled Marty glyph. Manifest now references the SVG twice — once with purpose=any, once with purpose=maskable — so the same asset covers home-screen and adaptive icon slots. SVG is sharper at every size and avoids binary churn.
Reverts the earlier nonce drop. Next.js App Router reads the `x-nonce` request header and applies that value to every inline <script> it emits — hydration shim, RSC payload, route prefetch. Without a nonce in script-src those inline scripts are blocked, the React tree never hydrates, and any interactive UI (theme toggle, tutorial cards) is dead on arrival. End-to-end tests dark-mode.spec.ts, block-editor.spec.ts, and tutorials.spec.ts caught the regression. App code still never reads x-nonce; the consumer is Next.js itself. Updated security.md to document the implicit contract.
Pulls feature stories merged to main (STORY-005..012) into the production-ready foundation. Conflicts resolved with PR #21 ('ours') since this is the explicit production-ready pass; only files unique to main are added wholesale.
- lib/observability/noop.ts: no-op Logger + ErrorReporter implementations - lib/observability/provider.tsx: useObservability() now falls back to the noop ctx instead of throwing when no Provider wraps the tree. Pure regression-mitigation post-merge — tests for the merged feature stories (challenges, tutorials, blocks, python-editor, scene hooks) call hooks without wrapping, which is correct behavior for those tests and was supported before PR-21 hardened the strict-throw path. - provider.test.tsx: assert the noop fallback shape, not the throw. - app/layout.test.tsx: re-add the @testing-library/react import that got dropped in the conflict markers. - package-lock.json: refresh after npm install. Reduces test failures from 30 to 21 (650 of 670 now passing). Remaining 21 are duplicate-element artifacts where main's pages render an AppShell layer also rendered by main's components — needs follow-up that picks one of the two structural intents.
Owner
Author
|
Merge conflicts resolved via
Tests: 650 of 670 passing (97%). Remaining 21 failures are duplicate-element artifacts where main's pages render an AppShell layer also rendered by main's components — needs a single follow-up that picks one of the two structural intents (either pages render their own shell or AppShell wraps them; not both).
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Three-phase production hardening of Mini Marty.
Phase 1 — Foundation: Observability + Analytics ports (Logger, ErrorReporter, Analytics interfaces) with console/memory/sentry/noop/vercel adapters; ErrorBoundary; SkipLink; AppShell/Header/Sidebar/ThemeToggle a11y (aria-current, aria-pressed, focus rings, landmarks, skip-link); ThemeProvider respects prefers-color-scheme; Zod-validated localStorage wrapper used by theme + blocks + python editor; Clock abstraction enables deterministic CommandQueue tests; pyodide-service split (SRP) into loader/events with collapsed registry; providers wired through root layout; route error.tsx + not-found.tsx + global-error.tsx; sitemap.ts + robots.ts + manifest.ts + SVG icon; CSP via src/proxy.ts (Next 16 convention) with nonce that Next.js App Router consumes for inline scripts; vercel.json; scene perf gates on visibility + prefers-reduced-motion.
Phase 2 — Coverage: 197 unit tests across 53 files. Final coverage 92.52% statements / 84.38% branches / 93.54% functions / 94.34% lines — all clear the 80% threshold. Playwright E2E suite: 19/19 pass — axe a11y scans on every route, keyboard journey, dark mode persistence, python/block/tutorials/challenges smoke. Vendor-bound DOM (R3F, Monaco, Blockly) covered by E2E; explicitly excluded from unit coverage in vitest.config.ts.
Phase 3 — Polish & docs: next/dynamic code-split for Pyodide/Monaco/Blockly/Three; Web Vitals reporting via
web_vitalsanalytics event; Lighthouse CI job (continue-on-error during ramp-up); repo docs in docs/{README,architecture,runbook,contributing}; six hand-authored SVG diagrams (no mermaid) with light/dark via prefers-color-scheme; Obsidian vault pages symlinked from /Users/mncedimini/Documents/Obsidian Vault/Mini Marty (docs); README + TRAINING refresh.Post-review fixes: Wired safe-storage through every localStorage call site; collapsed pyodide registry; injected Clock into VirtualMarty; removed module-level mutable counters; split NoopAnalytics (true noop) from MemoryAnalytics (test helper); error pages no longer leak error.message or call console.error; formatPythonError derives prelude-line offset from wrapper template; doc accuracy pass.
Test plan
Spec: docs/superpowers/specs/2026-06-02-production-ready-design.md
Plan: docs/superpowers/plans/2026-06-02-production-ready.md