ods-pages

ODS TODO

Companion to REGRESSION_LOG.md (test batches + bugs found). Loose rules: drop things in freely, re-sort when you revisit. Link to code paths the same way REGRESSION_LOG does so the list doubles as a jump-table.

---

Now — actively being worked on

• No focused initiative. Day-to-day maintenance mode: when a new cross-framework bug comes up, add a contract-first conformance scenario that pins it (CONTRIBUTING.md → “Conformance scenarios — contract-first”). Conformance count: 27 scenarios × 2 drivers = 54 parity tests. Test gate (Flutter 904 + React 1180+ + conformance + coverage thresholds + Stryker weekly) is solid.

Next — next 1–2 sessions

• Flutter chat panel for Edit-with-AI — ADR-0003’s chat mode is React-only today; one-shot is on both. Mirror the React ChatEditFlow in Flutter (message list + AbortController-style cancellation + <spec>...</spec> proposal protocol + per-card Apply / Discard). Reuses existing ai_provider.dart + ai_edit_prompt.dart; needs a sibling ai_chat_prompt.dart (mirror of ai-chat-prompt.ts) and a stateful chat screen. Closes the last big parity gap from ADR-0003 §5 (“Flutter feature parity for chat mode”).

Docs — priority 3 (pre-public polish)

• CHANGELOG.md — low value until releases start; relevant once the conformance suite pins spec versions.

Later — important, not urgent

• recordSource default-order parity — surfaced by s23 (record navigation). PocketBase defaults to created desc, so FakeDataService.query reverses results to mimic that; SQLite’s default scan returns insertion order. Same spec → different “first record” across frameworks. The s23 scenario currently sidesteps this with order-agnostic assertions. Real fix: add an optional sort directive to the firstRecord action (and/or pick a single canonical default) so firstRecord is deterministic for a given spec on every renderer.
• Conformance: deepen existing narrow scenarios — a few scenarios pin only initial state or config, not behavior: - s17 tabs: only asserts initial active state; add a clickTab(label) primitive + test that switching tabs changes which is active, and that the snapshot only exposes the active tab’s content (if that’s the actual contract). - s18 chart: only asserts config propagation; no aggregation math. Add a variant that seeds rows, snapshots the chart’s series values (would need seriesCount / seriesData in the snapshot, not just a hardcoded 1).
• Conformance: scenario catalog parity check — the scenario list (ids, names, capabilities) is duplicated between scenarios.ts and scenarios.dart. Extract to conformance/scenarios.json (just the metadata; run bodies stay per-language); add a meta-test on each side that asserts “every scenario id in the catalog has a scenario implementation.” Prevents the two sides from silently drifting in which scenarios they advertise.
• Generator code honors SettingsStore — code_generator.dart still uses getApplicationDocumentsDirectory() directly in its emitted code (see code_generator.dart:475). Low priority because the generated app runs on someone else’s machine, but worth revisiting for consistency.
• OS-keychain storage for AI API keys — ADR-0003 v1 stores the key plaintext (localStorage on React, settings.json on Flutter). Move to OS-keychain (keytar-equivalent on web is a non-starter, so React likely stays in localStorage or moves to a server proxy; Flutter has flutter_secure_storage). Wants its own ADR — different APIs per platform, different web/desktop story, masking-on-display already in place either way.
• AI cost warning before send — estimateCost exists on both providers (ai-provider.ts / ai_provider.dart) and is unit-tested, but neither screen surfaces the estimate before firing the request. Add a configurable threshold (default ~$0.10) — over it, show “this request is estimated at $X, send anyway?” before the call. Open question per ADR-0003 §5.
• PII redaction before AI send — seedData on a data source can carry real names / emails. Currently we send the spec verbatim. Add an opt-in toggle in AI Settings that runs a lightweight redactor over seedData[] arrays before the prompt is built (replace strings matching email / phone shapes with placeholders). Default off; surface the disclosure inline in the AI Settings panel either way. Open question per ADR-0003 §5.

Wishlist — ideas; not scheduled

• Mutation testing — Dart side — Stryker for React landed (2026-04-26: Frameworks/react-web/stryker.config.json, weekly .github/workflows/mutation.yml, npm run mutation). Mirror with mutation_test package on the Flutter side once the React baseline stabilises and we know whether it’s catching real gaps.
• Visual regression tests for theme rendering. Playwright snapshots or Percy. CLAUDE.md flags theming as a historical pain point; worth it once the API churn settles.
• Property-based tests — formula evaluator + further parser coverage — first slice landed 2026-04-26: spec-parser totality / soundness / minimal-valid in tests/unit/parser/spec-parser-properties.test.ts (fast-check) and the Dart mirror in test/parser/spec_parser_properties_test.dart (seeded RNG). Next: round-trip properties for the formula evaluator and aggregate evaluator.
• Accessibility tests — @axe-core/playwright is already a devDep, tests/e2e/accessibility/ is empty. A pass on key flows (login, dashboard, app home) would catch a lot cheaply.
• Third renderer — Swift/SwiftUI mobile-native, or a terminal-UI renderer for scripting/piping. Drives real pressure on the conformance driver contract (Path B).
• GitHub Issues migration — when the ODS spec goes public, the “drop it in TODO.md” loop becomes the wrong interface for external contributors. Move Later/Wishlist items to Issues, pin a Roadmap, keep Now/Next here.
• Public conformance badge — once Path B lands + the contract stabilizes, publish ods-conformance and let 3rd-party frameworks self-certify. Ecosystem play.
• Streaming AI responses — both providers support SSE-style token streaming; v1 of ADR-0003 waits for the full response. Better UX on long generations (especially full-spec rewrites in chat mode) at the cost of chunk parsing + cancellation plumbing in both AnthropicProvider and OpenAiProvider. ADR-0003 §5 open question; gated on whether users actually complain about latency.
• Tool-use / function-calling for AI — instead of “reply with the whole JSON spec”, ask the model to call structured tools (addField, renameDataSource, etc.). Cuts output tokens on large specs and gives us a typed surface to validate before apply. Cost: provider-specific tool schemas (Anthropic + OpenAI diverge here), more round-trips, the proposal protocol gets more complex than <spec>...</spec>. ADR-0003 §5 open question.
• Cross-session AI conversation persistence — chat history lives in component state today and dies on close. Persist per app (PocketBase collection on React, SQLite table on Flutter) so a builder can resume a long-running conversation. Storage + privacy questions (the spec gets pasted into every turn so the history accumulates copies of every revision). ADR-0003 §5 open question.
• AI cost-ceiling hard limit — v1 only has the threshold warning (see Later). A hard per-session cap that blocks further requests after $N spent would protect runaway iteration loops. Likely overkill until someone actually burns through their key. ADR-0003 §5 open question.

---

Done — recent (trim quarterly)

2026-04-28 — ADR-0003 (AI Build Helper, BYO API key) shipped phases 1–6

• Phase 1: provider abstraction — ai-provider.ts / ai_provider.dart with AnthropicProvider (Messages API) + OpenAiProvider (Chat Completions). Pure HTTP, transport-injected, 38 unit tests + 8 properties (~240 random inputs).
• Phase 2: AI Settings panel on both frameworks (provider radio + masked key + model dropdown + Test connection). React stores in localStorage (ods_ai_settings); Flutter stores in SettingsStore. jsdom localStorage shim added to vitest setup (the test environment was missing removeItem).
• Phase 3: React one-shot Edit-with-AI flow (textarea instruction → AI call → diff → Apply / Regenerate / Discard). EditWithAiScreen falls back to copy/paste when no key is configured. 12 prompt-builder tests cover ai-edit-prompt.ts.
• Phase 4: React multi-turn chat panel — ai-chat-prompt.ts adds the <spec>...</spec> proposal protocol; ChatEditFlow renders message bubbles + per-card Apply / Discard + AbortController-style Stop. Fixed a “Failed to execute ‘fetch’ on ‘Window’: Illegal invocation” regression by binding the default fetch impl to globalThis (regression test pinned).
• Phase 5: cross-framework conformance — ai:provider capability + AiRequestSnapshot + simulateAiRequest driver method on the OdsDriver contract; both drivers spin up the named provider against a fake transport and return the captured wire shape; s27 pins both endpoint URLs, both auth-header forms (x-api-key: vs authorization: Bearer), and the per-provider messages-array shape.
• Phase 6: Flutter mirror of one-shot edit — _EditWithAiScreen branches on SettingsStore.isAiConfigured (one-shot mode shows instruction → Generate → side-by-side before/after → Apply / Regenerate / Discard with spec validation; copy/paste stays as the no-key fallback). ai_edit_prompt.dart mirrors React’s helpers (12 unit tests).
• Build with AI menu fixed: split the misleading “Create New” (which was just a paste dialog) into a proper BuildWithAiScreen (one-shot AI flow when key configured, copy-prompt + paste-back fallback otherwise) plus a separate “Paste JSON” entry for importing existing specs.
• Conformance count: 26 → 27 scenarios × 2 drivers = 54 parity tests. Flutter test count: 865 → 904. ADR-0003 deferrals (chat panel on Flutter, OS-keychain, streaming, tool-use, cross-session persistence, cost ceiling, PII redaction) tracked in Next / Later / Wishlist above.

2026-04-26 — TDD discipline push (sessions 12+ commits)

• Conformance-first workflow doc in CONTRIBUTING.md and docs/testing.md — scenario goes in red on both drivers before implementation. ADR-0002 used as the worked example.
• 6 new conformance scenarios (s21–s26): theme config round-trip, detail fields, record navigation (firstRecord/nextRecord/onEnd), clickMenuItem, CURRENT_USER magic defaults, list defaultSort drives displayed row order. 3 of them surfaced real driver/parity gaps — driver methods added, gaps closed in the same flow. scenarios.ts.
• CI ↔ local parity tightened: workflows trigger on Frameworks/conformance/** (was a hidden path-filter gap); publish.sh adds npx tsc -b + flutter analyze so the static checks CI runs also fail locally. Conformance count bumped to 26 scenarios × 2 drivers = 52 parity tests.
• Coverage thresholds enforced locally + in CI: vitest per-folder (models 90%, parser 90%, engine 50%) plus a Dart tool/coverage_check.dart that parses lcov.info and enforces lib/engine 60% / lib/models 85% / lib/parser 80%. Both fail-non-zero on regression.
• Pure helper extraction: buildUpdatedSpecJson on both frameworks (theme-spec-writer.ts / theme_spec_writer.dart) replaces inline JSON munging in the admin save-to-spec path. 28 unit tests added (14 per framework).
• Property-based parser tests on both frameworks: fast-check for React (5 properties), seeded RNG for Flutter (5 properties) — totality / soundness / minimal-valid invariants pinned over ~600 random inputs.
• Flutter widget tests un-skipped + 42 in the gate. Diagnosed the multi-month “harness hang on Windows” as a FakeAsync-vs-sqflite_ffi interaction: the native-bridge timers sqflite_ffi schedules are intercepted by FakeAsync but never fired, so pumpAndSettle waits forever. Fix: bootEngineFor / disposeAllFor wrap setup in tester.runAsync to escape the FakeAsync zone; pumpAndSettle does fixed real-time pump rounds (16 × 100ms) instead. Plus a pumpUntilFound helper for cases where fixed timing isn’t enough. _test_harness.dart.
• Stale widget tests fixed: 7 tests had real bugs that compounded in the run-skip gap — SingleChildScrollView around a ListView (nested viewports), assertions for column headers / search inputs that the renderer hides under the “No data yet” empty state, a form referencing a formId not in the spec’s pages.
• Pre-commit + CI gate now runs Flutter widget tests too; total Flutter ~810 → 865.
• Eliminate widget-test flake: bumped pump rounds to 16 × 100ms (1.6s worst case) after a 5-run validation showed ~20% flake at the 8 × 75ms tuning. 5 / 5 consecutive full-gate runs now green.
• Mutation testing scaffolded (React side): Stryker config, npm run mutation, .gitignore updates, weekly .github/workflows/mutation.yml + workflow_dispatch. Threshold gate is break: 0 for now — switch to a real break threshold after 2-3 weekly runs settle.
• Process docs: regression-scenario rule in CONTRIBUTING (every bug fix lands with a failing-then-passing test) + .github/pull_request_template.md with a Test plan checklist.

2026-04-25 — ADR-0002 (Theme + Customizations redesign) shipped

• Phase 1: dropped OdsBranding entirely; replaced with OdsTheme { base, mode, headerStyle, overrides }. Logo / favicon / appIcon lifted to top-level OdsApp. Fonts moved onto the theme. Both parsers + 17 in-repo specs + ods-schema.json rewritten to the new shape — no parser shim.
• Phase 2: replaced the freeform font textbox with FontPicker.tsx — curated dropdown of system fonts + Google Fonts (lazy-loaded via injected <link> tags). UI label moved to “App identity & typography”.
• Phase 3: admins save theme back to the spec via PocketBase (React) — currentAppId + rawSpecJson plumbed through the app store; non-admins continue to write per-app localStorage.
• Phase 4: themeConfig() driver method + ThemeConfig type added to both contracts; conformance scenario s21 verifies both parsers produce the same {base, mode, headerStyle, overrides, logo, favicon} shape from the same spec.
• Phase 5: Flutter mirror of admin save-to-spec — AppEngine.loadedAppId + rawSpecJson + hotReplaceSpec, LoadedAppsStore.findById, dual-write in settings_screen.dart to update the spec via LoadedAppsStore.updateApp and hot-replace the in-memory model so the UI reflects the change immediately.

2026-04-24 — Docs pass

• CONTRIBUTING.md at root — explains the monorepo layout (where changes go by blast radius), local setup, the publish.sh test gate, when to add a conformance scenario, the test layers, and PR expectations. Refers to ARCHITECTURE / CONVENTIONS / TROUBLESHOOTING rather than duplicating.
• GLOSSARY.md at root — one-page reference for ODS vocabulary (spec, framework, builder, data source, ownership, cascade rename, formula, magic default, off-ramp, slug, visibleWhen, conformance scenario, etc.). Each entry points to where the concept lives in the spec or code.
• docs/testing.md — covers all five test layers (unit, component, integration, conformance, E2E), explains the React E2E folder convention (smoke/critical/regression/workflows/accessibility) and what goes where, includes the “where does my test go?” decision tree.
• Cleaned up pre-monorepo leftovers: moved Specification/CODE_OF_CONDUCT.md to root; removed the stale Specification/CONTRIBUTING.md.
• Updated README.md docs map to surface CONTRIBUTING / GLOSSARY / docs/testing.md / CONVENTIONS to visitors.

2026-04-24 — Ownership column auto-schema fix (both frameworks)

• Fixed Flutter DataStore.setupDataSources + React DataService.setupDataSources to auto-append the ownership column (ds.ownership.ownerField) to a data source’s fields when ds.ownership.enabled is true, unless the builder already declared it manually. Each has a small _fieldsWithOwnership / fieldsWithOwnership helper mirroring the other. Idempotent on re-runs because ensureTable / ensureCollection handle the “column already exists” case non-destructively.
• Removed the _owner-in-fields workaround from ownershipPerUser.json — s16 now passes against the canonical spec shape (ownership config + no explicit _owner field).
• Unblocks bundled examples like team-tasks-app.json which use the canonical shape and would otherwise fail on both Flutter (SQLite “no such column”) and React (PocketBase rejects unknown create fields).

2026-04-24 — Path B Session L (cascade parity fix + s20)

• Fixed cascade-rename parity bug. Flutter parser no longer converts flat-key cascade to nested shorthand — it preserves the canonical form as-is. Flutter runtime in AppEngine.executeActions now detects flat-key shape and reads parentField / childDataSource / childLinkField directly (instead of using cascadeMatchField as parentField, which broke when matchField differed from the renamed field). Legacy nested form kept as fallback.
• ActionHandler updated to prefer cascade.parentField over its withData-key deduction when computing cascadeOldValue.
• s20: cascade update renames matching children after renaming the parent. Uses matchField=’id’ + withData={name: ‘Projects’} + cascade with explicit parentField=’name’ — the pattern that failed before the fix. Passes on both drivers.
• Updated the legacy-flat-key model test to match the new canonical preservation behavior.
• cascadeRename capability declared on FlutterDriver; React already had it. Conformance total: 19 → 20 scenarios. React 1145 → 1146; Flutter 808 → 809. Test-count bump also reflects the batch-2 integration tests which now compile/run against the updated cascade runtime.

2026-04-24 — Later list rearrangement (document deferred coverage)

• Moved remaining untested conformance capabilities to the Later list with full fix direction: detail scenario, deepening tabs (clickTab primitive) and chart (series math) from initial-state to behavior, scenario-catalog parity check. Each kept as a standalone Later item so the Now/Next lists stay focused.

2026-04-24 — Path B Session K (s18 chart config + s19 kanban drag)

• s18: chart component snapshot preserves chartType, title, and dataSource config from spec. Narrow parity check — no aggregation math verified. ChartSnapshot added to the Dart contract; chart capability on FlutterDriver.
• s19: dragging a kanban card updates the row’s status field and the kanban snapshot counts shift accordingly. Introduced new driver primitive dragCard(dataSource, rowId, toStatus) on both OdsDriver contracts; both drivers implement it via the PUT data source (or fall back to the kanban’s own dataSource).
• KanbanSnapshot + KanbanColumn added to Dart contract; FlutterDriver emits it. kanban capability declared on FlutterDriver.
• Conformance total: 17 → 19 scenarios. React 1143 → 1145; Flutter 806 → 808.

2026-04-24 — Path B Session J (s17 tabs initial state)

• s17: tabs component snapshot reports labels in spec order with the first tab active by default. Pins the tabs snapshot contract across both frameworks.
• Added TabsSnapshot + TabsTab to the Dart contract; FlutterDriver now emits a TabsSnapshot for OdsTabsComponent.
• tabs capability declared on FlutterDriver (already on React).
• Conformance total: 16 → 17 scenarios. React 1142 → 1143; Flutter 805 → 806.

2026-04-24 — Path B Session I (s16 row-level ownership + Flutter schema bug surfaced)

• s16: ownership-enabled data source hides rows owned by other users. Alice + Bob each register + submit; each sees only their own rows in the list snapshot; dataRows (the god view) sees all rows regardless of session.
• Both drivers’ list snapshots now route through queryDataSource (React store) / engine.queryDataSource (Flutter) — so rowCount reflects what the current user would actually see, including ownership filtering. Previously both used the raw unfiltered query.
• auth:ownership capability declared on FlutterDriver (was already on React).
• Flutter schema bug surfaced (see Next list) — writing s16 revealed that Flutter’s DataStore doesn’t auto-add the _owner column for ownership-enabled data sources; the insert then fails on SQLite’s strict schema. Workaround in the scenario spec: declare _owner explicitly in fields.
• Conformance total: 15 → 16 scenarios. React 1141 → 1142; Flutter 804 → 805.

2026-04-24 — Path B Session H (s15 summary aggregates + cascade bug surfaced)

• s15: summary component value resolves aggregate expressions ({COUNT(tasks)}) against the data source; label passes through from spec. Exercises the summary capability on both drivers and the AggregateEvaluator / resolveAggregates cross-framework equivalence.
• Fixed ReactDriver’s snapshotComponent summary branch — previously read non-existent s.defaultValue and returned ‘’. Now uses s.value and resolves aggregates via the real AggregateEvaluator. Added SummarySnapshot to the Dart contract + summary capability on FlutterDriver.
• Seeded-data support added to FakeDataService.setupDataSources — mirrors the real DataService.setupDataSources so scenarios can declare seedData on local data sources.
• Cascade parity bug surfaced (see Next list) — writing a cascade scenario revealed that React’s handleCascade and Flutter’s AppEngine.executeActions disagree on which cascade shape is canonical. No spec satisfies both. Scenario deferred until the bug is fixed; new Next-list item captures the work.
• Conformance total: 14 → 15 scenarios. React 1140 → 1141; Flutter 803 → 804.

2026-04-24 — Path B Session G (s14 formulas)

• s14: formula fields compute from their dependencies and update on change. Tests both number formulas ({quantity} * {unitPrice}) and text interpolation ({firstName} {lastName}), plus the “any empty dependency → empty result” guard both evaluators enforce.
• Both drivers now evaluate formulas in formValues using their native evaluateFormula / FormulaEvaluator — the same evaluator the form renderer uses, so driver output mirrors what a user would see in the UI.
• formulas capability declared on both drivers. Conformance total: 13 → 14 scenarios. React 1139 → 1140; Flutter 802 → 803.

2026-04-24 — Path B Session F (s12 onEnd + s13 update rowAction)

• s12: submit action with onEnd navigate — pins Bug #2 fix (universal onEnd). New submitThenNavigate.json.
• s13: rowAction with action: update writes literal values to the matched row. Forces both drivers to extend clickRowAction beyond delete-only — the ReactDriver now dispatches update via executeActions, the FlutterDriver via engine.executeRowAction. New rowActionUpdate.json.
• action:update capability declared on both drivers. Conformance total: 11 → 13 scenarios. React 1137 → 1139; Flutter 800 → 802.

2026-04-24 — Path B Session E (shared scenario specs)

• Extracted 7 scenario specs to Frameworks/conformance/specs/ as pure JSON files — the single source of truth for both drivers. Kills the “write every scenario spec twice” tax that caused the label vs header parity bug earlier today.
• TS loader at Frameworks/conformance/src/load-spec.ts; Dart loader at Frameworks/flutter-local/test/conformance/load_spec.dart. Both read the same bytes.
• Scenario definitions shrank: scenarios.ts dropped 260+ lines, scenarios.dart dropped ~270 lines. Run bodies stay per-language — the per-driver assertion code is what needs to diverge.
• 22 parity tests still green (11 scenarios × 2 drivers).

2026-04-24 — Path B Session D (FlutterDriver full parity)

• FlutterDriver ported s06-s11 to full parity with ReactDriver. clickRowAction (delete), full visibleWhen evaluation (field + data), setClock via a driver-local _fakeNow that formValues lazily consults for CURRENTDATE/NOW defaults, and login / registerUser / currentUser / logout wired through the real AuthService + SQLite _ods_users table.
• Driver’s _boot now only sets skipAppAuth=true for single-user specs, so multi-user scenarios get proper AuthService.initialize() — creating _ods_users + _ods_user_roles tables.
• Conformance total: 11 scenarios × 2 drivers = 22 parity tests green. Flutter suite 794 → 800.

2026-04-24 — Path B Session C (FlutterDriver MVP + parity bug caught)

• New FlutterDriver mirroring the Dart contract (contract.dart), scenarios s01-s05 ported (scenarios.dart), runner at conformance_test.dart.
• Passes against the real AppEngine + SQLite (temp-dir per scenario). Flutter suite 789 → 794 tests.
• First parity bugs caught and fixed: (a) TS scenario specs used label on list columns; both React and Flutter parsers actually read header — TS silently accepted undefined, Flutter parser threw. Fixed specs to use header on both sides. (b) Dart OdsAction was missing the level field (TS had it after s03); added to the Dart model + fromJson.
• Flutter CI + publish.sh gates now include test/conformance. Parity drift between frameworks will block merges.

2026-04-24 — Path B Session B (auth in the driver + 2 scenarios)

• New FakePocketBase — in-memory stand-in implementing the subset of pb that AuthService touches (authStore, collection('users') CRUD + authWithPassword). Replaces the old mockPb() in conformance so the real AuthService runs against it — no parallel auth implementation.
• ReactDriver login + registerUser wired through AuthService (was throwing). Role defaults to the spec’s auth.defaultRole when caller omits it.
• s10 — register then login flow (no auto-login, correct credentials succeed, roles populated, logout returns to guest).
• s11 — login with wrong password returns false, session untouched; subsequent correct password succeeds.
• Conformance total: 9 → 11 scenarios; React suite 1135 → 1137.

2026-04-24 — Path B Session A (ReactDriver stubs + 4 scenarios)

• clickRowAction implemented — delegates to executeDeleteRowAction on the store; resolves non-_id matchFields by looking up the row first. Non-delete actions still throw (MVP scope). Pinned by s06 (scenarios.ts).
• visibleWhen evaluation in snapshots — mirrors the React renderer’s logic: field-based conditions read form state, data-based conditions query the data service for row counts. Pinned by s07 (field) + s08 (data count).
• setClock uses vi.setSystemTime + useFakeTimers; restored on unmount. formValues now lazily resolves CURRENTDATE/NOW defaults for unset fields using the current clock. Pinned by s09.
• Conformance total: 5 → 9 scenarios; React suite 1131 → 1135.

2026-04-24 — Org-level landing page

• Created One-does-simply/one-does-simply.github.io repo with a standalone index.html — now live at https://one-does-simply.github.io/. Inherits the indigo/purple brand palette from Specification/index.html; lists active (ods-pages) + planned (ods-chat, ods-workflow, ods-game) families.
• Updated the org’s public URL (blog field in the API) to the new landing page. Old stale URL (one-does-simply.github.io/Specification/) is now fully dead everywhere: org URL, repo homepages, runtime code.

2026-04-24 — Pages migration

• Enabled GitHub Pages on ods-pages (main branch, / root); URL base moves from https://one-does-simply.github.io/Specification/... (stale archived snapshot, can’t update) to https://one-does-simply.github.io/ods-pages/Specification/... (tracks main).
• Updated 13 tracked files (Flutter + React source, Flutter README) — runtime fetches for Examples, Templates, Themes, and the Build Helper prompt all point at the new URL. Old URL intentionally left to die.
• Seeded Specification/build-helper-prompt.txt (copy of the Flutter bundled asset) — the URL was a latent 404 on both old and new deployments; the React “Edit with AI” screen had been broken. Now resolves.

2026-04-24 — Cross-family setup (Session 2)

• CONVENTIONS.md at root — documents patterns every ODS family should copy (monorepo layout, publish.sh, TODO/REGRESSION_LOG format, ADR convention, CI pattern, CLAUDE.md) and what stays per-family. Calls the duplicate- don’t-extract decision explicitly.
• One-does-simply/.github repo created with profile/README.md. Visitors landing on github.com/One-does-simply now see an ODS family overview with active/planned siblings.

2026-04-24 — Public polish pass (Session 1)

• Root LICENSE (MIT) — mirrors Specification/LICENSE. Fixes “No license” label visitors see in the GitHub sidebar.
• Root SECURITY.md — points to private vulnerability reporting + lays out scope.
• README rewritten to reflect monorepo reality (dropped the “three sibling repositories” framing) + CI badges + framing as first family in the ODS ecosystem (README.md).
• Repo description + 8 topics set via gh repo edit (spec-driven, low-code, react, flutter, pocketbase, typescript, dart, monorepo) — improves discoverability.
• Flutter CI aligned with publish.sh: runs test/engine test/models test/parser test/integration --exclude-tags=slow instead of the whole tree. Widget tests were hanging on GH Linux runners (same root cause as the local Windows skip).

2026-04-24 — Monorepo consolidation closed out

• Upstream repos (Specification, Frameworks, BuildHelpers) archived on GitHub. Physical consolidation itself landed in commit d5c165e (2026-04-20) as a single bulk-add rather than a history-preserving git subtree add — original sub-tree histories now live in the archived upstreams.

2026-04-24 — TS strict gate restored

• Cleared pre-existing TS errors: _token rename in ColorCustomizer.tsx; _app/_page mis-prefix in code-generator.ts (8 errors from one signature); SendOptions type on the pb.send wrapper in pocketbase.ts; removed dead cardRotation helper + unused rotation memo in KanbanComponent.tsx.
• Re-added npx tsc -b step to .github/workflows/react.yml so type regressions block CI.

2026-04-24 — Root-level CI workflows

• Moved Flutter + React workflows to root .github/workflows/ (flutter.yml, react.yml). The nested copies under Frameworks/*/.github/workflows/ were orphaned after the monorepo consolidation (GitHub Actions only reads workflows from repo-root .github/workflows/) — neither framework had working CI. Adjusted paths for the new layout (single checkout, no separate Specification clone).
• Flutter workflow excludes slow-tagged Batch 9 perf tests (flutter test --exclude-tags=slow) so I/O-bound perf tests don’t flake the gate on cold runners.
• React E2E job preserved (Playwright + PocketBase via tests/e2e/global-setup.ts).

2026-04-23 — Small-win sweep

• OdsAction parser preserves level on showMessage (ods-action.ts); conformance s03 tightened to assert level: 'success'; driver dropped its unsafe cast (react-driver.ts).
• Batch 9 perf tests tagged slow via library-level @Tags(['slow']) in batch9_performance_test.dart; dart_test.yaml declares the tag; publish.sh runs with --exclude-tags=slow.
• Stale A couple of comments.txt TODO entry removed (file was already gone).
• appPrefix isolation unit tests — pin the multi-app collection contract at the real DataService boundary (data-service.test.ts).
• LoginScreen signup-clears-both-gates regression test — renders <LoginScreen> with needsAdminSetup=true, runs signup, asserts both needsLogin and needsAdminSetup clear; failure path leaves gates intact (LoginScreen.test.tsx).
• pocketbase.ts module-init guard — source-level test fails if pb.authStore.clear() creeps back into pocketbase.ts (pocketbase.test.ts).

2026-04-19 — Batch 8 E2E + gap closure + storage/auth infra + docs pass

• Batch 8: 22 Playwright regression tests on real PB (see REGRESSION_LOG.md)
• Automated PB startup in Playwright globalSetup — no manual steps
• PB users collection auto-create (AuthService.ensureUsersCollection)
• Fixed LoginScreen needsAdminSetup-on-signup latent bug
• Kanban drag-and-drop E2E coverage with data-ods-kanban-* attrs
• Removed pb.authStore.clear() from pocketbase.ts so admin sessions persist across navigation
• Unified Flutter user list — extracted FrameworkUserList, removed per-app _InlineUserList + UserManagementScreen
• CURRENT_USER.EMAIL token in Flutter form resolver + currentEmail on AuthService / FrameworkAuthService
• Storage folder bootstrap + first-run picker + Move Data UX
• OneDrive removal + migration to c:\Apps\One-does-simply
• .claude/settings.json allowlist tuned (fewer permission prompts)
• Workspace README.md + ARCHITECTURE.md
• docs/adr/ folder + ADR convention; conformance-driver doc lined up as ADR-0001 (physical move deferred until review lands)
• docs/TROUBLESHOOTING.md — 8 gotchas, symptom/cause/fix format
• Per-framework ARCHITECTURE.md for flutter-local and react-web

This site is open source. Improve this page.