cesnimda
20 KiB
Phase 1 completion summary
Completed 2026-09-10 on codex/ui-redesign-audit. This is an audit checkpoint, not a verified production release. No visual redesign, component-library migration, route restructuring, Phase 2 implementation or B07 work has begun. B06/F04 closure remains open.
The detailed architecture assessment, functionality ledger, route/workflow map, API/identity map, component inventory, test assessment and migration risks form this completed audit. The source census remains valid: the only src/app change since its checkpoint is the separately documented interview test assertion. This final completion commit changes documentation only.
1. Existing functionality to preserve
All 57 preservation IDs in the full ledger remain protected, including features that already satisfy the brief. These are capability groups, not a claim of 57 individual controls or complete runtime test coverage.
| Ledger IDs | Complete functional coverage by area |
|---|---|
| E01–E09 | Public plans; local and provider authentication; registration/verification/reset/email change; 2FA/recovery/trusted devices; sessions/logout; account identity/avatar/password; sidebar/command search; theme/localisation/shared feedback; cross-account state isolation |
| J01–J10 | NAV discovery and saved alerts; reviewed manual/URL/paste capture; extension/bookmarklet capture; table/mobile cards/filter/sort/pagination/columns/saved views; edit/bulk actions/trash/restore; Kanban; companies/contacts; offer comparison; JSON/CSV export and import/conversion |
| W01–W12 | Full job workspace and quick overlay/deep links; progression/timeline; analysis/match/strategy; checklist; CV selection/tailoring; cover letters/review/version restore; application drafts/frozen submitted packages; attachments; interview prep/debrief; reminders/calendar/notifications; asynchronous operations |
| C01–C12 | Canonical Career Profile and long-tail sections; overview/completeness/history; file extraction/diff/accept/discard; evidence bank; CV library/copy/publish/share; responsive editor/preview; item and section visibility/order/overrides; template-aware customization; page/document settings; undo/redo/save/history; section/document AI review; private/public PDF |
| M01–M06 | Job correspondence records; Gmail search/import/thread linking; global inbox/filter/attachment information; provider compose/reply/drafts/review/send/delivery uncertainty; Gmail review decisions; Google/Microsoft/IMAP connections |
| S01–S05 | Personal analytics/next actions; owner-local widget visibility/onboarding/notifications; follow-up/email/table/theme/language preferences; AI privacy/usage/Free–Pro/billing capabilities; account export/encrypted backup/lifecycle/deletion |
| A01–A03 | Admin user/role/reset/deletion guards; audit/undo/restore; separate operational diagnostics/email settings/AI probes/CV benchmarks/maintenance |
The route map accounts for 31 concrete patterns: 29 page patterns and two compatibility aliases. It also records not-found/error boundaries and non-route entry points: capture query strings, command search, notification links, workspace section/tab aliases, OAuth popups and public CV slugs. Embedded Documents, Analytics and Notifications functionality is inventoried even though these are not current standalone routes.
Protected workflows include discover/capture → review/create → analyse/match → select/tailor CV → prepare material → submit/freeze package → track/remind → interview/outcome; import CV → review canonical profile → create/edit/publish/export variant; inbox/import → link → draft/review/send; and account/session/admin lifecycle. Exact sub-actions and source owners are in the ledger; a redesigned page must not silently omit its nested functions.
2. Existing functionality needing UX improvement
These are preserve-and-improve, not missing features or authorization to implement them now. Assessment concerns coherence/discoverability unless a defect is explicitly identified in section 4. No new user study or visual/manual accessibility session is claimed.
| Existing capability / IDs | Already satisfies the brief | Remaining UX work |
|---|---|---|
| Navigation and command launcher, E07 | Collapsible sidebar, section support, mobile drawer, active child-route matching, Ctrl/Cmd+K job/company search | Main features currently share one broad Manage group. Improve grouping and discoverability; retain account/admin separation. Extra command capabilities are new scope below |
| Job views and workspace, J05–J07, W01–W12 | Desktop table/mobile cards, saved views, keyboard Kanban, full job page and shared quick overlay, next actions and artifact relationships | Present table/board as views of one workspace; clarify section hierarchy and full-page/preview transitions. Preserve every alias, status and nested workflow |
| Career and CV, C01–C12 | Canonical profile, document overrides, real template previews, page boundaries/automatic fit, drag and keyboard reorder, undo/history, desktop split/mobile edit-preview, PDF | Improve source-versus-override clarity, advanced-control discovery and preview focus. Keep advanced existing controls despite the brief's shorter suggested list |
| Contextual AI, W03/W05/W06/W09, C03/C11 | Context selection, source material, preview/diff and accept/reject already exist | Standardize current/suggested/evidence presentation and locked/error states. Presence of review UI is not proof that every AI output is factually correct |
| Dashboard/onboarding, S01–S02 | Metrics/trends, next actions, five owner-local visibility switches, guided checklist/dismissal | Improve hierarchy and empty/loaded states; additional widget/layout contracts are new work |
| Mail/documents/notifications, M01–M06, W06–W08/W11–W12 | Drafts, explicit send/review, delivery states, job-linked files, inbox/review, reminders and operations | Make destinations and status distinctions easier to find. Unified navigation relocates these capabilities; it does not replace their persistence contracts |
| Calendar and account/settings/admin, W11, E02–E06, S03–S05, A01–A03 | Follow-up/deadline Google/Outlook event creation and ICS; account controls; separate admin tools | Preserve provider availability/permission handling and clarify account versus career versus provider identity; keep operational diagnostics separate from personal data |
| Themes/localisation/responsiveness, E08 and all flows | MUI theme, light/dark/system, EN/Bokmål and existing workflow adaptations | Standardize tokens, layout/focus/error states and test populated EN/NB/mobile surfaces. New density and brand choices are separate requirements |
3. Genuinely new requirements and conditional additions
These distinctions come from comparing the brief to the current handlers/models, not from treating every desired noun as a missing page. Proposed defaults are not permission to change stored data.
| Brief requirement | Current boundary / incremental work | Scope decision |
|---|---|---|
| Dashboard ordering, resizing, reset and remembered layout | DashboardView persists only cards/activity/funnel/companies/skills visibility flags |
New layout behavior. Decide device-local versus cross-device persistence before backend/schema work; Phase 8 |
| Expanded command palette | QuickCommandDialog searches jobs/companies and has six static commands | Add CV/document search, CV creation, Career Profile, appearance/language and contextual commands where specified; reuse existing actions |
| Comfortable/compact application density | Theme/table display preferences exist; CV document density exists | New application preference affecting chosen dense surfaces; do not confuse it with document density |
| Configurable statuses and category mapping | JobPipeline preserves free-text/custom values; pipeline.ts defines fixed groups and an Other fallback | New configuration/mapping workflow. Preserve Waiting/Ghosted, raw values, analytics and rules; never replace statuses with the brief's suggested list |
| Per-task due dates and notifications | Checklist model stores status/order/completion; follow-up and deadline dates belong to the job | New task scheduling contract if adopted. Existing job reminders do not prove per-task scheduling |
| Optional editable weekly goals | Existing analytics and onboarding do not store a weekly target | New optional capability with supportive defaults; no universal application quota |
| Intent-led onboarding and suggested profile preferences | OnboardingChecklist links existing account/profile/job/provider steps; it does not collect the brief's initial job-seeker intent | New intake/personalization scope if adopted. Suggested desired-role/remote/salary preferences need an explicit data contract; do not infer them from job salary/location |
| Shared contextual-assistant entry and command actions | Domain-specific AI controls/wrappers exist; no persistent shell assistant is wired into the current route graph | New interaction surface/reuse work. A general-purpose chat platform is not required to satisfy contextual access |
| Aggregate product/admin dashboard and candidate metrics | AdminSystemPage is operational; personal dashboard analytics already exist | Cohort/adoption/conversion and CV-performance additions require defined measures, permitted aggregate data and availability. Candidate widget lists are not a promise every metric exists or must ship; correlation is not causation |
| Manual CV preview zoom | CvBuilderEditor automatically fits page width; drag reorder, pagination and history already exist | Manual zoom is a recommended incremental control, not a reason to rebuild the preview or declare all builder capabilities missing |
| Logo and brand treatment | Existing mark/theme remain functional | New logo is required by the brief; exact accent/name decisions are deferred. No automatic product rename |
| Broader event/provider support, priority/favourite and optional quick-drawer form | Existing calendar actions cover follow-up/deadline; current quick overlay is functional; suggested fields/providers are not all current contracts | Conditional product additions or presentation choices. Broader calendar is explicitly future scope. Do not remove existing integration or introduce schema merely to match illustrative examples |
Grouped navigation, a Jobs view switch, and clearer Documents/Analytics/Notifications destinations belong to section 2's reorganization of existing behavior; they do not automatically require new APIs or a new resource model. New analytical measures or cross-job queries must be scoped separately from relocating existing views.
4. Known baseline defects and technical/UX debt
The existing validated queue owns these defects; this final audit pass records them and does not fix or re-audit them. Prior independently reproduced findings remain distinct from source-only risks and unresolved production evidence.
| Status / owner | Known issue relevant to redesign | Required treatment |
|---|---|---|
| Open, F01 / D01–D03 | CV/profile/job lost-update and atomic-history/projection contracts | Preserve existing history; a mounted editor's save queue does not solve concurrent writers. Coordinate before changing persistence/conflict UX |
| Open, D04–D07 | Duplicate creation intent, concurrent/omitted byte accounting and export/history growth | Preserve current warnings/quotas; do not silently add idempotency, retention or storage redesign |
| Open, F14 / A01–A03 | Previously validated shared contrast and preview keyboard-scroll defect; thin loaded-state accessibility coverage | Track structural fixes separately; no compliance claim from shell-only axe samples |
| Open, W01/W05 | Calendar date/timezone correctness and private unsaved-draft recovery scope | Preserve current semantics while documenting defects. Do not persist private drafts across accounts or invent new storage |
| Open, R01b/R02/R03/R05 and paid-lifecycle queue | Production DB fallback validation, consistent backup/restore/rollback proof, worker survival, billing entitlement/deletion | Separate owners and release gates. No schema work before recovery proof; billing UI cannot establish entitlement correctness |
| Historical error fixed; R01a partly open | interview-prep.test.tsx:120 TS2571 reproduced, then corrected with a test-only matcher during authorized release repair |
Standalone tsc now passes. Explicit standalone CI gate is still missing; do not continue calling TS2571 an active failure |
| Current CI block | Run 743 browser worker SIGSEGV, then timeout/closed-browser errors; deploy skipped | Obtain runner diagnostics before another release attempt. Do not weaken gates or label the crash an application assertion failure |
| Source-backed maintainability/UX debt | Two router layers; large stateful forms; mixed breakpoints; hardcoded shell colors alongside theme; hardcoded English mobile accessible name | Small, tested future changes. Module size is not performance evidence; cleanup candidates are not approved deletions |
| Unverified risk | Router identity changes with display preferences; optional labels/select patterns; loaded EN/NB/theme/mobile/manual accessibility gaps | Targeted regression/manual checks needed. Do not promote these to reproduced security defects |
| Open, B06/F04 | Actual deployed ingress trust and bypass protection remain unproven | Separate production verification below; repository checks cannot close it |
See validated ticket boundaries, test limitations and regression-risk evidence. Identity isolation, record concurrency and creation idempotency remain three different contracts.
Functionality preservation contract
Every ledger ID defaults to preserve. Before migrating a workflow, record its destination, retained actions/state transitions, exact API/data meaning, permission/session boundary and regression evidence. Relocation/merger must retain capability; deprecation requires an explicit documented decision. No deletion is approved in this audit.
Preserve direct links/aliases/capture/OAuth flows; owner/session/role/entitlement checks; canonical Career Profile versus document overrides; AI preview/reject/apply and factual provenance; dirty/conflict recovery and private-state disposal; server drafts/history/frozen packages and uncertain-send behavior; publication versus private downloads; B01–B05 safety contracts; typed date/JSON/CSV/PDF formats and Norwegian text. Existing backend defects are recorded for their own tickets, not redefined as acceptable intended behavior.
A migrated workflow needs same-account success and cross-account denial, populated/empty/error/locked states, EN/NB, light/dark, desktop/mobile, keyboard/focus and artifact checks appropriate to its behavior. A screenshot, module presence or aggregate test count alone is insufficient preservation evidence.
Frontend architecture and stack assessment
Retain the current stack. React 19/strict TypeScript is a client-rendered SPA packaged by Next 16 static export, with React Router 7 owning application routes and nginx handling static/deep-link/API routing. MUI 7, MUI X, Emotion and MUI icons already provide the foundation; styling combines central theme/global CSS and substantial local sx. TanStack Table is used, not TanStack Query. No Redux/Zustand/cache-platform replacement is warranted.
Local React state and focused contexts own forms/auth/theme/language/plan/feedback; shared Axios owns cookies/XSRF/B02 request provenance/cancellation and blob downloads. The API census contains 266 call expressions; ownership is server-enforced. Improve shared component contracts and existing tokens before page migration. No demonstrated limitation justifies replacing the component library, routing or backend.
Test and CI baseline
| Evidence | Result / limit |
|---|---|
Latest completed CI, run 743 at main 27657f3 |
916/916 backend, zero skips; 67 suites / 298 frontend tests; lint, dependency audits, Compose/safety/parser and production builds passed |
| Same CI browser gate | 6 passed / 12 failed; initial worker SIGSEGV, retry timeout and closed-browser errors. Deploy skipped; not a green release |
Local release checks at a861ca4, unchanged source since |
18/18 browser, standalone tsc, lint and static build passed; installed npm audit zero vulnerabilities. Local success does not replace CI |
| Prior B06 focused tests | 166 backend, 3 Compose, 6 native Production ingress, 9 native B04 HTTP checks passed; no actual Traefik/nginx topology |
| Accessibility/test census | 76 frontend test files, 106 backend test-source files; general axe samples eight routes in English, not all loaded interactions/manual accessibility |
| Final audit-only validation | 621 local links valid, 57 unique ledger IDs, 31 route patterns, 266 API rows, all 120 production source files match the census; only intervening src change is the documented test assertion. Secret scan and diff checks pass; application suites were not rerun for prose edits |
Exact commands, historical failures and classification are preserved in TEST_INVENTORY and the release report. Previous CI/build changes were separate authorized release/baseline repairs: Compose availability, Linux fixture/runtime/TLS-test portability, blocking dependency advisories and the test-only TS2571 correction. They are not redesign implementation, and this final pass adds none.
Unresolved production validation
B06 implementation complete, but B06 runtime/production closure and F04 remain open. Intended path: Traefik → host 3000 → nginx → internal API 8080. Docker Desktop's inference-socket failure prevented the real local container chain. No verified production SSH alias was available; no credentials or CI secrets were obtained.
Actual nginx startup/fail-closed configuration/restart, Traefik socket peer and trusted CIDR, backend client IP/scheme, real-chain B02/B04 behavior, host/container listeners, firewall/NAT/IPv6 and independent direct-port denial still need evidence. Public DNS/valid HTTPS were observed, but intended-address inventory and the HTTP redirect to unreachable HTTPS port 444 remain unresolved. Live /health last observed 2026-09-10 05:07 UTC, version 299, ready; this is not deployment of the pushed audit repairs. Exact safe operator commands and expected evidence remain in the B06 report.
Recommended Phase 2 starting point
Start only when separately authorized: a bounded functional-baseline session, with no layout or route changes. Add the explicit standalone type-check script/CI step while coordinating with runner repair; map the 57 preservation IDs to existing assertions and add the highest-risk missing tests for route/overlay/capture parity, populated CV/profile overrides and AI accept/reject. Preserve real API/B02 identity tests and B01/B03/B04/B05 suites.
Exit requires a documented pass/fail/unavailable baseline, meaningful loaded-workflow proof and a preservation checklist tied to tests. Existing defects stay separately ticketed; failed tests are not made green by weakening expectations. Record-integrity and recovery prerequisites govern later persistence/schema work. Phase 3 navigation proposals and Phase 4 visuals follow their own checkpoints; no stable/live tag or F04 closure is warranted while their evidence is missing. Branding, product additions and cross-device layout decisions do not block this audit's completion.