cesnimda
4.9 KiB
Career Profile UX
ARC-08; PRES-C01–PRES-C04, PRES-C07, PRES-C10, PRES-C11, PRES-E06, PRES-E09. Sources: CareerProfilePage, career sections, CvVariant, Phase 2 canonical/sibling isolation regression.
One canonical source
Career Profile is the account-owned structured career singleton. Account /profile owns login identity/email/security; CVs own identified settings and item overrides. Label every edit destination before editing, including AI review:
| Scope label | What changes | What must not change |
|---|---|---|
| Career Profile — shared source | Canonical selected profile fields after explicit profile save/import acceptance | Account login/email/security; existing CV overrides; frozen submitted packages |
| This CV only — [name] | Selected variant settings/item override | Canonical profile, sibling CVs, other owners |
| This application — [role/company] | Job draft, letter, answers or existing job-specific artifact | Canonical data; unrelated application drafts |
| Suggestion — not saved | Review proposal only | Any stored user-owned content until deliberate acceptance/save |
Live CVs can reflect later canonical edits where they inherit the source. Never promise that editing Career Profile leaves all rendered live CVs unchanged. Existing overridden fields remain overridden; frozen submissions remain historical. Show this consequence near profile Save and the variant's source label. An existing published live CV also needs clear explanation; it is not an immutable submitted snapshot.
Page organization
Overview gives completeness and actionable missing information, recent import/history and CV links. Edit offers personal information, summary, work, education, skills, languages, interests and all existing long-tail/custom sections. Evidence has its existing sourced statement/title/category/tags list/add/delete. Import review/history stays a distinct activity with operation status.
Do not force onboarding before access. Profile completeness is guidance, not a public employability score or requirement to invent content. No “missing skill” autofill from job requirements. Desired role/location/remote/salary preferences are NEW-13 only where not currently represented; current job fields do not satisfy career intent preferences.
Preserve avatar editing, raw CV text copy, manual edit, history restore, extraction rebuild/improve/reprocess, retry/cancel and low-confidence selections. Do not reduce the canonical schema to the shorter illustrative section list.
Import and AI review
Choose/upload source → processing with operation ID → compare extracted proposal with current profile → select supported changes (including low-confidence controls) → accept or discard. Display source document/field/confidence evidence when returned. Confidence alone does not verify truth.
Discard leaves canonical data intact. Failed extraction retains original and recovery actions; cancellation is not success. Low-confidence or conflicting personal claims require explicit user verification. User edits the proposal before acceptance where supported. No silent replacing of profile or automatic propagation from an AI assistant.
Privacy/entitlement/provider availability explains disabled actions and offers manual editing. Never send unselected attachment content or retain private extracted text in global shell state.
Source and override interaction
Variant item references retain stable source keys. “Edit Career Profile” navigates to source; “Customize in this CV” creates/edits the existing override. “Use profile value” resets only that override, showing current source value first. Profile item removal can leave stale selections/overrides: surface an unresolved source link with repair choices, not silent reassignment to a similarly named item. Exact missing-source behavior needs a focused fixture before changing it.
Do not add one-click “promote all CV changes to profile.” If later requested, it needs an explicit field-level review and same-account server contract. No generic synchronization framework or schema move.
Preservation and migration gate
Current save/history/projection behavior remains, but F01 record concurrency is unresolved. Plan source/override wording independently; coordinate any save/history/conflict owner edits with D01–D03 and recovery gates. B02 alone is not optimistic concurrency.
Before restructuring: native A→B stale-state denial; canonical/sibling unchanged after variant override; profile inherited-field propagation and override reset; extraction accept/discard/low-confidence/no-provider states; history restore; long sections and loaded keyboard names. Phase 2 already proves a headline override path, not every item/restore/import combination.