## Outcome

Reviewed baseline `572ef44fc330911673ce190b7abfbbf210bff77c` read-only. The current contracts already support verified TTWROR/XIRR and a True Wealth confirmed/modelled series, but the default wealth history is confirmed-only and the UI conflates “complete,” “confirmed,” and “verified.” Correction markers and source/KPI sparklines are not represented.

## Current contract/UI map

- `frontend/src/pages/PortfolioPage.vue:29-42`
  - Four separate sections: overview, performance, data/diagnostics, strategy.
  - Overview mounts only `WealthCockpitPanel`; verified performance is a separate route.
  - Refresh keys are independent; confirmed imports increment both.

- `frontend/src/components/wealth/WealthCockpitPanel.vue`
  - Fetches `getWealthCockpit(period)` (`:163-196`).
  - Period vocabulary: `ytd | previous_year | 12m | all` (`:10-16`, API type `portfolio.ts:139`).
  - Default history is `cockpit.history.points` and a single SVG polyline (`:39-48`, `:193`).
  - Source badges/statuses come from readiness, freshness, reconciliation, performance, value basis and import coverage (`:92-121`).
  - Diagnostics are already collapsed (`:131-155`).
  - No source sparklines, correction markers, axes, point labels, or explicit per-point evidence basis.
  - Despite the copy saying gaps are not supplemented, sparse dates are connected continuously by one polyline.

- `frontend/src/components/performance/PortfolioPerformancePanel.vue`
  - Eagerly loads coverage, performance, wealth, setup, scheduler status, and True Wealth projections (`:171-230`).
  - Period vocabulary differs: `ytd | 1y | 3y | 5y | all | custom` (`:23-30`, `:178`, `:213`).
  - KPI row exposes current value, contributions, result, XIRR, TTWROR and quality (`:144-151`).
  - True Wealth alone uses explicit confirmed/modelled points from `TrueWealthPerformanceView` (`:133-142`, `:228`).
  - General `PortfolioPerformance.time_series` is forcibly treated as confirmed in chart construction; model provenance is lost (`:228`).
  - `ttwror_series` exists in the API but is not rendered.
  - XIRR is only scalar.
  - The chart has cashflow and anchor markers but no correction marker (`:153-164`).
  - `canShowSeries` unnecessarily requires both TTWROR and XIRR complete for non-True-Wealth valuation charts (`:191-199`).

- API client: `frontend/src/api/portfolio.ts`
  - Wealth contract: `WealthCockpit`, `WealthSourceRow`, `WealthDiagnostic` (`:85-157`).
  - Verified performance: `PortfolioPerformance`, `PerformanceCoverage` (`:11-36`).
  - True Wealth model contract: `TrueWealthPerformanceView`, `TrueWealthModelPreview` (`:80-81`).
  - Endpoints:
    - `getWealthCockpit` (`:181`)
    - coverage/performance (`:186-190`, `:202-207`)
    - True Wealth view/model (`:194-195`).

- Backend:
  - Routes and response models: `src/jarvis_finance/api/routers/overview.py:130-146,217-229,303-328,474-524`.
  - Wealth schema: `api/schemas/wealth_cockpit.py:12-127`.
  - Performance schema: `api/schemas/portfolio_performance.py:8-142`.
  - Wealth history intentionally uses exact common stored dates, without interpolation: `services/wealth_cockpit.py:_household_history`, especially `:264-380`.
  - `build_wealth_cockpit` assembles history, six KPIs, readiness and diagnostics at `:1070-1264`.
  - `build_portfolio_performance` computes TTWROR/XIRR and emits `time_series`, `ttwror_series`, cashflows and quality at `services/portfolio_performance.py:852-1274`.
  - Immutable supersession exists and is cutoff-reproducible, but correction lineage is discarded before serialization; see `tests/unit/test_portfolio_performance_foundation.py:test_snapshot_versions_are_immutable_and_cutoff_reproducible`.

## Recommended additive contracts

Add explicit facts rather than deriving “verified” from source-name strings:

```ts
export type EvidenceBasis = 'confirmed' | 'modelled'
export type SeriesQuality = 'complete' | 'partial' | 'unavailable'

export type ModelledWealthPoint = {
  at: string
  value_chf: string
  basis: EvidenceBasis
  quality_status: Exclude<SeriesQuality, 'unavailable'>
  anchor: boolean
  coverage: {
    included_sources: number
    expected_sources: number
  }
}

export type ValuationCorrectionMarker = {
  valuation_at: string
  recorded_at: string
  scope: 'portfolio' | 'postfinance' | 'truewealth' | 'crypto'
  kind: 'valuation_superseded'
  previous_value_chf: string
  replacement_value_chf: string
}

export type WealthDevelopment = {
  status: SeriesQuality
  points: ModelledWealthPoint[]
  correction_markers: ValuationCorrectionMarker[]
  reason_codes: string[]
}
```

Add to `WealthCockpit`:

```ts
development: WealthDevelopment
```

Optionally add source-level sparklines where aggregate development cannot be reused:

```ts
// on WealthSourceRow
trend_points: Array<{
  at: string
  value_chf: string
  basis: EvidenceBasis
  quality_status: 'complete' | 'partial'
}>
```

For verified performance, preserve legacy `time_series` and add an explicit versioned projection:

```ts
export type VerifiedValuationPoint = {
  at: string
  value: string
  basis: EvidenceBasis
  anchor: boolean
  quality_status: 'complete' | 'partial'
}

export type ReturnSeriesPoint = {
  at: string
  value: string
}

export type PerformanceVerification = {
  ttwror: 'verified' | 'provisional' | 'unavailable'
  xirr: 'verified' | 'provisional' | 'unavailable'
  reason_codes: string[]
}
```

Add to `PortfolioPerformance`:

```ts
valuation_series_v2: VerifiedValuationPoint[]
xirr_series: ReturnSeriesPoint[]       // backend-computed rolling/endpoint XIRR
correction_markers: ValuationCorrectionMarker[]
verification: PerformanceVerification
```

Backend equivalents should be exact Pydantic models with `ConfigDict(extra="forbid")`, added to:

- `api/schemas/wealth_cockpit.py`
- `api/schemas/portfolio_performance.py`

Do not call modelled values “verified.” A metric is verified only when its metric quality and period coverage are complete and its required valuation inputs are confirmed. Keep TTWROR cumulative and XIRR annualized labels explicit.

## Minimal Vue redesign

1. **Default overview**
   - Keep `WealthCockpitPanel` as the primary screen.
   - Render `cockpit.development` as “Modellierte Vermögensentwicklung,” with:
     - solid confirmed segments;
     - dashed modelled segments;
     - explicit `Modelliert`/`Bestätigt` badge;
     - cashflow and correction markers;
     - no return terminology.
   - Derive overview KPI sparklines from `development.points`; use `source.trend_points` for source cards.

2. **Collapsed verified area**
   - In `PortfolioPage.vue`, place an initially closed `<details>` below the standard cockpit:
     - summary: “Verifizierte Performance & Diagnose”
     - lazy-mount `PortfolioPerformancePanel` only after first expansion.
   - Keep `/portfolio/performance` as a direct expanded route for compatibility.
   - Add a small `embedded`/`compactHeader` prop to `PortfolioPerformancePanel` rather than duplicating it.

3. **Performance panel**
   - Use `verification.ttwror/xirr` for badges.
   - Render existing `ttwror_series`; render additive `xirr_series` as the metric sparklines.
   - Use `valuation_series_v2` for confirmed/modelled styling.
   - Render `correction_markers` separately from cashflows and anchors.
   - Decouple valuation-chart visibility from XIRR/TTWROR completeness; return metrics may remain unavailable while stored valuation points are still chartable.

4. **Shared period control**
   - Extract a small `PortfolioPeriodControl.vue`.
   - Canonical UI presets: `ytd | 1y | 3y | 5y | all | custom`.
   - Convert presets to explicit `from/to` client-side for performance.
   - Either extend the wealth endpoint to explicit `from/to`, or add `1y/3y/5y/custom` to its validated preset contract. Do not silently map `1y` to current `12m` while retaining two vocabularies.
   - Persist period in the URL so overview and expanded diagnostics use the same period.

5. **Laptop/iPad**
   - Change the performance filter from `lg:grid-cols-5` to a safer 12-column layout: scope/period 3 columns each, dates 3 each, apply button full-width or right-aligned below at tablet widths.
   - Keep the wealth history/distribution split at `xl`; it already stacks below 1280 px.
   - Fix `WealthCockpitPanel.vue:105`: `xl:grid-cols-4` is nested inside a one-third-width source card and produces four tiny columns. Use `sm:grid-cols-2` consistently.
   - Use horizontally scrollable segmented period controls only if each target remains at least 44 px.
   - Add explicit probes for 1366×768, 1440×900, iPad portrait 768×1024 and landscape 1024×768.

## Contract mismatch and semantic risks

- Frontend marks backend-required performance fields optional:
  - `ttwror_version`, `xirr_version`, `attribution_version`, `ttwror_series`, `attribution` in `portfolio.ts:20-31`.
  - Backend requires them in `portfolio_performance.py:129-142`.
- Frontend marks coverage classification/cashflow status optional (`portfolio.ts:35`), while backend requires both (`portfolio_performance.py:107-108`).
- `WealthCockpitResponse` forbids top-level extras, so frontend-only `development` additions will be stripped/rejected unless schema and service change together.
- Wealth nested `kpis`, `history`, `totals`, `policy` remain broad dictionaries (`wealth_cockpit.py:115-127`), allowing backend/frontend drift despite top-level validation.
- “Bestätigt” currently maps directly from quality status `complete` (`PortfolioPerformancePanel.vue:212`); computational completeness is not evidence provenance.
- Portfolio performance’s “Aktueller erfasster Wert” uses whole-household wealth, including bank cash, instead of investment closing value (`PortfolioPerformancePanel.vue:190`). This conflicts with the performance scope.
- General performance chart labels all points confirmed (`:228`) even where modelled valuation sources may be present.
- Sparse wealth observations are joined visually across calendar gaps (`WealthCockpitPanel.vue:193`).
- Correction lineage is available in storage but not exposed after canonical valuation selection.
- `all` chooses the earliest date across every coverage row, not the selected scope (`PortfolioPerformancePanel.vue:213`).
- Client-clock preset boundaries and backend period boundaries can differ around timezone/day rollover.
- There is no committed browser/layout suite; Vitest/jsdom cannot prove overflow, target sizes, or chart geometry.

## Tests to add/update

- Backend:
  - `tests/unit/test_wealth_cockpit_v1.py`
    - ordered/unique/no-future development points;
    - partial coverage stays modelled, never confirmed;
    - correction markers respect `data_cutoff`;
    - exact Pydantic/OpenAPI fields.
  - `tests/unit/test_portfolio_performance_foundation.py`
    - `valuation_series_v2` provenance;
    - verified status fails closed for modelled/partial inputs;
    - deterministic `xirr_series`;
    - superseded valuation marker without exposing internal IDs.
- Frontend:
  - `PortfolioPage.test.ts`: verified panel closed and unfetched initially; lazy load on expansion; direct performance route remains expanded.
  - `WealthCockpitPanel.test.ts`: confirmed/modelled badge and segment styling, source sparklines, correction marker, sparse-gap behavior.
  - `PortfolioPerformancePanel.test.ts`: separate verified/provisional badges, TTWROR/XIRR sparklines, corrections, valuation chart visible independently of return availability, shared period changes refresh all surfaces.
  - Add browser tests for the four target viewports, no horizontal overflow, 44 px controls, keyboard/ARIA details behavior and period persistence.

## Verification and workspace state

- Frontend focused suite: **24 passed**.
- Backend focused suite: **48 passed** using system Python.
- Repository remained clean at the requested SHA.
- No files created or modified.
- The repository-local `.venv` was absent and the runtime venv lacked pytest; system `python3 -m pytest` succeeded.