genshin-assistant/docs/PROJECT.md

# Project

This document is the source of truth for project intent, scope, runtime facts, and operational expectations.

For implementation structure, see [ARCHITECTURE.md](ARCHITECTURE.md). For engineering standards, see [CONVENTIONS.md](CONVENTIONS.md).
For the current app status, see [CURRENT_STATUS.md](CURRENT_STATUS.md).
For the next scan-result and artifact-inventory product phase, see
[scanner-results-inventory-roadmap.md](scanner-results-inventory-roadmap.md).
For the 2026-07-09 scanner merge evidence, see [MERGE_READINESS.md](MERGE_READINESS.md).
For Gitea push/authentication setup, see [GITEA_AUTH.md](GITEA_AUTH.md).
For the completed evaluation/offline phases and executed live acceptance, see
[ARTIFACT_EVALUATION_OFFLINE_VALIDATION_2026-07-10.md](ARTIFACT_EVALUATION_OFFLINE_VALIDATION_2026-07-10.md)
and [NATIVE_SCANNER_LIVE_ACCEPTANCE_2026-07-10.md](NATIVE_SCANNER_LIVE_ACCEPTANCE_2026-07-10.md),
plus the current-build [NATIVE_SCANNER_LIVE_SMOKE_2026-07-11.md](NATIVE_SCANNER_LIVE_SMOKE_2026-07-11.md).
For the score-free Build-Fit V1 data and safety contract, see
[BUILD_FIT_CONTRACT_V1.md](BUILD_FIT_CONTRACT_V1.md).
For the current English-default UI, local-result removal, timing, and reviewed
OCR-corpus checkpoint, see
[LOCALIZATION_LOCAL_DELETION_TIMING_CHECKPOINT_2026-07-11.md](LOCALIZATION_LOCAL_DELETION_TIMING_CHECKPOINT_2026-07-11.md).

## Project Identity

| Field | Value |
| --- | --- |
| Project name | Genshin Artifact Assistant |
| Status | Artifact-first native scanner with historical live-accepted dev 5/20/50/100, packaged bounded/UI paths, and a complete packaged 2,211-Artifact inventory soak; scanner-first Galaxy UI, explicit review/promotion gates, deterministic roll-efficiency, an English-default/Deutsch-selectable app UI, local-only artifact removal, durable timing evidence, and an offline-tested source-bound Build-Fit evidence-ranking layer. The current-build elevated-dev 50-artifact smoke now passes its functional/safety gate; the 3/s performance target remains separate and unproven. |
| Platform | Windows desktop |
| Target users | Genshin Impact players who want artifact decisions without complex optimizer setup |
| Runtime | Electron app with React UI and TypeScript |

## Vision

Genshin Artifact Assistant should make artifact management feel like a no-brainer. The user opens Genshin, runs a scan, and receives simple decisions: what is good, who can use it, what can probably be marked as trash, and which builds are currently available.

The app is not intended to replace deep min-max tools. It prioritizes time savings, confidence, and understandable recommendations over perfect theorycrafting.

## Goals

- Make artifact scanning stable enough that a normal user can trust it without babysitting every click.
- Build one local canonical Genshin data package for artifact sets, pieces, slots, stats, and characters.
- Parse artifact name, slot, main stat, substats, set, equipped state, and confidence deterministically against that package.
- Save weak or failed reads automatically as review samples and turn corrections into reusable local fixes.
- Present finished scan results as a compact artifact list instead of a debug-heavy live stats surface.
- Make scan readiness, the primary scan action, progress, and safe next steps understandable without opening technical views.
- Make the primary scan cover the recognized owned Artifact inventory by
  default, while allowing an explicit limit by Artifact count or grid rows.
- Keep extraction confidence separate from artifact value so uncertain OCR becomes review, not a misleading low score.
- Provide a browsable local artifact inventory with detail views before promoting broader recommendations.
- Die sichtbare Sternreihe einer nativen Artefaktkarte direkt erfassen: Nur
  explizit bestätigte 5★-Artefakte erhalten eine Roll-Efficiency-Bewertung;
  bestätigte 1–4★-Artefakte bleiben als manuell prüfbare Ergebnisse erhalten.
- Keep the app offline-first and usable without optimizer imports, Enka, HoYoLAB,
  or any external scanner as a core dependency.
- Re-introduce recommendations only after the scanner base is trustworthy.

## Non-Goals

- No memory reads, process hooks, game modification, packet inspection, or anti-cheat bypassing.
- Keine automatische Bergung, Quick-Select-Auswahl, Löschung, Fütterung,
  Verbesserung, Sperr-/Entsperraktionen oder Ressourcenverbrauch — auch nicht
  für erkannte 1–4★-Artefakte.
- No advanced formula editor or full power-user optimizer in the MVP.
- No cloud sync by default.

## Functional Requirements

| ID | Requirement | Priority | Status |
| --- | --- | --- | --- |
| FR-001 | List capture sources and automatically prefer the detected Genshin window when available. | Must | Implemented |
| FR-002 | Read one currently opened artifact reliably from the local screen and show its parsed result. | Must | Implemented |
| FR-003 | Generate and maintain a local canonical Genshin data package for sets, pieces, slots, stats, characters, aliases, and UI profiles. | Must | Implemented baseline |
| FR-004 | Parse artifact fields only through deterministic matching, validation, and derivation against the canonical package. | Must | Implemented baseline |
| FR-005 | Run a stable automatic inventory scan: detect grid, click tile, verify detail change, parse, store, continue, scroll, resume. | Must | Implemented for visible-inventory baseline |
| FR-006 | Save low-confidence, failed, conflicting, or stale scans automatically as review samples with reason codes. | Must | Implemented baseline |
| FR-007 | Apply local learned fixes from review corrections before every new parse. | Must | Implemented baseline |
| FR-008 | Keep the scan UI operator-friendly: readiness and primary action first, preview plus result rail during scanning, debug in collapsed or secondary surfaces. | Must | Implemented scanner-first baseline |
| FR-009 | Provide account-level artifact triage after scanner trust is acceptable. | Should | Pending |
| FR-010 | Provide 1-3 simple build suggestions per character from owned artifacts after scanner trust is acceptable. | Should | Pending |
| FR-011 | Farming overlay for reward scans. | Later | Prototype shell |
| FR-012 | Show active scan results as a minimal right-side rail with artifact number, name or compact fallback, value score, and status pill. | Should | Implemented baseline |
| FR-013 | Provide a scanned artifact inventory view with compact score pills, filters, sorting, and click-through detail. | Should | Implemented baseline |
| FR-014 | Provide artifact detail evaluation with screenshot/crops, parsed fields, OCR confidence, value reasons, and optional upgrade projection. | Should | Implemented; packaged visual acceptance passed, under-level projection remains corpus-backed |
| FR-015 | Provide clear loading, success, review, blocked, and error feedback through progress, spinners, skeletons, and mutation toasts without hiding durable state. | Should | Implemented UI baseline |
| FR-016 | Die sichtbare Sternreihe je nativer Artefaktkarte erfassen; nur explizit bestätigte 5★-Artefakte bewerten, bestätigte 1–4★-Artefakte ohne Wertung mit Bild und Details in einer separaten schreibgeschützten `Nicht 5★`-Ansicht zeigen. | Must | Implementiert; aktueller 50er-Smoke bestanden |
| FR-017 | Provide an English-default application interface and let the user switch the renderer to German in Settings without changing the scanner contract. | Should | Implemented; built-renderer acceptance passed, elevated packaged-executable runtime remains separate |
| FR-018 | Let the user remove an incorrectly scanned artifact from local app data only, including its local crop when safe, without any Genshin mutation. | Should | Implemented, unit tested, and confirmation paths exercised; no data-removing test was run |

## Non-Functional Requirements

| Category | Requirement | Measurement |
| --- | --- | --- |
| Safety | Never perform irreversible in-game actions. | Code review and manual test |
| Evaluation safety | Explizit visuelle 1–4★-Evidenz wird ausgeschlossen; fehlende, widersprüchliche oder unsichere Stern-Evidenz geht in `Review`. | Result-/Replay-Tests und Detailprüfung |
| Performance | Single artifact read should feel interactive and batch scan should not stall on false progress. | Capture latency monitored manually; auto-scan stops on blocked verification |
| Performance | A 100-artifact visible-inventory run should finish cleanly with low review/miss rates and report timing evidence. | `npm run scan:goal:validated` or `npm run scan:goal:validated:wait` quality-gated report |
| Privacy | Captures and parsed data stay local by default. | No remote upload in scanner path |
| Reliability | Uncertain OCR must be visible to the user. | Confidence and details view |
| Score integrity | Extraction confidence and artifact value are separate concepts. | Review state can block or qualify a value score |
| Learning loop | Scanner mistakes should become reusable local review samples. | `review-samples.jsonl` |
| Maintainability | Scanner heuristics must be isolated and documented. | Parser tests, scan-loop tests, data generator, review sample pipeline |
| Usability | The default path should expose one primary next action and keep advanced controls secondary. | Scanner-first navigation and desktop visual acceptance |
| Accessibility | Focus, status, loading, and motion must remain understandable without relying on animation or color alone. | Keyboard review, ARIA status semantics, and `prefers-reduced-motion` |

## Tech Stack

| Area | Choice | Notes |
| --- | --- | --- |
| Desktop shell | Electron 43 | Windows local app and overlay windows |
| Frontend | React + TypeScript + Vite | UI and client state |
| Styling | CSS design tokens with dark Galaxy glass surfaces | Cinematic depth with restrained violet, cyan, gold, and semantic state accents |
| OCR | Tesseract.js prototype plus deterministic normalization/derivation | OCR alone is not trusted as the decision source |
| Capture | Electron desktopCapturer plus Windows GDI Smart Capture | GDI path is used for Genshin Smart Capture reliability |
| Input automation | C# sidecar with elevated dev runner when needed | Live-validated for read-only inventory selection clicks; see `docs/AUTOMATION_LIVE_SCAN.md` |
| Tests | Vitest 4 + TypeScript checks | OCR eval, helper/cleanup safety, Build-Fit contract/combination/profile/ranking/context gates, localization/deletion/timing tests, packaged acceptance runners, and saved-run replay |
| Packaging | electron-builder | Configured in `package.json` |

## Runtime

| Environment | Entry Point | Notes |
| --- | --- | --- |
| Local dev | `npm run dev` | Starts Vite and Electron |
| Local dev with automation | `npm run dev:admin` | Required when `GenshinImpact.exe` is elevated; Windows blocks lower-integrity cursor/click input |
| Production build | `npm run build` | Builds React and Electron main process |
| Complete non-live gate | `npm run acceptance:offline` | Runs tests, OCR/assessment gates, saved-run validation, fresh package build/verification, audits, and diff checks |
| Unpacked package | `npm run package:offline-check` | Builds helper/app and verifies ASAR/resources without launching Genshin |
| Saved-run replay | `npm run scan:native:replay` | Re-evaluates the newest existing native scan without capture or input |
| Saved-run evidence set | `npm run scan:native:validate:saved` | Validates requested accepted targets (bounded 20/50/100 or full 2,211), files, review provenance, replay, and projection boundaries without capture or input |
| Preview | `npm run preview` | Browser preview only; capture bridge is unavailable |

## Current State Review

### What already works

- The app can enumerate capture sources and often identify the Genshin window automatically.
- Single-artifact capture is no longer blind full-screen OCR; it produces detail crops, OCR blocks, parsed fields, confidence, and notes.
- A local canonical data package already exists in `src/data/genshinGameData.json`, generated from `genshin-db`.
- The parser already uses known sets, pieces, slots, stat aliases, set aliases, character aliases, and derived slot/set mapping.
- Review samples, learned replacements, parser notes, and stored artifacts already persist locally.
- The scan surface has a current-session result rail that starts empty; older
  persisted Artifacts remain available in the dedicated inventory view.
- The auto-scan loop is no longer a naive click spammer: it has preflight, verification, miss handling, page fingerprinting, and stop conditions.
- The scanner now has a validated visible-inventory path: 32 safe artifact
  targets per page, lookup-derived fields, fast OCR crop profile, and a
  quality-gated live soak runner.
- Elevated live automation is validated in the current dev environment:
  `/automation/probe-click?index=1` changed the selected artifact and
  `/scanner/start?limit=2` completed with 2/2 verified reads and 0 misses.
- A 2026-07-08 visible-inventory 50-artifact run completed cleanly with
  50/50 parsed and stored, 0 review, 0 duplicates, and 0 misses. It is stable
  but still too slow for the 2-3 artifacts/second target.
- Later 2026-07-08 direct-GDI hot-path runs completed 20/20 parsed with
  0 misses and 0 review. The best clean 20-artifact iteration reached
  7285 ms, or roughly 2.75 artifacts/second; the final stable
  `2026-07-08-direct-gdi-reviewfix` run completed in 7973 ms. The
  3 artifacts/second target remains unproven.
- The same direct-GDI path completed a 100-artifact run with 100/100 parsed,
  0 review, 0 misses, and 42064 ms elapsed across 4 pages.
- Review samples can now be exported with `npm run eval:review-candidates` into
  a Git-ignored human-labeling worklist. This is the next quality phase before
  adding more OCR corpus cases or trusting review queue data as labels.
- Native IK-style Artifact capture is wired through the C# helper. It writes
  card crops and run artifacts for downstream OCR/parse processing instead of
  making React do per-artifact work in the hot capture loop.
- The native result processor tails flushed capture jobs while the helper is
  still scanning, evaluates them with four bounded workers, publishes ordered
  live deltas, and writes incremental `scan-results.json` plus
  `processing-report.json`. Parser confidence and IK set/piece/slot conflicts
  remain reviewable, and store persistence stays opt-in.
- The Inventory view now has an Artifact-only pipeline surface for scope,
  native capture, OCR queue, review gate, promotion, and evidence. Native rows
  expose crop previews, IK/GOOD metadata, dry-run promotion state, and a
  `Naechster Schritt` card.
- The active UI intentionally hides weapon, material, and character-detail IK
  catalog coverage until those values are actually scanned.
- Clean Artifact results now receive deterministic `roll-efficiency-v1`
  evaluations. Inventory/detail show score reasons and legal roll breakdown;
  build fit remains explicitly deferred.
- Under-leveled, unambiguously identified 5-star Artifacts with four known
  substats receive a detail-only Worst/Middle/Best projection. 4-star,
  rarity-ambiguous, incomplete, or review data receives none.
- Existing 20/50/100 native results passed five-repeat offline determinism
  replay; the approved 100-result run evaluated 100/100 with zero unknown.
- An Electron 43 unpacked app passed 27/27 static package checks and live launch
  from `%TEMP%`. ASAR renderer, preload, IPC, packaged C# helper, IK resources,
  elevation telemetry, Inventory UI, and a packaged 5-result dry run passed.
- The renderer now uses a scanner-first information architecture: Scanner,
  Artifacts, Review, and the clearly labeled Builds preview form the workspace;
  Overlay and Diagnostics remain secondary tools.
- The Scanner surface leads with Genshin/app readiness, one primary start/stop
  action, a three-step empty state, explicit scan phases, and one modal for
  scan scope, capture source, and manual tools. Inventory loading uses
  skeletons/spinners, and confirmed
  promotion, review, import, and export mutations report through accessible
  toasts.
- Shared Galaxy tokens, focus-visible treatment, micro-motion, and a
  `prefers-reduced-motion` fallback now define the visual/feedback baseline.
- The renderer now defaults to English on a fresh, missing, or invalid locale
  setting. Users can choose German from Settings; the preference is local to the
  app and updates the document language. This changes application copy only:
  Smart Capture remains validated for the English Genshin UI.
- Artifact detail now exposes a small, confirmed local-delete action. Native
  scan results are tombstoned rather than rewritten, their crop is removed only
  when it is contained in the selected run directory, and an exact linked Store
  record is removable only through an explicit local confirmation. No delete
  path calls the input helper or changes Genshin.
- Native runs now retain an atomic `run-timing.json` record from request start
  through durable result reconciliation. It measures capture and processing
  separately and derives end-to-end duration from one wall-clock interval
  instead of adding overlapping phases.
- The 136 historical full-inventory Review samples were audited against their
  crops. The confirmed corpus now distinguishes 128 explicit non-five-star
  entries, six exact duplicated-decimal repairs, one slot-safe artifact-name
  correction, and one incomplete entry that intentionally remains `Review`.

### What is still structurally weak

- The scan experience is still partly orchestrated from `src/App.tsx`, which makes behavior changes harder than they should be.
- Broader scan soak testing has reached clean 20-, 45-, and 100-artifact runs
  with 0 misses on the current engine. The current 2026-07-09 100-artifact run
  completed `100/100` verified and parsed with `0` review and `0` misses.
- OCR quality is still inconsistent enough that some fields are recovered by fallback and derivation more often than they should be.
- Learned fixes currently focus on text replacements; they do not yet update crop offsets, UI profile variants, or scanner targeting rules in a structured way.
- The scan page now has a minimalist recent-results rail, and the Inventory view
  can inspect native Artifact results, crop previews, IK match state, promotion
  dry-runs, pipeline risk/status, roll-efficiency reasons, and upgrade
  projection. Single-result review/edit/approve is implemented and batch
  review is intentionally unavailable.
- The native IK-style path has fresh later-session 5/20/50/100 scale evidence,
  deterministic replay, focus-loss and stop evidence, plus a packaged 5-result
  runtime smoke. Non-16:9 and other UI-language profiles remain open.
- Store promotion from native `scan-results.json` now supports one selected,
  confirmed result at a time with main-process revalidation and a durable log.
  Batch promotion intentionally remains unavailable.
- Native Review results can be corrected and approved or rejected one at a
  time. Approval revalidates IK identity, canonical main values, and legal
  substat rolls, then feeds the existing review-to-eval candidate pipeline.
- Recommendations and build logic exist, but build-aware fit must be designed
  separately from the now-implemented context-free roll-efficiency score.
- The Builds navigation item now reads only the newest complete native run and
  can surface up to three source-bound, read-only profile suggestions. It must
  not imply character ownership, team, weapon, rotation, DPS, or optimizer
  certainty; the built-renderer acceptance now passes, while an elevated
  packaged-executable runtime remains a separate packaging check.
- The historical 136-item Review queue is not a source of labels by itself.
  Its audit added only visually confirmed parser/corpus cases; the one
  incomplete crop remains fail-closed in `Review`.
- Repeatability and 3 artifacts/second are still open; speed work should not
  outrank result clarity, inventory UX, or corpus growth while the current path
  is stable.

### Current product conclusion

The app has crossed from OCR-demo/prototype into an Artifact-first scanner app.
The broadest proven live path is still the visible-inventory scan flow: the
operator opens Artifact inventory with a visible detail card, the app verifies
the state, scans read-only, persists parsed artifacts, and keeps uncertain data
reviewable. The newer native IK-style path is the intended high-speed direction:
the helper captures Artifact card crops quickly, while OCR, parsing, review,
promotion, and value evaluation run downstream.

The next product phase is not broad category expansion. Weapons, materials, and
character details stay out of active scope. Packaged/live acceptance passed on
2026-07-10 for the bounded 5/20/50/100 scanner and packaged 5er. A later real
Settings-UI run also accepted the revised Artifact-limit path at 5/5 through
preflight, capture, processing, parsing, and the session result rail with no
review, processing error, or persistence. The packaged Settings UI then
accepted a one-row limit as an eight-Artifact target at 8/8 with the same safe
outcome. The complete Settings-UI full-owned-inventory soak then accepted the
OCR-owned target 2,211 with exact 2,211 captures/results across 70 pages. After
one evidence-driven parser correction, the current packaged reprocess completed
2,211/2,211 with 136 Review (6.15%), zero errors/writes, and zero saved-run
validation issues. The current Build-Fit checkpoint now has four sourced
profiles, profile-expiry gates, a deterministic non-Roll-Efficiency evidence
ranker, explicit session-only Furina context, and a read-only Builds surface.
The renderer now defaults to English with a German Settings option, and local
artifact removal is deliberately confined to app-local result/store records and
contained crops. The current scanner also writes durable shared-clock timing
evidence and has a visually reviewed corpus correction for the audited
136-entry Review set. These current-build changes are offline/unit validated,
and the built-renderer Builds/UI acceptance passed, including the local-delete
confirmation probes. That is not an elevated packaged-executable claim. The
current elevated development 50-artifact native smoke now passes with complete
timing evidence and zero writes; its `28,022 ms` end-to-end result does not
prove the separate 3/s target. The remaining Review corpus continues to grow
only from visually confirmed labels. The review exporter now prioritizes
retrievable retained crops and serializes only safe capture/run references; its
confirmation preparer refuses OCR-only candidates so parser guesses cannot be
promoted without visual evidence.

## Product Direction

- Artifact scanning is the first-class feature.
- Character optimization returns only after scan quality is trustworthy.
- Team building stays out of the critical path until artifact ingestion is stable.
- External scanners are not runtime dependencies. The app owns its scan
  choreography, OCR, and quality gates.
- Self-learning stays deterministic and local first: review samples, aliases, crop offsets, and UI profile tuning before any ML retraining discussion.
- The scan workspace should be an operator surface: the left area shows preview
  when idle and owns capture-versus-target progress during a run; the right
  rail contains current-session results plus independent evaluation progress.
  Stop, status, and review access remain visible. Detailed debug evidence
  belongs in diagnostics, summaries, or artifact detail.
- Primary navigation should follow the user journey: scan, inspect artifacts,
  review uncertainty, then enter future build guidance. Overlay and diagnostics
  remain tools instead of competing with that journey.
- Feedback should match the action: progress for long-running scans, skeletons
  for content loading, inline state for durable blockers, and toasts for the
  outcome of explicit mutations.
- Artifact value scoring must not hide OCR uncertainty. `Review` is a distinct
  outcome, not just a weak artifact score.
- Explizit bestätigte visuelle 1–4★-Artefakte sind keine
  Bergungs-/Lösch-Empfehlung: Sie erscheinen nur in der schreibgeschützten
  `Nicht 5★`-Ansicht mit Bild und Details zur manuellen Prüfung. Eine
  fehlende oder widersprüchliche Sternreihe bleibt `Review`.

## Execution Plan

### Phase 0 - Stabilize the operator surface

Outcome:
- Scan page reduced to source, main preview, result panel, primary scan actions, and compact status.
- Diagnostics, logs, crops, confidence breakdown, review queue, and learning internals moved behind modal or drawer entry points.
- Scan completion popup summarizes scanned, stored, duplicates, review samples, blocked reason, and elapsed time.

Status:
- Implemented as the scanner-first Galaxy baseline. Readiness and the primary
  scan action lead the page, the left surface owns capture progress, and the
  larger right surface owns ordered live results plus compact independent
  evaluation progress. The current-session result rail starts empty, advanced
  controls open in a modal without shifting the workspace, and Diagnose remains
  a separate scrollable tool. The
  Artifact-limit path and consolidated progress/result layout passed a real
  five-result Settings-UI run, the one-row path at 8/8, and the full-owned path
  at 2,211/2,211 with current-package Review below the 15% gate.

### Phase 1 - Canonical game data package

Outcome:
- `scripts/generate-genshin-data.cjs` emits one stricter package contract for:
  - artifact sets
  - artifact pieces
  - slot-by-piece mapping
  - stats and allowed mains by slot
  - characters
  - aliases
  - UI profiles
  - source version metadata
- Parser regression tests run against saved review samples and known bad cases.
- Parser stops "free guessing" outside the canonical package.

Status:
- Implemented as a generated lookup package in `src/data/genshinGameData.json`
  with validation and parser integration. Continue regenerating and expanding
  aliases deliberately when Genshin data or OCR samples require it.

### Phase 2 - Deterministic parser hardening

Outcome:
- Name, slot, set, main stat, and equipped fields are parsed through layered validation:
  1. direct OCR cleanup
  2. alias normalization
  3. exact package match
  4. constrained fuzzy match
  5. safe derivation from piece/slot/value references
- Main stat/value inference is tightened with slot constraints and reference tables.
- Bad parses automatically generate structured review reasons.
- Equipped-character parsing is canonical-data constrained: noisy known names can
  match through aliases/fuzzy lookup, but unknown footer fragments stay
  `Not detected` instead of being stored as invented character names.

Status:
- Implemented for the merge baseline. Parser tests cover canonical set/slot/stat
  matching, equipped-character footer noise, known aliases, and unsafe one-letter
  fragments. Continue growing the confirmed review corpus before tightening
  thresholds further.

### Phase 3 - Scanner core rebuild

Outcome:
- Auto-scan becomes a dedicated engine with explicit states:
  - preflight
  - grid detection
  - click target
  - wait stable
  - detail verify
  - parse
  - store or review
  - next tile
  - row scroll
  - resume or stop
- Progress counts only when a new verified artifact or duplicate signature is confirmed.
- Repeated pages, unchanged detail cards, blocked cursor movement, and scroll failures stop the scan with diagnosis instead of producing fake progress.
- Fast artifact-read captures include the equipped footer when an equipped
  marker is visible; preflight and polling captures still skip expensive OCR.

Status:
- Implemented and merged for the visible-inventory path. Live validation on
  2026-07-09 covered `20/20` verified/parsed with `0` review and `0` misses,
  equipped-character persistence, unlocked lock state, positive locked state,
  and locked persistence. Explicit entry-mode experiments remain separate from
  the normal merge-ready path.

### Phase 4 - Input automation replacement

Outcome:
- Replace the production automation path with a persistent Windows sidecar dedicated to:
  - focus
  - move
  - click
  - scroll
  - capture
  - probe
- Session probe decides which input mode works before auto-scan is unlocked.
- Auto-scan never starts on a session that cannot prove one successful detail-card change.

Status:
- Read-only C# helper path and elevated dev startup are validated for the
  visible-inventory scanner baseline. Direct inventory, Paimon-menu, and
  auto-entry modes remain Dev-Control experiments and should be tested with low
  limits before being promoted.

### Phase 5 - Learning loop that actually compounds

Outcome:
- Weak scans save themselves as review samples automatically.
- User corrections update local:
  - text replacements
  - alias maps
  - crop offsets
  - UI profile adjustments
  - constrained set/piece/slot fixes
- "Apply learned fixes" runs before every parse.
- Review samples become both parser regression fixtures and learning inputs.

Status:
- Prepared in code for text replacements, field aliases, constrained fixes,
  crop adjustment proposals, and UI-profile adjustment proposals. Crop/profile
  changes still require live review before being auto-applied. The current
  review-corpus checkpoint adds only visually confirmed cases: exact duplicate
  decimal punctuation is repaired, and a partial artifact name is canonicalized
  only when independent, high-confidence slot evidence leaves one safe match.
  Ambiguous or incomplete crops remain `Review`.

### Phase 6 - Minimal scan result rail and artifact inventory

Outcome:
- The scan page shows a large screenshot/preview and a compact right-side rail
  of finished artifact evaluations.
- Each row shows scan number, artifact name or compact fallback, value score,
  and a colored status pill.
- Debug stats, confidence breakdowns, and OCR internals move out of the main
  scan surface.
- A new artifact inventory menu provides compact browsing, filters, sorting,
  and click-through details.

Status:
- Started with a scan result rail in the scan surface and an `Inventory` view.
  The rail shows native `scan-results.json` entries from the current app scan
  session after post-processing and does not preload older stored/native rows
  at startup. The inventory browser can filter,
  sort, inspect native/store/snapshot rows, and preview native card crops from
  the selected run directory. It also exposes the current IK inventorylist
  version, active Artifact-only scope, and pipeline status for capture,
  post-processing, review, promotion, and evidence. Native Artifact rows carry
  IK/GOOD match status from the post-capture processor and a dry-run promotion
  decision that shows whether a native result is speicherbar, already stored,
  review-only, or blocked without writing to the store. Detail review,
  deterministic roll-efficiency reasons, evaluated/open filters, score sorting,
  and under-level projection are implemented. The browser prioritizes the
  collection, filters, selected detail, and explicit next actions; technical
  pipeline evidence remains available in collapsible detail. See
  [scanner-results-inventory-roadmap.md](scanner-results-inventory-roadmap.md)
  for the implementation phases and acceptance criteria. The separate,
  read-only `Nicht 5★` filter retains explicitly recognized 1–4★ native
  results with their crop and details, excludes them from roll evaluation, and
  exposes no in-game action. A confirmed local-delete action can remove a
  wrong native result from the visible local inventory through an append-only
  tombstone and safely contained crop cleanup; it can remove a Store-only row
  or exact linked Store record only locally and never changes Genshin.

### Phase 7 - Artifact detail evaluation and upgrade projection

Outcome:
- Artifact detail explains parsed fields, OCR confidence, scoring reasons, and
  review needs.
- Upgrade projection is available only for unambiguously identified 5-star data
  with four known legal substats and is labeled as probabilistic, with
  worst/middle/best projected value scores.
- Low-confidence OCR disables or qualifies value conclusions instead of showing
  false certainty.

Status:
- Implemented as `roll-efficiency-v1` plus detail-only
  `roll-efficiency-projection-v1`. Static/build/replay, fresh dev 5/20/50/100,
  packaged runtime, and final Inventory visual acceptance pass. The new live
  100 contained only +20 Artifacts, so under-level projections remain covered
  by the confirmed saved-run corpus rather than a new live sample.

### Phase 8 - Recommendations come back on top of a trusted scanner

Outcome:
- Account snapshot and build suggestions are only promoted once scan quality is high enough to trust owned artifacts.
- Recommendations explain uncertainty and surface conflicts instead of pretending perfect certainty.

Status:
- `build-fit-contract-v1` continues to keep per-Artifact and five-piece
  assessments score-free. On top of it, the offline-tested
  `build-fit-evidence-ranking-v1` selects at most three non-overlapping safe
  combinations from the newest complete native run. The sourced corpus now
  covers Furina Solo-Hydro C0-C1, Neuvillette Charged Attack, Raiden off-field
  Hyperbloom, and Nahida off-field Deepwood. Profiles expire explicitly;
  Roll-Efficiency and OCR confidence are not ranking inputs. Furina's required
  whole-build ER target accepts only explicit, session-only user context and
  otherwise stays deferred. The Builds surface explains Set/Main-Stat/Substat
  coverage, sources, warnings, and uncertainty without writing data or sending
  game input. A built-renderer acceptance now covers fresh English locale,
  German switching/restoration, Builds loading/context/focus, reduced motion,
  overflow, and native plus Store local-delete confirmation probes. It is not evidence of an
  elevated packaged-executable launch; that runtime boundary remains separate.
  See
  [BUILD_FIT_CONTRACT_V1.md](BUILD_FIT_CONTRACT_V1.md).

## Immediate Next Validation Order

1. The final current-build offline gate passed after the localization, deletion,
   timing, corpus, and acceptance-runner changes: `npm run acceptance:offline`
   completed TypeScript, 72 test files / 444 tests, OCR evaluation, saved-run
   validation, package verification (27/27), both dependency audits, and the
   diff check.
2. Preserve the passing built-renderer acceptance as renderer evidence, but
   separately verify an elevated packaged-executable launch/UAC bridge before a
   binary release claim. It is not a substitute for scanner smoke evidence.
3. The current elevated development 50-artifact smoke passed with full
   reconciliation, zero Review/errors/writes, and
   `requestToResultsReconciledMs=28,022`. This is a functional/safety gate;
   it does not claim 3 Artifacts/s.
4. Completed on 2026-07-11: generated workspace output was cleaned and the
   coherent checkpoint was committed and pushed as `8b9f948`. A
   package-runtime/UAC test is required only for that separate claim.
5. The visual-evidence review exporter/preparer passed focused checks plus
   `npm run lint`, `npm test` (72 files / 446 tests), and `npm run build`.
   Its read-only export of the current 177-record local queue found three
   retrievable crops, all already-confirmed `native-review-approved` cases;
   no new corpus label was promoted.
6. Continue review-corpus growth only from new visually confirmed crops. Keep
   explicit 1–4★ evidence excluded, ambiguous evidence in `Review`, and reject
   aliases that cannot be justified by independent field evidence.
7. Refresh source-bound profiles before expiry and add Golden plus
   Review/low-confidence adversarial fixtures for each additional profile.
8. Keep the bounded wheel reset as the production baseline. Benchmark a `B`
   plus Artifact-tab reopen only as a separate guarded experiment before any
   reset strategy change.
9. Keep Build-Fit profile suggestions read-only and Artifact-only until an
   explicit product decision adds owned-character, weapon, team, and rotation
   inputs with their own sources.

## Open Questions

| Question | Status |
| --- | --- |
| Is the current C# helper sufficient for production packaging, or does a later Rust/C++ sidecar still materially reduce latency or packaging risk? | Open |
| When should UI-profile learning be allowed to change crop geometry automatically versus requiring review approval? | Open |
| What scan-quality threshold is high enough before profile suggestions can be called live-accepted? | Scanner gate passed for accepted 16:9 English Artifact scope; Build-Fit unit/repository gate and built-renderer acceptance passed, while elevated packaged-executable runtime and current scanner smoke remain separate |
| What exact first-pass value formula should drive the `0-100` artifact score before build-aware recommendations exist? | Resolved for V1: legal roll-tier efficiency only; build fit remains separate |
| Which upgrade projection model is honest enough for early UX: deterministic roll buckets, probability-weighted outcomes, or a deliberately simple best/middle/worst estimate? | Resolved for V1: legal Worst/Middle/Best roll-efficiency bounds with explicit uncertainty |
| Which Genshin UI languages should be supported after English once the scanner contract is stable? | Open |