feat(scanner): add native artifact pipeline

Add native IK-style capture processing, Artifact Inventory, explicit promotion and single-result review. Confirm the three live OCR corrections in the eval corpus and preserve extraction/value separation.
This commit is contained in:
AzuTear
2026-07-09 23:30:42 +02:00
parent 28d60eb915
commit 639b0b7f59
92 changed files with 13606 additions and 1703 deletions
+355 -45
View File
@@ -6,17 +6,33 @@ import path from "node:path";
import { fileURLToPath } from "node:url";
import { createWorker, PSM } from "tesseract.js";
import { createInputHelperService, type InputHelperService } from "./services/inputHelper.js";
import {
createNativeScannerProcessingService,
nativeScannerProcessStats,
type NativeCaptureJobPayload,
} from "./services/nativeScannerProcessingService.js";
import { pngBufferToBitmap } from "./services/pngBitmap.js";
import { createRepositoryContext, type RepositoryContext } from "./bootstrap/repositoryContext.js";
import { registerIpcHandlers } from "./bootstrap/ipcBootstrap.js";
import { createDevControlServer } from "./devControlServer.js";
import { createAppWindowManager, type AppWindowManager } from "./appWindowManager.js";
import { createGoodFileService, type GoodFileService } from "./services/goodFileService.js";
import type { IkArtifactCatalog } from "../src/lib/ikArtifactMatcher.js";
import type { AppSnapshot } from "../src/types/domain.js";
import type {
CaptureOptions,
CaptureResult,
GoodDatabase,
NativeScannerCatalogStatus,
NativeScannerDataStatus,
NativeScannerImageLoadStatus,
NativeScannerPreflightStatus,
NativeScannerProcessStatus,
NativeScannerPromotionStatus,
NativeScannerReviewArtifactInput,
NativeScannerReviewStatus,
NativeScannerResultsLoadStatus,
NativeScannerRunStatus,
OcrResult,
AppRuntimeInfo,
ScannerCommand,
@@ -54,7 +70,7 @@ app.commandLine.appendSwitch("disable-gpu-sandbox");
const __dirname = path.dirname(fileURLToPath(import.meta.url));
const isDev = Boolean(process.env.VITE_DEV_SERVER_URL);
const APP_RUNTIME_STARTED_AT = new Date().toISOString();
const APP_RUNTIME_SIGNATURE = "2026-07-08-direct-gdi-reviewfix";
const APP_RUNTIME_SIGNATURE = "2026-07-09-native-artifact-pipeline";
let registeredHotkeys: Record<string, boolean> = {};
let devControlServer: Server | null = null;
@@ -214,6 +230,305 @@ async function publishScannerStatus(status: ScannerStatusPayload) {
return { ok: true };
}
function nativeScannerDataDir() {
const candidates = [
process.env.IK_INVENTORYLISTS_DIR,
path.join(process.cwd(), "data", "ik-inventorylists"),
path.join(app.getAppPath(), "data", "ik-inventorylists"),
path.join(process.resourcesPath, "ik-inventorylists"),
].filter((candidate): candidate is string => Boolean(candidate));
return candidates.find((candidate) => existsSync(path.join(candidate, "version.txt"))) ?? candidates[0];
}
function nativeScannerOutputRoot() {
return path.join(app.getPath("userData"), "native-scans");
}
function nativeScannerStats(status: NativeScannerRunStatus): Record<string, number> {
return {
clicked: Number(status.clicked ?? 0),
attempted: Number(status.captured ?? 0),
verified: Number(status.captured ?? 0),
parsed: 0,
stored: 0,
review: 0,
duplicates: 0,
misses: 0,
pages: Number(status.pages ?? 0),
elapsedMs: Number(status.activeMs ?? 0),
activeScanMs: Number(status.activeMs ?? 0),
queued: Number(status.queued ?? 0),
captured: Number(status.captured ?? 0),
};
}
function publishNativeScannerStatus(status: NativeScannerRunStatus, extra: Partial<ScannerStatusPayload> = {}) {
const summary = {
mode: "Native IK Scanner",
status: status.status,
...nativeScannerStats(status),
targetCount: status.target,
gridLabel: status.message,
runId: status.runId,
outputRoot: status.outputRoot,
runDir: status.runDir,
manifestPath: status.manifestPath,
jobsPath: status.jobsPath,
lastArtifactPath: status.lastArtifactPath,
};
scannerDevStatus = {
...scannerDevStatus,
running: Boolean(status.running),
reviewStatus: status.message || "Native scanner ready.",
captureStatus: status.status,
stats: nativeScannerStats(status),
summary,
automationLog: [
...(scannerDevStatus.automationLog ?? []).slice(-10),
status.message || `native scanner ${status.status}`,
],
nativeScanner: status,
...extra,
updatedAt: new Date().toISOString(),
};
return scannerDevStatus;
}
async function nativeScannerDataStatus(): Promise<NativeScannerDataStatus> {
return getInputHelperService().nativeScannerDataStatus(nativeScannerDataDir());
}
async function nativeScannerCatalog(): Promise<NativeScannerCatalogStatus> {
return getInputHelperService().nativeScannerCatalog(nativeScannerDataDir());
}
async function loadNativeIkArtifactCatalog(): Promise<IkArtifactCatalog | null> {
const catalog = await nativeScannerCatalog();
return catalog.data.valid ? { artifacts: catalog.artifacts } : null;
}
async function nativeScannerPreflight(options: { category?: string } = {}): Promise<NativeScannerPreflightStatus> {
const preflight = await getInputHelperService().nativeScannerPreflight(nativeScannerDataDir(), options.category ?? "artifacts");
scannerDevStatus = {
...scannerDevStatus,
nativeScannerPreflight: preflight,
lookupStatus: preflight.data,
updatedAt: new Date().toISOString(),
};
return preflight;
}
async function nativeScannerStart(options: { limit?: number; category?: string } = {}): Promise<NativeScannerRunStatus> {
const status = await getInputHelperService().nativeScannerStart({
dataDir: nativeScannerDataDir(),
outputRoot: nativeScannerOutputRoot(),
limit: options.limit ?? 100,
category: options.category ?? "artifacts",
});
publishNativeScannerStatus(status, { lookupStatus: await nativeScannerDataStatus().catch(() => undefined) });
return status;
}
async function nativeScannerStop(): Promise<NativeScannerRunStatus> {
const status = await getInputHelperService().nativeScannerStop();
publishNativeScannerStatus(status);
return status;
}
async function nativeScannerStatus(): Promise<NativeScannerRunStatus> {
const status = await getInputHelperService().nativeScannerStatus();
publishNativeScannerStatus(status);
return status;
}
function resolveNativeProcessRunDir(runDir?: string) {
const native = scannerDevStatus.nativeScanner as NativeScannerRunStatus | undefined;
const candidate = runDir?.trim() || native?.runDir || "";
if (!candidate) return "";
const root = path.resolve(nativeScannerOutputRoot());
const resolved = path.resolve(candidate);
const relative = path.relative(root, resolved);
if (relative.startsWith("..") || path.isAbsolute(relative)) return "";
return resolved;
}
async function buildNativeCardCaptureResult(
imagePath: string,
job: NativeCaptureJobPayload,
): Promise<CaptureResult> {
const sourceImage = nativeImage.createFromPath(imagePath);
const size = sourceImage.getSize();
if (!size.width || !size.height) throw new Error(`Native card crop is empty: ${imagePath}`);
const detailRect = { x: 0, y: 0, width: size.width, height: size.height };
const inventoryRect = { x: 0, y: 0, width: 1, height: 1 };
const crops = createCrops(
sourceImage,
size,
detailRect,
inventoryRect,
{
ocrMode: "artifact",
ocrProfile: "full",
omitFullFrame: true,
omitInventoryPreview: true,
omitCropImages: true,
omitLockState: true,
},
undefined,
{ skipOcr: false },
);
const croppedPayload = crops
.filter((crop) => crop.ocrEnabled !== false)
.map((crop) => crop.ocrImage
? {
id: crop.id,
label: crop.label,
image: crop.ocrImage,
}
: null)
.filter((crop): crop is OcrCropPayload => Boolean(crop));
const recognized = await runOcrOnCropsWithTimeout(croppedPayload, "current", 6500);
return {
id: `native-card-${job.sequence}`,
name: path.basename(imagePath),
width: size.width,
height: size.height,
dataUrl: "",
capturedAt: new Date().toISOString(),
captureTarget: "genshin-client",
detailFingerprint: `${path.basename(imagePath)}:${size.width}x${size.height}`,
ocr: recognized.ocr,
ocrTimedOut: recognized.timedOut,
ocrSkipped: false,
crops: crops.map((crop) => ({
id: crop.id,
label: crop.label,
rect: {
x: crop.rect.x,
y: crop.rect.y,
width: crop.rect.width,
height: crop.rect.height,
},
dataUrl: crop.dataUrl,
})),
artifactDetail: {
present: true,
confidence: 70,
orangeHits: 0,
greenHits: 0,
textHits: 0,
},
layout: {
aspect: aspectRatioLabel(size),
isSixteenNine: false,
warning: "Native card-crop processing uses the detail card as its own coordinate space.",
},
timings: {
prepareMs: 0,
ocrMs: 0,
totalMs: 0,
ocrProfile: "full",
ocrEngine: "current",
ocrWorkerPoolSize: OCR_WORKER_POOL_SIZE,
cropCount: croppedPayload.length,
ocrSkipped: false,
},
};
}
function createNativeScannerService() {
return createNativeScannerProcessingService({
resolveRunDir: resolveNativeProcessRunDir,
buildCaptureResult: buildNativeCardCaptureResult,
loadArtifacts: () => getArtifactStoreRepository().loadAll(),
saveArtifacts: (records) => getArtifactStoreRepository().saveMany(records),
loadIkArtifactCatalog: loadNativeIkArtifactCatalog,
saveReviewSample: (sample) => getReviewSamplesRepository().append(sample),
});
}
async function nativeScannerProcessRun(options: { runDir?: string; persist?: boolean; limit?: number } = {}): Promise<NativeScannerProcessStatus> {
const service = createNativeScannerService();
const status = await service.processRun(options);
scannerDevStatus = {
...scannerDevStatus,
reviewStatus: `Native post-processing: ${status.parsed}/${status.processed} parsed, ${status.review} review, ${status.stored} stored.`,
stats: {
...(scannerDevStatus.stats ?? {}),
...nativeScannerProcessStats(status),
},
nativeScannerProcessing: status,
automationLog: [
...(scannerDevStatus.automationLog ?? []).slice(-10),
`native process: ${status.parsed}/${status.processed} parsed, report ${status.reportPath}`,
],
updatedAt: new Date().toISOString(),
};
return status;
}
async function nativeScannerLoadResults(options: { runDir?: string; limit?: number } = {}): Promise<NativeScannerResultsLoadStatus> {
const service = createNativeScannerService();
const loaded = await service.loadResults(options);
scannerDevStatus = {
...scannerDevStatus,
nativeScannerResults: loaded,
automationLog: [
...(scannerDevStatus.automationLog ?? []).slice(-10),
`native results: ${loaded.results.length}/${loaded.total} loaded from ${loaded.path || "scan-results.json"}`,
],
updatedAt: new Date().toISOString(),
};
return loaded;
}
async function nativeScannerPromoteResults(options: { runDir?: string; resultIds: string[] }): Promise<NativeScannerPromotionStatus> {
const service = createNativeScannerService();
const status = await service.promoteResults(options);
scannerDevStatus = {
...scannerDevStatus,
reviewStatus: status.ok
? `Native promotion: ${status.promoted} promoted, ${status.alreadyStored} already stored.`
: `Native promotion blocked: ${status.error ?? "unknown error"}`,
nativeScannerPromotion: status,
automationLog: [
...(scannerDevStatus.automationLog ?? []).slice(-10),
`native promotion: ${status.promoted}/${status.requested}, log ${status.logPath || "unavailable"}`,
],
updatedAt: new Date().toISOString(),
};
return status;
}
async function nativeScannerReviewResult(options: {
runDir?: string;
resultId: string;
action: "approve" | "reject";
artifact?: NativeScannerReviewArtifactInput;
note?: string;
}): Promise<NativeScannerReviewStatus> {
const service = createNativeScannerService();
const status = await service.reviewResult(options);
scannerDevStatus = {
...scannerDevStatus,
reviewStatus: status.ok
? `Native review ${status.action}: ${status.resultId}`
: `Native review blocked: ${status.error ?? "unknown error"}`,
nativeScannerReview: status,
automationLog: [
...(scannerDevStatus.automationLog ?? []).slice(-10),
`native review ${status.action}: ${status.ok ? "ok" : status.error}`,
],
updatedAt: new Date().toISOString(),
};
return status;
}
async function nativeScannerLoadImage(options: { runDir?: string; imagePath: string }): Promise<NativeScannerImageLoadStatus> {
const service = createNativeScannerService();
return service.loadImage(options);
}
async function readRuntimeInfo() {
try {
const result = await getInputHelperService().getRuntimeInfo();
@@ -446,6 +761,9 @@ function focusMainWindow() {
}
function sendScannerCommand(command: ScannerCommand | "probe-click") {
if (command === "stop") {
void nativeScannerStop().catch(() => undefined);
}
getAppWindowManager().sendScannerCommand(command);
}
@@ -468,6 +786,15 @@ function startDevControlServer() {
sendScannerCommand,
clickScreen: clickScreenCommand,
scannerStatus: () => ({ ...scannerDevStatus, appBuild: appRuntimeInfo(), ocrWarmup: getOcrWarmupStatus() }),
nativeScannerDataStatus,
nativeScannerCatalog,
nativeScannerPreflight,
nativeScannerStart,
nativeScannerStop,
nativeScannerStatus,
nativeScannerProcessRun,
nativeScannerLoadResults,
nativeScannerLoadImage,
warmOcr: (engine) => warmOcrWorkerPool(engine),
loadReviewSamples,
listCaptureSources,
@@ -483,15 +810,13 @@ function createOverlayWindow() {
getAppWindowManager().createOverlayWindow();
}
// Inventory Kamera keeps a pool of native Tesseract engines and scans artifact
// fields concurrently. Our fast artifact profile has four useful OCR parameter
// groups, so the default pool is four workers unless the machine is smaller or
// GAA_OCR_WORKERS explicitly overrides it.
// The fast artifact profile has four useful OCR parameter groups, so the
// default pool is four workers unless the machine is smaller or GAA_OCR_WORKERS
// explicitly overrides it.
const OCR_WORKER_POOL_SIZE = resolveOcrWorkerPoolSize();
const IK_TRAINEDDATA_LANG = "genshin_fast_09_04_21";
type OcrWorker = Awaited<ReturnType<typeof createWorker>>;
type OcrWorkerEngine = "current" | "ik-traineddata";
type OcrWorkerEngine = "current";
type OcrCropPayload = { id: string; label: string; image: Buffer };
type OcrWarmupStatus = {
engine: OcrWorkerEngine;
@@ -513,11 +838,9 @@ type OcrWorkerPoolState = {
const ocrWorkerPools: Record<OcrWorkerEngine, OcrWorkerPoolState> = {
current: { poolPromise: null, runQueue: Promise.resolve() },
"ik-traineddata": { poolPromise: null, runQueue: Promise.resolve() },
};
const ocrWarmupStatuses: Record<OcrWorkerEngine, OcrWarmupStatus> = {
current: { engine: "current", status: "cold", workerPoolSize: OCR_WORKER_POOL_SIZE },
"ik-traineddata": { engine: "ik-traineddata", status: "cold", workerPoolSize: OCR_WORKER_POOL_SIZE },
};
function resolveOcrWorkerPoolSize() {
@@ -541,42 +864,13 @@ function appRuntimeInfo(): AppRuntimeInfo {
}
function ocrEngineFromOptions(options: CaptureOptions = {}): OcrWorkerEngine {
return options.ocrEngine === "ik-traineddata" ? "ik-traineddata" : "current";
}
function ikTessdataCandidates() {
const envDir = process.env.IK_TESSDATA_DIR;
return [
envDir,
path.resolve(process.cwd(), "data", "tessdata"),
path.resolve(process.cwd(), "work", "Inventory_Kamera", "InventoryKamera", "tessdata"),
path.resolve(process.cwd(), "work", "refs", "Inventory_Kamera", "InventoryKamera", "tessdata"),
path.resolve(process.cwd(), "..", "_ik_ref_fork", "InventoryKamera", "tessdata"),
path.resolve(process.cwd(), "..", "_ik_ref", "InventoryKamera", "tessdata"),
path.resolve(process.env.USERPROFILE ?? "", "Desktop", "_ik_ref_fork", "InventoryKamera", "tessdata"),
path.resolve(process.env.USERPROFILE ?? "", "Desktop", "_ik_ref", "InventoryKamera", "tessdata"),
process.resourcesPath ? path.resolve(process.resourcesPath, "tessdata") : "",
].filter(Boolean) as string[];
}
function findIkTessdataDir() {
return ikTessdataCandidates().find((candidate) => existsSync(path.join(candidate, `${IK_TRAINEDDATA_LANG}.traineddata`))) ?? "";
void options;
return "current";
}
function getOcrWorkerOptions(engine: OcrWorkerEngine) {
if (engine !== "ik-traineddata") return { lang: "eng", options: undefined };
const langPath = findIkTessdataDir();
if (!langPath) {
throw new Error(`IK traineddata not found. Set IK_TESSDATA_DIR or place ${IK_TRAINEDDATA_LANG}.traineddata in data/tessdata.`);
}
return {
lang: IK_TRAINEDDATA_LANG,
options: {
langPath,
gzip: false,
cachePath: app.isReady() ? path.join(app.getPath("userData"), "tessdata-cache") : path.resolve(process.cwd(), "outputs", "tessdata-cache"),
},
};
void engine;
return { lang: "eng", options: undefined };
}
function getOcrWorkerPool(engine: OcrWorkerEngine) {
@@ -594,7 +888,7 @@ function getOcrWorkerPool(engine: OcrWorkerEngine) {
}
async function resetOcrWorker(engine?: OcrWorkerEngine) {
const engines: OcrWorkerEngine[] = engine ? [engine] : ["current", "ik-traineddata"];
const engines: OcrWorkerEngine[] = engine ? [engine] : ["current"];
await Promise.all(engines.map(async (engineId) => {
const state = ocrWorkerPools[engineId];
const broken = state.poolPromise;
@@ -710,7 +1004,6 @@ function warmOcrWorkerPool(engine: OcrWorkerEngine = "current") {
function getOcrWarmupStatus() {
return {
current: ocrWarmupStatuses.current,
"ik-traineddata": ocrWarmupStatuses["ik-traineddata"],
};
}
@@ -1334,7 +1627,7 @@ async function buildCaptureResult(
: await runOcrOnCropsWithTimeout(croppedPayload, ocrEngine);
const ocrMs = Date.now() - ocrStartedAt;
const totalMs = Date.now() - buildStartedAt;
const captureOcrEngine: CaptureOptions["ocrEngine"] = ocrEngine === "ik-traineddata" ? "ik-traineddata" : "current";
const captureOcrEngine: CaptureOptions["ocrEngine"] = "current";
const count = parseInventoryCount(recognized.ocr);
return {
@@ -1458,6 +1751,23 @@ function initializeAppLifecycle() {
moveMainWindowOffGenshin: async () => moveMainWindowOffGenshin(),
focusGenshinForScanStart: () => focusGenshinForScanStart(),
publishScannerStatus: (status: ScannerStatusPayload) => publishScannerStatus(status),
nativeScannerDataStatus: () => nativeScannerDataStatus(),
nativeScannerCatalog: () => nativeScannerCatalog(),
nativeScannerPreflight: (options?: { category?: string }) => nativeScannerPreflight(options),
nativeScannerStart: (options?: { limit?: number; category?: string }) => nativeScannerStart(options),
nativeScannerStop: () => nativeScannerStop(),
nativeScannerStatus: () => nativeScannerStatus(),
nativeScannerProcessRun: (options?: { runDir?: string; persist?: boolean; limit?: number }) => nativeScannerProcessRun(options),
nativeScannerLoadResults: (options?: { runDir?: string; limit?: number }) => nativeScannerLoadResults(options),
nativeScannerPromoteResults: (options: { runDir?: string; resultIds: string[] }) => nativeScannerPromoteResults(options),
nativeScannerReviewResult: (options: {
runDir?: string;
resultId: string;
action: "approve" | "reject";
artifact?: NativeScannerReviewArtifactInput;
note?: string;
}) => nativeScannerReviewResult(options),
nativeScannerLoadImage: (options: { runDir?: string; imagePath: string }) => nativeScannerLoadImage(options),
readRuntimeInfo: () => readRuntimeInfo(),
loadSnapshotFromDisk: () => loadSnapshotFromDisk(),
saveSnapshotToDisk: (snapshot: AppSnapshot) => saveSnapshotToDisk(snapshot),