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
@@ -0,0 +1,496 @@
import fs from "node:fs/promises";
import path from "node:path";
import { pngBufferToBitmap } from "./pngBitmap.js";
import { createNativeScannerResultWorkflowService } from "./nativeScannerResultWorkflowService.js";
import { parseArtifactCandidate } from "../../src/lib/artifactOcrParser.js";
import { toStoredArtifact } from "../../src/lib/artifactStore.js";
import { matchParsedArtifactToIk, type IkArtifactCatalog } from "../../src/lib/ikArtifactMatcher.js";
import { createStoredScanResultEntry } from "../../src/lib/scanResultEntry.js";
import type {
ArtifactStoreLoadResult,
ArtifactStoreSaveResult,
CaptureResult,
NativeScannerImageLoadStatus,
NativeScannerProcessStatus,
NativeScannerPromotionStatus,
NativeScannerReviewArtifactInput,
NativeScannerReviewStatus,
NativeScannerResultsLoadStatus,
ReviewSamplePayload,
} from "../../src/types/global.js";
import type { ScanResultCategory, StoredArtifactRecord, StoredScanResultEntry } from "../../src/types/storage.js";
export type NativeCaptureJobPayload = {
sequence: number;
category?: string;
page?: number;
row?: number;
col?: number;
capturedAt?: string;
relativePath?: string;
absolutePath?: string;
};
export interface NativeScannerProcessingService {
processRun(options?: { runDir?: string; persist?: boolean; limit?: number }): Promise<NativeScannerProcessStatus>;
loadResults(options?: { runDir?: string; limit?: number }): Promise<NativeScannerResultsLoadStatus>;
promoteResults(options: { runDir?: string; resultIds: string[] }): Promise<NativeScannerPromotionStatus>;
reviewResult(options: {
runDir?: string;
resultId: string;
action: "approve" | "reject";
artifact?: NativeScannerReviewArtifactInput;
note?: string;
}): Promise<NativeScannerReviewStatus>;
loadImage(options: { runDir?: string; imagePath: string }): Promise<NativeScannerImageLoadStatus>;
}
interface NativeScannerProcessingServiceDependencies {
resolveRunDir(runDir?: string): string;
buildCaptureResult(imagePath: string, job: NativeCaptureJobPayload): Promise<CaptureResult>;
loadArtifacts?: () => Promise<ArtifactStoreLoadResult>;
saveArtifacts(records: StoredArtifactRecord[]): Promise<Pick<ArtifactStoreSaveResult, "added" | "updated"> & Partial<ArtifactStoreSaveResult>>;
saveReviewSample?: (sample: ReviewSamplePayload) => Promise<{ ok: boolean }>;
loadIkArtifactCatalog?: () => Promise<IkArtifactCatalog | null>;
}
type ProcessedNativeJob = {
result: NativeScannerProcessStatus["results"][number];
scanResult: StoredScanResultEntry;
storedRecord: StoredArtifactRecord | null;
};
const POST_CAPTURE_QUEUE_CONCURRENCY = 4;
export function createNativeScannerProcessingService(
deps: NativeScannerProcessingServiceDependencies,
): NativeScannerProcessingService {
const resultWorkflows = createNativeScannerResultWorkflowService(deps);
return {
async processRun(options = {}) {
const started = Date.now();
const runDir = deps.resolveRunDir(options.runDir);
if (!runDir) {
return emptyProcessStatus("No native scanner runDir available.");
}
const { jobsPath, jobs } = await readNativeCaptureJobs(runDir);
const limit = Math.max(1, Math.min(jobs.length, Math.round(options.limit ?? jobs.length)));
const selectedJobs = jobs.slice(0, limit);
const runId = path.basename(runDir);
const ikCatalog = deps.loadIkArtifactCatalog
? await deps.loadIkArtifactCatalog().catch(() => null)
: null;
const processedJobs = await mapWithConcurrency(
selectedJobs,
POST_CAPTURE_QUEUE_CONCURRENCY,
(job) => processNativeCaptureJob({
deps,
ikCatalog,
job,
persist: Boolean(options.persist),
runDir,
runId,
}),
);
const results = processedJobs.map((entry) => entry.result);
const scanResults = processedJobs.map((entry) => entry.scanResult);
const recordsToPersist = processedJobs
.map((entry) => entry.storedRecord)
.filter((record): record is StoredArtifactRecord => Boolean(record));
let stored = 0;
if (options.persist && recordsToPersist.length > 0) {
const saved = await deps.saveArtifacts(recordsToPersist);
stored = saved.added + saved.updated;
}
const status: NativeScannerProcessStatus = {
ok: true,
runDir,
jobsPath,
reportPath: path.join(runDir, "processing-report.json"),
scanResultsPath: path.join(runDir, "scan-results.json"),
processed: results.length,
parsed: results.filter((result) => result.parsed).length,
review: results.filter((result) => result.needsReview).length,
stored,
errors: results.filter((result) => result.error).length,
elapsedMs: Date.now() - started,
queueConcurrency: Math.min(POST_CAPTURE_QUEUE_CONCURRENCY, selectedJobs.length),
persisted: Boolean(options.persist),
results,
};
await fs.writeFile(status.scanResultsPath, JSON.stringify(scanResults, null, 2), "utf8");
await fs.writeFile(status.reportPath, JSON.stringify(status, null, 2), "utf8");
return status;
},
async loadResults(options = {}) {
const runDir = deps.resolveRunDir(options.runDir);
if (!runDir) {
return emptyResultsStatus("No native scanner runDir available.");
}
const resultsPath = path.join(runDir, "scan-results.json");
try {
const raw = await fs.readFile(resultsPath, "utf8");
const parsed = JSON.parse(raw);
const results = Array.isArray(parsed)
? parsed.filter(isStoredScanResultEntry)
: [];
const limit = Number.isFinite(options.limit)
? Math.max(1, Math.min(results.length, Math.round(options.limit ?? results.length)))
: results.length;
return {
ok: true,
runDir,
path: resultsPath,
total: results.length,
results: results.slice(-limit),
};
} catch (error) {
return {
...emptyResultsStatus(error instanceof Error ? error.message : String(error)),
runDir,
path: resultsPath,
};
}
},
promoteResults: resultWorkflows.promoteResults,
reviewResult: resultWorkflows.reviewResult,
async loadImage(options) {
const runDir = deps.resolveRunDir(options.runDir);
if (!runDir) {
return emptyImageStatus("No native scanner runDir available.");
}
const resolvedRunDir = path.resolve(runDir);
const imagePath = path.isAbsolute(options.imagePath)
? path.resolve(options.imagePath)
: path.resolve(resolvedRunDir, options.imagePath);
if (!isPathInside(resolvedRunDir, imagePath)) {
return { ...emptyImageStatus("Image path is outside the native scanner run directory."), runDir: resolvedRunDir, path: imagePath };
}
if (!/\.png$/i.test(imagePath)) {
return { ...emptyImageStatus("Native scanner previews must be PNG files."), runDir: resolvedRunDir, path: imagePath };
}
try {
const stat = await fs.stat(imagePath);
if (stat.size > 20 * 1024 * 1024) {
return { ...emptyImageStatus("Native scanner preview is too large."), runDir: resolvedRunDir, path: imagePath };
}
const buffer = await fs.readFile(imagePath);
const bitmap = pngBufferToBitmap(buffer);
return {
ok: true,
runDir: resolvedRunDir,
path: imagePath,
dataUrl: `data:image/png;base64,${buffer.toString("base64")}`,
width: bitmap.width,
height: bitmap.height,
};
} catch (error) {
return {
...emptyImageStatus(error instanceof Error ? error.message : String(error)),
runDir: resolvedRunDir,
path: imagePath,
};
}
},
};
}
export function nativeScannerProcessStats(status: NativeScannerProcessStatus): Record<string, number> {
return {
processed: status.processed,
parsed: status.parsed,
review: status.review,
stored: status.stored,
errors: status.errors,
elapsedMs: status.elapsedMs,
queueConcurrency: status.queueConcurrency,
};
}
async function processNativeCaptureJob({
deps,
ikCatalog,
job,
persist,
runDir,
runId,
}: {
deps: NativeScannerProcessingServiceDependencies;
ikCatalog: IkArtifactCatalog | null;
job: NativeCaptureJobPayload;
persist: boolean;
runDir: string;
runId: string;
}): Promise<ProcessedNativeJob> {
const category = nativeJobCategory(job.category);
const imagePath = job.absolutePath || (job.relativePath ? path.join(runDir, job.relativePath) : "");
if (!category.artifactProcessingSupported) {
return reviewJobResult({
category: category.scanResultCategory,
error: `Native post-capture processing for category '${category.nativeCategory}' is not implemented yet; IK catalog is available only.`,
imagePath,
job,
runId,
});
}
if (!imagePath || !(await fileExists(imagePath))) {
const error = "Card crop image missing.";
return reviewJobResult({ category: category.scanResultCategory, error, imagePath, job, runId });
}
try {
const capture = await deps.buildCaptureResult(imagePath, job);
const parsed = parseArtifactCandidate(capture);
const ikMatch = parsed && ikCatalog ? matchParsedArtifactToIk(parsed, ikCatalog) : null;
const notes = [...new Set([...(parsed?.notes ?? []), ...(ikMatch?.notes ?? [])])];
const needsReview = parsedArtifactNeedsReview(parsed) || Boolean(ikMatch && !ikMatch.matched);
const canPersist = parsedArtifactCanPersist(parsed, needsReview);
const shouldPersist = Boolean(persist && parsed && canPersist && !needsReview);
const storedRecord = parsed && shouldPersist
? toStoredArtifact(parsed, "native-ik-scan", needsReview, capture.locked)
: null;
return {
result: {
sequence: job.sequence,
category: category.scanResultCategory,
page: job.page,
row: job.row,
col: job.col,
imagePath,
parsed: Boolean(parsed),
artifactName: parsed?.name,
setName: parsed?.setName,
slot: parsed?.slot,
confidence: parsed?.confidence ?? 0,
needsReview,
persisted: shouldPersist,
ikMatch: ikMatch ?? undefined,
notes,
ocr: capture.ocr ?? [],
},
scanResult: createStoredScanResultEntry({
runId,
sequence: job.sequence,
page: job.page,
row: job.row,
col: job.col,
category: category.scanResultCategory,
source: "native-ik-scan",
imagePath,
parsed,
needsReview,
confidence: parsed?.confidence ?? 0,
capturedAt: job.capturedAt,
artifactRecordId: storedRecord?.id,
persistedArtifact: shouldPersist,
notes,
locked: capture.locked,
ikMatch: ikMatch ?? undefined,
}),
storedRecord,
};
} catch (error) {
const message = error instanceof Error ? error.message : String(error);
return reviewJobResult({ category: category.scanResultCategory, error: message, imagePath, job, runId });
}
}
function reviewJobResult({
category,
error,
imagePath,
job,
runId,
}: {
category?: ScanResultCategory;
error: string;
imagePath: string;
job: NativeCaptureJobPayload;
runId: string;
}): ProcessedNativeJob {
const scanResultCategory = category ?? nativeJobCategory(job.category).scanResultCategory;
return {
result: {
sequence: job.sequence,
category: scanResultCategory,
page: job.page,
row: job.row,
col: job.col,
imagePath,
parsed: false,
needsReview: true,
confidence: 0,
ocr: [],
error,
},
scanResult: createStoredScanResultEntry({
runId,
sequence: job.sequence,
page: job.page,
row: job.row,
col: job.col,
category: scanResultCategory,
source: "native-ik-scan",
imagePath,
parsed: null,
needsReview: true,
confidence: 0,
capturedAt: job.capturedAt,
error,
notes: [error],
}),
storedRecord: null,
};
}
function nativeJobCategory(category: string | undefined): {
nativeCategory: string;
scanResultCategory: ScanResultCategory;
artifactProcessingSupported: boolean;
} {
const nativeCategory = (category ?? "artifacts").trim().toLowerCase() || "artifacts";
if (nativeCategory === "artifact" || nativeCategory === "artifacts") {
return { nativeCategory, scanResultCategory: "artifact", artifactProcessingSupported: true };
}
if (nativeCategory === "weapon" || nativeCategory === "weapons") {
return { nativeCategory, scanResultCategory: "weapon", artifactProcessingSupported: false };
}
if (nativeCategory === "character" || nativeCategory === "characters") {
return { nativeCategory, scanResultCategory: "character", artifactProcessingSupported: false };
}
if (nativeCategory === "material" || nativeCategory === "materials") {
return { nativeCategory, scanResultCategory: "material", artifactProcessingSupported: false };
}
return { nativeCategory, scanResultCategory: "unknown", artifactProcessingSupported: false };
}
async function mapWithConcurrency<TInput, TOutput>(
items: readonly TInput[],
concurrency: number,
worker: (item: TInput, index: number) => Promise<TOutput>,
) {
const output = Array<TOutput>(items.length);
let nextIndex = 0;
const workerCount = Math.max(1, Math.min(items.length, Math.floor(concurrency)));
await Promise.all(Array.from({ length: workerCount }, async () => {
while (nextIndex < items.length) {
const index = nextIndex++;
output[index] = await worker(items[index], index);
}
}));
return output;
}
async function readNativeCaptureJobs(runDir: string): Promise<{ jobsPath: string; jobs: NativeCaptureJobPayload[] }> {
const jobsPath = path.join(runDir, "capture-jobs.jsonl");
const raw = await fs.readFile(jobsPath, "utf8");
const jobs = raw
.split(/\r?\n/)
.map((line) => line.trim())
.filter(Boolean)
.map((line) => JSON.parse(line) as NativeCaptureJobPayload)
.filter((job) => Number.isFinite(job.sequence));
return { jobsPath, jobs };
}
async function fileExists(filePath: string) {
try {
await fs.access(filePath);
return true;
} catch {
return false;
}
}
function emptyProcessStatus(error: string): NativeScannerProcessStatus {
return {
ok: false,
runDir: "",
jobsPath: "",
reportPath: "",
scanResultsPath: "",
processed: 0,
parsed: 0,
review: 0,
stored: 0,
errors: 1,
elapsedMs: 0,
queueConcurrency: 0,
persisted: false,
results: [],
error,
};
}
function emptyResultsStatus(error: string): NativeScannerResultsLoadStatus {
return {
ok: false,
runDir: "",
path: "",
total: 0,
results: [],
error,
};
}
function emptyImageStatus(error: string): NativeScannerImageLoadStatus {
return {
ok: false,
runDir: "",
path: "",
dataUrl: "",
width: 0,
height: 0,
error,
};
}
function isPathInside(root: string, candidate: string) {
const relative = path.relative(root, candidate);
return relative === "" || (!relative.startsWith("..") && !path.isAbsolute(relative));
}
function isStoredScanResultEntry(value: unknown): value is StoredScanResultEntry {
if (!value || typeof value !== "object") return false;
const entry = value as Partial<StoredScanResultEntry>;
return typeof entry.id === "string"
&& typeof entry.runId === "string"
&& Number.isFinite(entry.sequence)
&& typeof entry.source === "string"
&& typeof entry.imagePath === "string"
&& typeof entry.extractionStatus === "string"
&& typeof entry.valueStatus === "string"
&& Array.isArray(entry.notes);
}
function parsedArtifactNeedsReview(parsed: ReturnType<typeof parseArtifactCandidate>) {
if (!parsed) return true;
if (parsed.confidence < 78) return true;
const criticalFields: Array<"name" | "slot" | "mainStat" | "mainValue" | "setName"> = ["name", "slot", "mainStat", "mainValue", "setName"];
if (criticalFields.some((field) => (parsed.fields[field]?.confidence ?? 0) < 70)) return true;
if (parsed.substats.length === 0) return true;
return parsed.notes.some((note) => /not confidently parsed|substats look incomplete|likely OCR misread/i.test(note));
}
function parsedArtifactCanPersist(parsed: ReturnType<typeof parseArtifactCandidate>, needsReview: boolean) {
if (!parsed) return false;
if (parsed.name === "Unknown artifact") return false;
if (parsed.slot === "Unknown slot") return false;
if (parsed.setName === "Unknown set") return false;
if (parsed.mainStat === "Unknown main stat") return false;
if (parsed.mainValue === "?") return false;
if (parsed.substats.length === 0) return false;
if (!needsReview && parsed.confidence < 68) return false;
if (needsReview && parsed.confidence < 60) return false;
return true;
}