Track guide OCR snapshot progress

2026-05-28 19:35:42 +07:00
8 changed files with 187 additions and 60 deletions
@@ -232,6 +232,74 @@ describe("GuideStore", () => {
 		await expect(fs.readFile(html.path, "utf-8")).resolves.toContain("<!doctype html>");
 	});

+	it("resumes OCR without reprocessing completed snapshots", async () => {
+		const recognizedSnapshotIds: string[] = [];
+		const store = new GuideStore(recordingsDir, {
+			ocrClient: {
+				recognize: async (snapshot) => {
+					recognizedSnapshotIds.push(snapshot.id);
+					return [];
+				},
+			},
+		});
+		await store.startSession(115);
+		const firstMarker = await store.addMarker({
+			recordingId: 115,
+			kind: "hotkey",
+			timeMs: 100,
+			label: "Ctrl+F12 marker",
+			normalizedX: 0.25,
+			normalizedY: 0.35,
+		});
+		const secondMarker = await store.addMarker({
+			recordingId: 115,
+			kind: "hotkey",
+			timeMs: 300,
+			label: "Ctrl+F12 marker",
+			normalizedX: 0.6,
+			normalizedY: 0.7,
+		});
+		const firstEvent = firstMarker.event;
+		const secondEvent = secondMarker.event;
+		await store.writeSnapshot({
+			recordingId: 115,
+			eventId: firstEvent?.id ?? "",
+			timeMs: 100,
+			offsetMs: 0,
+			width: 800,
+			height: 600,
+			pngBytes: new Uint8Array([1, 2, 3]).buffer,
+		});
+		await store.writeSnapshot({
+			recordingId: 115,
+			eventId: secondEvent?.id ?? "",
+			timeMs: 300,
+			offsetMs: 0,
+			width: 800,
+			height: 600,
+			pngBytes: new Uint8Array([4, 5, 6]).buffer,
+		});
+
+		await store.runOcr({
+			recordingId: 115,
+			snapshotIds: [`snapshot-${firstEvent?.id}`],
+		});
+		expect(recognizedSnapshotIds).toEqual([`snapshot-${firstEvent?.id}`]);
+
+		const resumedSession = await store.runOcr({ recordingId: 115 });
+		expect(recognizedSnapshotIds).toEqual([
+			`snapshot-${firstEvent?.id}`,
+			`snapshot-${secondEvent?.id}`,
+		]);
+		expect(resumedSession.snapshots.every((snapshot) => snapshot.ocrCompletedAt)).toBe(true);
+
+		await store.runOcr({ recordingId: 115 });
+		expect(recognizedSnapshotIds).toEqual([
+			`snapshot-${firstEvent?.id}`,
+			`snapshot-${secondEvent?.id}`,
+		]);
+	});
+
 	it("repairs generic hotkey marker text and attaches AI draft artifacts", async () => {
 		const store = new GuideStore(recordingsDir, {
 			ocrClient: {
@@ -58,6 +58,8 @@ const VALID_EVENT_SOURCES = new Set<GuideEventSource>([
 	"review-ui",
 ]);

+const guideOcrJobsByRecordingId = new Map<string, Promise<GuideSession>>();
+
 export class GuideStoreError extends Error {
 	constructor(
 		readonly code: GuideErrorCode,
@@ -259,50 +261,103 @@ export class GuideStore {
 	}

 	async runOcr(input: RunGuideOcrInput): Promise<GuideSession> {
-		const session = await this.readSession(input.recordingId);
-		const requestedIds = new Set(input.snapshotIds ?? []);
-		const snapshots =
-			requestedIds.size > 0
-				? session.snapshots.filter((snapshot) => requestedIds.has(snapshot.id))
-				: session.snapshots;
-		if (snapshots.length === 0) {
-			throw new GuideStoreError("guide-invalid-input", "No guide snapshots are available for OCR.");
+		const recordingId = normalizeGuideRecordingId(input.recordingId);
+		if (!recordingId) {
+			throw new GuideStoreError("guide-invalid-input", "OCR run is missing recordingId.");
 		}

-		const ocrClient =
-			this.dependencies.ocrClient ??
-			DefaultGuideOcrClient.fromConfig(await this.dependencies.ocrConfigProvider?.getOcrConfig());
-		const shouldFocusOcrSnapshots =
-			this.dependencies.focusOcrSnapshots ?? this.dependencies.ocrClient === undefined;
-		const eventsById = new Map(session.events.map((event) => [event.id, event]));
-		const blocks: OcrBlock[] = [];
-		try {
-			for (const snapshot of snapshots) {
-				const focusedSnapshot = shouldFocusOcrSnapshots
-					? await createFocusedOcrSnapshot({
-							snapshot,
-							event: eventsById.get(snapshot.eventId),
-							outputDir: session.outputDir,
-						})
-					: { snapshot };
-				const recognizedBlocks = await ocrClient.recognize(focusedSnapshot.snapshot);
-				blocks.push(...remapFocusedOcrBlocks(recognizedBlocks, focusedSnapshot.transform));
+		const previousJob =
+			guideOcrJobsByRecordingId.get(recordingId)?.catch(() => undefined) ?? Promise.resolve();
+		const nextJob = previousJob.then(async () => {
+			let session = await this.readSession(recordingId);
+			const requestedIds = new Set(input.snapshotIds ?? []);
+			const snapshots =
+				requestedIds.size > 0
+					? session.snapshots.filter((snapshot) => requestedIds.has(snapshot.id))
+					: session.snapshots;
+			if (snapshots.length === 0) {
+				throw new GuideStoreError(
+					"guide-invalid-input",
+					"No guide snapshots are available for OCR.",
+				);
 			}
-		} catch (error) {
-			throw new GuideStoreError(
-				"guide-ocr-unavailable",
-				error instanceof Error ? error.message : "OCR failed.",
-				true,
-			);
-		}

-		const snapshotIds = new Set(snapshots.map((snapshot) => snapshot.id));
+			const completedSnapshotIds = new Set(
+				session.snapshots
+					.filter((snapshot) => isSnapshotOcrCompleted(snapshot, session.ocrBlocks))
+					.map((snapshot) => snapshot.id),
+			);
+			const pendingSnapshots = snapshots.filter(
+				(snapshot) => !completedSnapshotIds.has(snapshot.id),
+			);
+			if (pendingSnapshots.length === 0) {
+				if (session.status === "ocr-ready") {
+					return session;
+				}
+				const readySession = touchSession({
+					...session,
+					status: "ocr-ready",
+					candidates: buildGuideStepCandidates(session),
+				});
+				await this.writeSession(readySession);
+				return readySession;
+			}
+
+			const ocrClient =
+				this.dependencies.ocrClient ??
+				DefaultGuideOcrClient.fromConfig(await this.dependencies.ocrConfigProvider?.getOcrConfig());
+			const shouldFocusOcrSnapshots =
+				this.dependencies.focusOcrSnapshots ?? this.dependencies.ocrClient === undefined;
+			const eventsById = new Map(session.events.map((event) => [event.id, event]));
+			try {
+				for (const snapshot of pendingSnapshots) {
+					const focusedSnapshot = shouldFocusOcrSnapshots
+						? await createFocusedOcrSnapshot({
+								snapshot,
+								event: eventsById.get(snapshot.eventId),
+								outputDir: session.outputDir,
+							})
+						: { snapshot };
+					const recognizedBlocks = await ocrClient.recognize(focusedSnapshot.snapshot);
+					const blocks = remapFocusedOcrBlocks(recognizedBlocks, focusedSnapshot.transform);
+					session = await this.writeOcrSnapshotProgress(session, snapshot.id, blocks);
+				}
+			} catch (error) {
+				throw new GuideStoreError(
+					"guide-ocr-unavailable",
+					error instanceof Error ? error.message : "OCR failed.",
+					true,
+				);
+			}
+
+			return session;
+		});
+		guideOcrJobsByRecordingId.set(recordingId, nextJob);
+		try {
+			return await nextJob;
+		} finally {
+			if (guideOcrJobsByRecordingId.get(recordingId) === nextJob) {
+				guideOcrJobsByRecordingId.delete(recordingId);
+			}
+		}
+	}
+
+	private async writeOcrSnapshotProgress(
+		session: GuideSession,
+		snapshotId: string,
+		blocks: OcrBlock[],
+	): Promise<GuideSession> {
 		const updatedOcrBlocks = [
-			...session.ocrBlocks.filter((block) => !snapshotIds.has(block.snapshotId)),
+			...session.ocrBlocks.filter((block) => block.snapshotId !== snapshotId),
 			...blocks,
 		];
+		const completedAt = new Date().toISOString();
+		const updatedSnapshots = session.snapshots.map((snapshot) =>
+			snapshot.id === snapshotId ? { ...snapshot, ocrCompletedAt: completedAt } : snapshot,
+		);
 		const draftSession = {
 			...session,
+			snapshots: updatedSnapshots,
 			ocrBlocks: updatedOcrBlocks,
 		};
 		const updatedSession = touchSession({
@@ -679,6 +734,7 @@ function normalizeGuideSnapshot(input: unknown): GuideSnapshot | null {
 	const eventId = normalizeString(input.eventId);
 	const pathValue = normalizeString(input.path);
 	const markedPath = normalizeOptionalString(input.markedPath);
+	const ocrCompletedAt = normalizeOptionalString(input.ocrCompletedAt);
 	const timeMs = normalizeNonNegativeNumber(input.timeMs);
 	const offsetMs = normalizeOptionalNumber(input.offsetMs);
 	const width = normalizePositiveInteger(input.width);
@@ -694,7 +750,23 @@ function normalizeGuideSnapshot(input: unknown): GuideSnapshot | null {
 	) {
 		return null;
 	}
-	return { id, eventId, timeMs, offsetMs, path: pathValue, markedPath, width, height };
+	return {
+		id,
+		eventId,
+		timeMs,
+		offsetMs,
+		path: pathValue,
+		markedPath,
+		ocrCompletedAt,
+		width,
+		height,
+	};
+}
+
+function isSnapshotOcrCompleted(snapshot: GuideSnapshot, ocrBlocks: OcrBlock[]): boolean {
+	return (
+		Boolean(snapshot.ocrCompletedAt) || ocrBlocks.some((block) => block.snapshotId === snapshot.id)
+	);
 }

 function normalizeOcrBlock(input: unknown): OcrBlock | null {
@@ -974,21 +974,13 @@ try {
  $graphics.SmoothingMode = [System.Drawing.Drawing2D.SmoothingMode]::AntiAlias
  $graphics.DrawImage($source, 0, 0, $source.Width, $source.Height)
  $shortSide = [Math]::Max(1, [Math]::Min($source.Width, $source.Height))
-  $haloRadius = [Math]::Min(14, [Math]::Max(8, [Math]::Round($shortSide * 0.012)))
-  $dotRadius = [Math]::Min(6, [Math]::Max(3, [Math]::Round($shortSide * 0.0045)))
-  $lineWidth = [Math]::Max(1, [Math]::Round($shortSide * 0.0015))
+  $dotRadius = [Math]::Min(7, [Math]::Max(4, [Math]::Round($shortSide * 0.005)))
  $x = [Math]::Min($source.Width, [Math]::Max(0, ${marker.x.toFixed(4)}))
  $y = [Math]::Min($source.Height, [Math]::Max(0, ${marker.y.toFixed(4)}))
-  $haloBrush = [System.Drawing.SolidBrush]::new([System.Drawing.Color]::FromArgb(87, 250, 204, 21))
-  $ringPen = [System.Drawing.Pen]::new([System.Drawing.Color]::FromArgb(184, 239, 68, 68), $lineWidth)
  $dotBrush = [System.Drawing.SolidBrush]::new([System.Drawing.Color]::FromArgb(235, 220, 38, 38))
  try {
-    $graphics.FillEllipse($haloBrush, $x - $haloRadius, $y - $haloRadius, $haloRadius * 2, $haloRadius * 2)
-    $graphics.DrawEllipse($ringPen, $x - $haloRadius, $y - $haloRadius, $haloRadius * 2, $haloRadius * 2)
    $graphics.FillEllipse($dotBrush, $x - $dotRadius, $y - $dotRadius, $dotRadius * 2, $dotRadius * 2)
  } finally {
-    $haloBrush.Dispose()
-    $ringPen.Dispose()
    $dotBrush.Dispose()
  }
  $bitmap.Save($outputPath, [System.Drawing.Imaging.ImageFormat]::Png)
@@ -1,12 +1,12 @@
 {
 	"name": "openscreen",
-	"version": "1.4.8",
+	"version": "1.4.9",
 	"lockfileVersion": 3,
 	"requires": true,
 	"packages": {
 		"": {
 			"name": "openscreen",
-			"version": "1.4.8",
+			"version": "1.4.9",
 			"dependencies": {
 				"@fix-webm-duration/fix": "^1.0.1",
 				"@pixi/filter-drop-shadow": "^5.2.0",
@@ -1,7 +1,7 @@
 {
 	"name": "openscreen",
 	"private": true,
-	"version": "1.4.8",
+	"version": "1.4.9",
 	"type": "module",
 	"packageManager": "npm@10.9.4",
 	"engines": {
@@ -367,7 +367,11 @@ export function GuidePanel({ recordingId, videoPath, videoSourcePath }: GuidePan
 				});
 				setSession(current);
 			}
-			if (current.ocrBlocks.length === 0 && current.snapshots.length > 0) {
+			const ocrCompletedSnapshotIds = new Set(current.ocrBlocks.map((block) => block.snapshotId));
+			const hasPendingOcr = current.snapshots.some(
+				(snapshot) => !snapshot.ocrCompletedAt && !ocrCompletedSnapshotIds.has(snapshot.id),
+			);
+			if (hasPendingOcr) {
 				const ocrResult = await window.electronAPI.guide.runOcr({
 					recordingId: current.recordingId,
 				});
@@ -56,6 +56,7 @@ export interface GuideSnapshot {
 	offsetMs: number;
 	path: string;
 	markedPath?: string;
+	ocrCompletedAt?: string;
 	width: number;
 	height: number;
 }
@@ -177,17 +177,7 @@ function drawSnapshotMarker(
 	point: { x: number; y: number },
 ) {
 	const shortSide = Math.max(1, Math.min(canvas.width, canvas.height));
-	const haloRadius = clampNumber(Math.round(shortSide * 0.012), 8, 14);
-	const dotRadius = clampNumber(Math.round(shortSide * 0.0045), 3, 6);
-	const lineWidth = Math.max(1, Math.round(shortSide * 0.0015));
-
-	context.beginPath();
-	context.arc(point.x, point.y, haloRadius, 0, Math.PI * 2);
-	context.fillStyle = "rgba(250, 204, 21, 0.34)";
-	context.fill();
-	context.lineWidth = lineWidth;
-	context.strokeStyle = "rgba(239, 68, 68, 0.72)";
-	context.stroke();
+	const dotRadius = clampNumber(Math.round(shortSide * 0.005), 4, 7);

 	context.beginPath();
 	context.arc(point.x, point.y, dotRadius, 0, Math.PI * 2);