Queue OCR & ASR work

package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@ztimson/ai-utils",
-	"version": "0.7.5",
+	"version": "0.7.7",
 	"description": "AI Utility library",
 	"author": "Zak Timson",
 	"license": "MIT",

src/asr.ts

@@ -110,30 +110,28 @@ function prepareAudioBuffer(file: string): [string, Float32Array] {
 }
 
 parentPort?.on('message', async ({ file, speaker, model, modelDir, token }) => {
+	let tempFile = null;
 	try {
 		if(!whisperPipeline) whisperPipeline = await pipeline('automatic-speech-recognition', `Xenova/${model}`, {cache_dir: modelDir, quantized: true});
 
-		// Prepare audio file
 		const [f, buffer] = prepareAudioBuffer(file);
-
+		tempFile = f !== file ? f : null;
-		// Fetch transcript and speakers
 		const hasDiarization = await canDiarization();
 		const [transcript, speakers] = await Promise.all([
 			whisperPipeline(buffer, {return_timestamps: speaker ? 'word' : false}),
 			(!speaker || !token || !hasDiarization) ? Promise.resolve(): runDiarization(hasDiarization, f, modelDir, token),
 		]);
-		if(file != f) rmSync(f, { recursive: true, force: true });
 
-		// Return any results / errors if no more processing required
 		const text = transcript.text?.trim() || null;
 		if(!speaker) return parentPort?.postMessage({ text });
 		if(!token) return parentPort?.postMessage({ text, error: 'HuggingFace token required' });
 		if(!hasDiarization) return parentPort?.postMessage({ text, error: 'Speaker diarization unavailable' });
 
-		// Combine transcript and speakers
 		const combined = combineSpeakerTranscript(transcript.chunks || [], speakers || []);
 		parentPort?.postMessage({ text: combined });
 	} catch (err: any) {
 		parentPort?.postMessage({ error: err.stack || err.message });
+	} finally {
+		if(tempFile) rmSync(tempFile, { recursive: true, force: true });
 	}
 });

src/audio.ts

@@ -5,37 +5,59 @@ import {canDiarization} from './asr.ts';
 import {dirname, join} from 'path';
 
 export class Audio {
+	private busy = false;
+	private currentJob: any;
+	private queue: Array<{file: string, model: string, speaker: boolean | 'id', modelDir: string, token: string, resolve: any, reject: any}> = [];
+	private worker: Worker | null = null;
+
 	constructor(private ai: Ai) {}
 
+	private processQueue() {
+		if(this.busy || !this.queue.length) return;
+
+		this.busy = true;
+		const job = this.queue.shift()!;
+		if(!this.worker) {
+			this.worker = new Worker(join(dirname(fileURLToPath(import.meta.url)), 'asr.js'));
+			this.worker.on('message', this.handleMessage.bind(this));
+			this.worker.on('error', this.handleError.bind(this));
+		}
+
+		this.currentJob = job;
+		this.worker.postMessage({file: job.file, model: job.model, speaker: job.speaker, modelDir: job.modelDir, token: job.token});
+	}
+
+	private handleMessage({text, warning, error}: any) {
+		const job = this.currentJob!;
+		this.busy = false;
+		if(error) job.reject(new Error(error));
+		else {
+			if(warning) console.warn(warning);
+			job.resolve(text);
+		}
+		this.processQueue();
+	}
+
+	private handleError(err: Error) {
+		if(this.currentJob) {
+			this.currentJob.reject(err);
+			this.busy = false;
+			this.processQueue();
+		}
+	}
+
 	asr(file: string, options: { model?: string; speaker?: boolean | 'id' } = {}): AbortablePromise<string | null> {
 		const { model = this.ai.options.asr || 'whisper-base', speaker = false } = options;
 		let aborted = false;
 		const abort = () => { aborted = true; };
-
 		let p = new Promise<string | null>((resolve, reject) => {
-			const worker = new Worker(join(dirname(fileURLToPath(import.meta.url)), 'asr.js'));
+			this.queue.push({file, model, speaker, modelDir: <string>this.ai.options.path, token: <string>this.ai.options.hfToken,
-			const handleMessage = ({ text, warning, error }: any) => {
+				resolve: (text: string | null) => !aborted && resolve(text),
-				worker.terminate();
+				reject: (err: Error) => !aborted && reject(err)
-				if(aborted) return;
-				if(error) reject(new Error(error));
-				else {
-					if(warning) console.warn(warning);
-					resolve(text);
-				}
-			};
-			const handleError = (err: Error) => {
-				worker.terminate();
-				if(!aborted) reject(err);
-			};
-			worker.on('message', handleMessage);
-			worker.on('error', handleError);
-			worker.on('exit', (code) => {
-				if(code !== 0 && !aborted) reject(new Error(`Worker exited with code ${code}`));
 			});
-			worker.postMessage({file, model, speaker, modelDir: this.ai.options.path, token: this.ai.options.hfToken});
+			this.processQueue();
 		});
 
-		// Name speakers using AI
 		if(options.speaker == 'id') {
 			if(!this.ai.language.defaultModel) throw new Error('Configure an LLM for advanced ASR speaker detection');
 			p = p.then(async transcript => {

src/vision.ts

@@ -2,8 +2,26 @@ import {createWorker} from 'tesseract.js';
 import {AbortablePromise, Ai} from './ai.ts';
 
 export class Vision {
+	private worker: any = null;
+	private queue: Array<{ path: string, resolve: any, reject: any }> = [];
+	private busy = false;
 
-	constructor(private ai: Ai) { }
+	constructor(private ai: Ai) {}
+
+	private async processQueue() {
+		if(this.busy || !this.queue.length) return;
+		this.busy = true;
+		const job = this.queue.shift()!;
+		if(!this.worker) this.worker = await createWorker(this.ai.options.ocr || 'eng', 2, {cachePath: this.ai.options.path});
+		try {
+			const {data} = await this.worker.recognize(job.path);
+			job.resolve(data.text.trim() || null);
+		} catch(err) {
+			job.reject(err);
+		}
+		this.busy = false;
+		this.processQueue();
+	}
 
 	/**
 	 * Convert image to text using Optical Character Recognition
@@ -11,13 +29,16 @@ export class Vision {
 	 * @returns {AbortablePromise<string | null>} Promise of extracted text with abort method
 	 */
 	ocr(path: string): AbortablePromise<string | null> {
-		let worker: any;
+		let aborted = false;
-		const p = new Promise<string | null>(async res => {
+		const abort = () => { aborted = true; };
-			worker = await createWorker(this.ai.options.ocr || 'eng', 2, {cachePath: this.ai.options.path});
+		const p = new Promise<string | null>((resolve, reject) => {
-			const {data} = await worker.recognize(path);
+			this.queue.push({
-			await worker.terminate();
+				path,
-			res(data.text.trim() || null);
+				resolve: (text: string | null) => !aborted && resolve(text),
+				reject: (err: Error) => !aborted && reject(err)
 			});
-		return Object.assign(p, {abort: () => worker?.terminate()});
+			this.processQueue();
+		});
+		return Object.assign(p, {abort});
 	}
 }