Queue OCR & ASR work
All checks were successful
Publish Library / Build NPM Project (push) Successful in 35s
Publish Library / Tag Version (push) Successful in 6s

This commit is contained in:
2026-02-20 19:05:19 -05:00
parent 473424ae23
commit 790608f020
4 changed files with 78 additions and 32 deletions

View File

@@ -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'));
const handleMessage = ({ text, warning, error }: any) => {
setTimeout(() => worker.terminate(), 1000);
if(aborted) return;
if(error) reject(new Error(error));
else {
if(warning) console.warn(warning);
resolve(text);
}
};
const handleError = (err: Error) => {
setTimeout(() => worker.terminate(), 1000);
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}`));
this.queue.push({file, model, speaker, modelDir: <string>this.ai.options.path, token: <string>this.ai.options.hfToken,
resolve: (text: string | null) => !aborted && resolve(text),
reject: (err: Error) => !aborted && reject(err)
});
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 => {