16 lines
688 B
JavaScript
16 lines
688 B
JavaScript
import { acquireQvac, releaseQvac, loadAuxiliaryModel, qvacSdk, withQvacMaster, qvacBusy } from './qvac-master.js';
|
|
|
|
export class QvacPerception {
|
|
constructor({ model = process.env.JARVIS_OCR_MODEL || 'OCR_LATIN' } = {}) { this.model = model; this.modelId = null; }
|
|
async ocr(image) {
|
|
if (qvacBusy()) throw new Error('OCR skipped while the language model is answering');
|
|
await acquireQvac();
|
|
try {
|
|
this.modelId ||= await loadAuxiliaryModel(this.model, {});
|
|
const sdk = await qvacSdk();
|
|
const result = await withQvacMaster(() => sdk.ocr({ modelId: this.modelId, image, stream: false }));
|
|
return result.blocks;
|
|
} finally { releaseQvac(); }
|
|
}
|
|
}
|