Doc Agent

import Tesseract from 'tesseract.js'; import type { LogLevel, OcrProgressCallback, StreamCallback } from './types'; /** * Helper to emit log via stream callback */ function emitLog( onStream: StreamCallback | undefined, level: LogLevel, message: string, data?: Record<string, unknown> ): void { onStream?.({ type: 'log', level, message, data }); } /** * OCR all images in parallel using tesseract.js * @returns Concatenated text with page markers */ export async function ocrImages( images: Buffer[], onProgress?: OcrProgressCallback, onStream?: StreamCallback ): Promise<string> { if (images.length === 0) return ''; try { const totalPages = images.length; // Process all pages in parallel const results = await Promise.all( images.map(async (image, index) => { const pageNum = index + 1; try { const result = await Tesseract.recognize(image, 'eng', { logger: (m) => { if (onProgress && m.status === 'recognizing text') { onProgress(pageNum, totalPages, m.progress, m.status); } }, }); emitLog(onStream, 'debug', `OCR completed for page ${pageNum}`, { page: pageNum, textLength: result.data.text.length, }); return { page: pageNum, text: result.data.text }; } catch (error) { emitLog(onStream, 'error', `OCR failed for page ${pageNum}`, { page: pageNum, error: String(error), }); return { page: pageNum, text: '' }; } }) ); // Concatenate with page markers const ocrText = results .filter((r) => r.text.trim()) .map((r) => `--- Page ${r.page} ---\n${r.text.trim()}`) .join('\n\n'); emitLog(onStream, 'info', `OCR complete: ${totalPages} pages, ${ocrText.length} chars`, { totalPages, totalTextLength: ocrText.length, }); return ocrText; } catch (error) { emitLog(onStream, 'error', 'OCR batch failed', { error: String(error) }); return ''; } }