Doc Agent

import { readFileSync } from 'node:fs'; import type { Config, DocumentData } from '@doc-agent/core'; import { extractWithGemini } from './providers/gemini'; import { extractWithOllama } from './providers/ollama'; import type { ExtractOptions } from './types'; /** * Extract structured data from a document (PDF or image) * Routes to the appropriate AI provider based on config */ export async function extractDocument( filePath: string, config: Config, options?: ExtractOptions ): Promise<DocumentData> { options?.onStream?.({ type: 'log', level: 'info', message: `Starting extraction with ${config.aiProvider}`, data: { filePath, provider: config.aiProvider }, }); const fileBuffer = readFileSync(filePath); const base64 = fileBuffer.toString('base64'); if (config.aiProvider === 'gemini') { return extractWithGemini(filePath, base64, config); } if (config.aiProvider === 'ollama') { return extractWithOllama(filePath, base64, config, 0, options?.onStream); } throw new Error(`Provider ${config.aiProvider} not yet implemented`); }