pfm-ocr/backend/pfm-web-app/test_notfound_images.js

const fs = require('fs');
const path = require('path');

const PIPELINE_URL = 'http://localhost:8000/layout-parsing';
const PROXY_URL = 'http://localhost:8000/api/vllm-proxy/v1/chat/completions';
const UPLOADS_DIR = path.join(__dirname, '..', 'uploads');

const notFoundImages = [
  '1782888211457-rotated_1782888204591_rotated_1782888199962_rotated_1782888195402_CAP7202641176939787142.jpg',
  '1782875064223-CAP2007290974474760139.jpg',
  '1782884586047-CAP9169719214882332189.jpg',
  '1782893409879-CAP4330863738757813156.jpg'
];

async function testImage(filename) {
  console.log(`\n========================================`);
  console.log(`TESTING FILE: ${filename}`);
  console.log(`========================================`);

  const filePath = path.join(UPLOADS_DIR, filename);
  if (!fs.existsSync(filePath)) {
    console.error(`File does not exist on disk: ${filePath}`);
    return;
  }

  const fileBuffer = fs.readFileSync(filePath);
  const base64Image = fileBuffer.toString('base64');

  // Test 1: Hit the Layout Parsing Pipeline API
  console.log('\n--- Test 1: Layout Parsing Pipeline (Standard) ---');
  try {
    const res = await fetch(PIPELINE_URL, {
      method: 'POST',
      headers: { 'Content-Type': 'application/json' },
      body: JSON.stringify({
        file: base64Image,
        matchHistoryJob: false,
        useLayoutDetection: true,
        fileType: 1,
        useDocUnwarping: false,
        useDocOrientationClassify: true
      })
    });

    if (res.ok) {
      const data = await res.json();
      const markdown = data.result?.layoutParsingResults?.[0]?.markdown?.text || data.layoutParsingResults?.[0]?.markdown?.text || '';
      console.log('Resulting Markdown snippet (first 300 chars):');
      console.log(markdown.substring(0, 300));
      console.log(`\nDoes it contain PO, SO, DO or Tanggal?`);
      console.log(`- "Tanggal": ${/Tanggal/i.test(markdown)}`);
      console.log(`- "SO": ${/SO/i.test(markdown)}`);
      console.log(`- "DO": ${/DO/i.test(markdown)}`);
      console.log(`- "PO": ${/PO/i.test(markdown)}`);
    } else {
      console.error(`Pipeline returned status ${res.status}: ${await res.text()}`);
    }
  } catch (err) {
    console.error('Pipeline test failed:', err.message);
  }

  // Test 2: Direct vLLM completions with simple extraction prompt
  console.log('\n--- Test 2: Direct vLLM Simple Extraction ---');
  try {
    const res = await fetch(PROXY_URL, {
      method: 'POST',
      headers: { 'Content-Type': 'application/json' },
      body: JSON.stringify({
        model: "PaddleOCR-VL-1.6-0.9B",
        messages: [
          {
            role: "user",
            content: [
              {
                type: "image_url",
                image_url: { url: `data:image/jpeg;base64,${base64Image}` }
              },
              {
                type: "text",
                text: "Read the top right section of the document. Extract Tanggal, No. SO, No. DO, and No. PO."
              }
            ]
          }
        ],
        temperature: 0.1,
        max_tokens: 300
      })
    });

    if (res.ok) {
      const data = await res.json();
      const content = data.choices?.[0]?.message?.content;
      console.log('vLLM Response:');
      console.log(content);
    } else {
      console.error(`vLLM proxy returned status ${res.status}: ${await res.text()}`);
    }
  } catch (err) {
    console.error('vLLM test failed:', err.message);
  }
}

async function runAll() {
  for (const filename of notFoundImages) {
    await testImage(filename);
  }
}

runAll();