postcast-engine/scripts/test-zero-character.js

// Тест: генерируем 3 ракурса Зеро. Каждая картинка — до 5 попыток (GPT-5.2 капризничает с tool_choice).

const fs = require('fs');
const path = require('path');
const axios = require('axios');
const FormData = require('form-data');

const ROOT = '/var/www/zeropost-engine';
process.chdir(ROOT);
require('dotenv').config({ path: path.join(ROOT, '.env') });
const config = require(path.join(ROOT, 'src/config'));
const settings = require(path.join(ROOT, 'src/services/settings'));
const { query } = require(path.join(ROOT, 'src/config/db'));

const UPLOADS_DIR = process.env.UPLOADS_DIR || '/var/www/zeropost-uploads';
let sharp = null;
try { sharp = require('sharp'); } catch {}

// Базовое описание персонажа — повторяется в каждом промпте чтобы сохранять консистентность.
const CHARACTER_BASE = `
Character: a small, friendly mascot named Zero. 
Body: a soft rounded square shape (like a plump pixel or a chubby tetris block), about the size of a small companion creature. 
Color: bright emerald green (#10b981) with a slight gradient to teal, soft and matte. 
Face: two simple round black dot eyes, a small understated mouth that expresses the current emotion. 
Personality: friendly, curious, enthusiastic, a bit nerdy. 
Style: clean modern vector illustration, flat design with soft shading, no outlines. 
Background: warm off-white (#fafaf9) with subtle geometric shapes in light teal/emerald, like Notion/Linear/Anthropic editorial illustrations. 
Composition: 1:1 square format with comfortable padding, character is the main focus.
Strictly: no text, no letters, no logos, no humans, no realistic robots, no glowing nodes, no circuits.
`.trim();

const POSES = [
  {
    name: 'avatar',
    desc: `Front-facing portrait pose. Zero is looking directly at the viewer with curious wide-open eyes and a small confident smile. Sits centered in the frame. Used as channel avatar.`,
  },
  {
    name: 'laptop',
    desc: `Zero sits at a tiny laptop (the laptop is also stylized geometric, in cream/emerald colors). Eyes focused on the screen, small thoughtful expression. Cozy setup with a small plant nearby. The mascot is enthusiastic about coding.`,
  },
  {
    name: 'eureka',
    desc: `Zero has just discovered something exciting. One small stick arm raised in the air with a soft glowing dot above the head representing an idea. Eyes wide and sparkly, big enthusiastic smile. Confetti or stylized geometric sparkles around.`,
  },
];

async function generateOne(prompt, attempt = 1) {
  const MAX_ATTEMPTS = 5;
  const model = process.env.AI_MODEL_IMAGE_VIA_RESPONSES || 'gpt-5.2';
  const wrapped = `Use the image_generation tool to create the following illustration. Do not write any text response, only call the tool.\n\n${prompt}`;
  try {
    const res = await axios.post(
      `${config.ai.baseUrl}/responses`,
      {
        model, input: wrapped,
        tools: [{ type: 'image_generation' }],
        tool_choice: { type: 'image_generation' },
      },
      {
        headers: { Authorization: `Bearer ${config.ai.imageApiKey}` },
        timeout: 300_000,
      }
    );
    const output = res.data?.output || [];
    const imgCall = output.find(o => o.type === 'image_generation_call');
    if (!imgCall || !imgCall.result) {
      if (attempt < MAX_ATTEMPTS) {
        console.log(`[zero]   retry attempt=${attempt + 1} (no image_generation_call)`);
        return generateOne(prompt, attempt + 1);
      }
      throw new Error(`No image after ${MAX_ATTEMPTS} attempts`);
    }
    return Buffer.from(imgCall.result, 'base64');
  } catch (err) {
    if (attempt < MAX_ATTEMPTS) {
      const msg = err.response?.data?.error?.message || err.message;
      console.log(`[zero]   retry attempt=${attempt + 1} (${msg.slice(0, 60)})`);
      return generateOne(prompt, attempt + 1);
    }
    throw err;
  }
}

async function processBytes(bytes, name) {
  const ts = Date.now();
  const webpName = `zero-${name}-${ts}.webp`;
  const webpPath = path.join(UPLOADS_DIR, webpName);
  if (sharp) {
    await sharp(bytes)
      .resize(1024, 1024, { fit: 'cover' })
      .webp({ quality: 88 })
      .toFile(webpPath);
  } else {
    fs.writeFileSync(webpPath, bytes);
  }
  return webpPath;
}

(async () => {
  try {
    console.log('[zero] generating 3 poses sequentially (with retries)...');
    const results = [];
    for (const p of POSES) {
      console.log(`[zero] starting ${p.name}...`);
      const prompt = `${CHARACTER_BASE}\n\nPose: ${p.desc}`;
      const bytes = await generateOne(prompt);
      const localPath = await processBytes(bytes, p.name);
      console.log(`[zero] ${p.name} → ${localPath}`);
      results.push({ name: p.name, path: localPath });
    }

    // Шлём все 3 в TG как media group с подписью
    const { rows: chs } = await query(`SELECT * FROM channels WHERE id=1`);
    const channel = chs[0];
    const tgBase = await settings.get('TELEGRAM_API_BASE', 'https://api.telegram.org');

    const form = new FormData();
    form.append('chat_id', String(channel.tg_channel_id));

    const media = results.map((r, i) => ({
      type: 'photo',
      media: `attach://photo${i}`,
      caption: i === 0
        ? `🧪 Тест Зеро — три позы:\n1. аватар  2. за ноутом  3. эврика`
        : undefined,
    }));
    form.append('media', JSON.stringify(media));
    results.forEach((r, i) => {
      form.append(`photo${i}`, fs.createReadStream(r.path));
    });

    const res = await axios.post(`${tgBase}/bot${channel.bot_token}/sendMediaGroup`, form, {
      headers: form.getHeaders(),
      timeout: 60000,
      maxContentLength: Infinity,
      maxBodyLength: Infinity,
    });

    const msgIds = (res.data?.result || []).map(m => m.message_id);
    console.log(`[zero] sent media group, message_ids=${msgIds.join(',')}`);
    console.log('[zero] DONE');
    process.exit(0);
  } catch (err) {
    console.error('[zero] FAILED:', err.response?.data || err.message);
    process.exit(1);
  }
})();