Can I vibe code TinyIMG?

Build a lightweight, automated image optimization and AI alt-tag microservice for a Shopify store using Node.js, Express, BullMQ, Sharp, and the OpenAI Vision API.

1. DATA MODEL & ARCHITECTURE:
- PostgreSQL database using Prisma ORM.
- Models: Store (id, shopifyDomain, accessToken), ImageBackup (id, originalUrl, optimizedUrl, shopifyImageId, productId, altText, status, createdAt).
- Background processing powered by Redis and BullMQ.

2. WORKFLOWS & LOGIC:
- Listen to Shopify webhooks: 'products/create' and 'products/update'.
- Extract image URLs from the product payload. Enqueue an optimization job for each new or updated image.
- Worker Process:
  a) Download the raw image buffer from the Shopify CDN URL.
  b) Save a copy of the original image buffer to Cloudflare R2 / AWS S3 bucket for rollback storage.
  c) Use 'sharp' library to compress the image: convert format to WebP, quality 80, strip EXIF metadata, and enforce maximum dimensions (2048x2048).
  d) Upload optimized image buffer to S3 / Cloudflare R2 and retrieve public URL.
  e) Call OpenAI API (gpt-4o-mini) passing the image URL and product title: 'Generate an SEO-friendly image alt tag under 120 characters describing this product image in context of: {productTitle}. Output plain text only.'
  f) Execute Shopify GraphQL Mutation 'productImageUpdate' to update the image's alt text and source URL on the Shopify product media record.
  g) Record job outcome in ImageBackup table.

3. ERROR HANDLING & CONCURRENCY:
- Wrap Shopify API calls in rate-limiting middleware respecting 'X-Shopify-Shop-Api-Call-Limit' headers.
- Implement exponential backoff retries for OpenAI API and Shopify GraphQL requests.
- If OpenAI alt text generation fails, fallback to using the Product Title as default alt text.

4. OUT OF SCOPE:
- Frontend embedded app interface (use environment variables for single-store setup).
- Complex page speed analytics engines or theme Liquid code injection.

$ each button prefixes agent-specific run instructions · build your own product, never copy proprietary code, trademarks or designs