Can I vibe code CartUp AI?

Build a self-hosted e-commerce search and recommendation microservice using Node.js, Typesense, and PostgreSQL.

1. DATA MODEL & INDEXING:
- Create a Product schema containing id, title, description, tags, price, in_stock (boolean), primary_image_url, and embeddings array.
- Set up a Typesense collection for 'products' configured for hybrid search: text fields (title, description, tags) + vector field (embedding, 1536 dimensions).
- Build a sync pipeline that listens to product webhook events (create, update, delete) to update both PostgreSQL and Typesense.
- Whenever a product is saved, trigger OpenAI API (text-embedding-3-small) on title + description and upsert vector data to Typesense.

2. SEARCH API & MERCHANDISING:
- Endpoint POST /api/search: accepts query string, page, limit, price_min, price_max, sort_by.
- Execute hybrid search on Typesense combining vector search and BM25 text match with weights.
- Filter out out-of-stock items (in_stock=false) unless specified otherwise.
- Merchandising logic: Allow defining 'boost' rules (e.g., if query matches 'jacket', pin SKU_123 to rank 1) and 'synonym' dictionaries.

3. RECOMMENDATION ENGINE:
- Create an event tracking endpoint POST /api/telemetry to record events: view_item, add_to_cart, purchase with session_id, user_id, sku.
- Endpoint GET /api/recommendations/related?sku=XYZ: calculate product similarity using vector cosine distance in Typesense.
- Endpoint GET /api/recommendations/frequently-bought-together?sku=XYZ: query Postgres order histories for products co-occurring in orders containing sku=XYZ in the last 90 days, ranked by frequency.

4. FRONTEND WIDGET:
- Create a plain vanilla JS script embed (<script src='search.js'>) that injects a search modal trigger into standard e-commerce templates.
- Show fast dynamic autocomplete dropdown with product image, price, and highlighted title match as the user types (debounce 150ms).
- Include telemetry tracking functions on cart/checkout actions.

5. EDGE HANDLING:
- Handle OpenAI API timeouts by falling back strictly to BM25 keyword search in Typesense.
- Ensure API responses are cached using Redis for exact repeated search queries for 5 minutes.

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