Can I vibe code Bloomreach Discovery?

Build a high-performance E-Commerce Search and Merchandising microservice in Node.js/TypeScript using Express, Qdrant (vector search), and PostgreSQL (relational catalog and rules storage).

The system must handle three core operations: Catalog Ingestion, Search Query Pipeline, and Visual Merchandising Overrides.

DATA MODELS (PostgreSQL / Prisma):
1. Product: id, external_id, title, description, category_path, price, inventory_count, brand, attributes (JSONB), status ('active'|'draft').
2. MerchandisingRule: id, search_term, pinned_product_ids (text array), hidden_product_ids (text array), boost_factors (JSONB - e.g. {"brand:Nike": 1.5}), valid_from, valid_to.
3. SearchEvent: id, query, user_id, clicked_product_id, converted (boolean), created_at.

CATALOG INGESTION API:
- POST /api/v1/products/sync: Bulk ingestion endpoint accepting up to 1,000 products per batch.
- Persist relational metadata to PostgreSQL.
- Generate text embedding using OpenAI text-embedding-3-small combining title, description, brand, and category_path.
- Upsert dense vector payload to Qdrant collection 'catalog_products' containing vector, price, brand, inventory_count, and category_path as indexed payload fields.

SEARCH QUERY PIPELINE (GET /api/v1/search?q=query&category=cat&page=1&limit=20):
1. Normalization & Rules Check: Normalize query string (lowercase, trimmed). Query PostgreSQL for active MerchandisingRules matching search_term.
2. Hybrid Vector Search:
   - Generate embedding for 'q' via OpenAI.
   - Execute Qdrant vector search filtering out inventory_count <= 0 and matching category if provided.
   - Fetch top 100 candidate product IDs with similarity scores.
3. Ranking & Merchandising Pipeline:
   - Apply boost factors from active MerchandisingRule (multiply similarity score by factor).
   - Inject pinned_product_ids at explicit index positions (0-indexed).
   - Filter out hidden_product_ids entirely.
   - Sort final array and paginate.
4. Response Payload: Return items array (with full PostgreSQL product metadata), total_count, current_page, and query_id (for event tracking).

ANALYTICS & AUTOMATED BOOSTING:
- POST /api/v1/events: Record user search clicks and conversions associated with query_id.
- Implement a background worker using node-cron that runs hourly to calculate Click-Through Rate (CTR) per product-query pair from SearchEvent. If CTR > 15%, automatically insert a temporary dynamic boost factor (+20% score multiplier) for that product in future identical searches.

FAILURE MODES & OUT OF SCOPE:
- Fallback: If Qdrant is unavailable, fall back to PostgreSQL ILIKE keyword matching over title and brand.
- Out of scope: Automated multi-language translation, frontend widgets/UI components, real-time multi-currency exchange conversion.

Provide fully working Express routes, Prisma schema definitions, Qdrant client configurations, and raw SQL/vector queries in standard TypeScript.

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