Can I vibe code Prefixbox?

Build a full-stack E-commerce Hybrid Search and Autocomplete Service using TypeScript, Node.js, and Meilisearch/Qdrant. 

DATA MODEL & ENTITIES:
- Store: id, api_key, platform ('shopify'|'magento'|'custom'), settings
- Product: id, store_id, external_id, title, description, vendor, product_type, tags, price, in_stock, score_boost, dynamic_attributes (JSONB)
- Variant: id, product_id, title, sku, price, image_url, attributes (JSONB)
- SearchRule: id, store_id, query_trigger, action_type ('pin'|'boost'|'hide'|'banner'), target_product_id, banner_url, priority
- SynonymGroup: id, store_id, terms (array of strings)

CORE FUNCTIONALITY:
1. Catalog Syncing:
   - Webhook endpoints to ingest product creation, update, and deletion events from Shopify and BigCommerce.
   - Batch indexing pipeline to parse catalog attributes and push text fields into Meilisearch/Typesense alongside generated OpenAI text-embedding-3-small embeddings in Qdrant.

2. Hybrid Search Engine API:
   - Endpoint GET /api/search?q={query}&filters={json}&page={page}:
   - Perform Reciprocal Rank Fusion (RRF) combining BM25 keyword matching with dense vector similarity.
   - Apply SearchRules: apply custom product score boosts, pin specific product IDs to top slots (e.g., position 1), and inject promotional banners into search results.
   - Return paginated product hits with metadata, available dynamic facets (category, price range, brand), and execution time (<50ms target).

3. Rich Autocomplete API:
   - Endpoint GET /api/autocomplete?q={query}:
   - Return instant suggestions split into categories: Query Suggestions, Suggested Products (with thumbnails/prices), and Matching Categories/Brands.
   - Implement strict sub-30ms response caching using Redis.

4. Merchandising Dashboard:
   - Admin UI to create and manage drag-and-drop Product Pinning, Boost Rules based on inventory/margin, Synonym mapping, and Custom Redirects for zero-result queries.

EXPLICITLY OUT OF SCOPE:
- Conversational LLM AI Chat widgets.
- Multilingual translation pipelines.
- Native CDP/user-history personalization algorithms.

FAILURE MODES TO HANDLE:
- Debounce high-frequency autocomplete requests on front-end widgets.
- Fallback gracefully to BM25 keyword search if OpenAI API embedding generation times out or fails.
- Exclude out-of-stock variants from top results unless explicit rules state otherwise.

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