Can I vibe code Celebros?

Build an open-source hybrid site search engine and light merchandising service for an ecommerce store using Next.js (App Router), Typesense, and Tailwind CSS.

1. DATA MODEL:
- Store Products: id, title, description, vendor, category, tags, price, in_stock, dynamic_attributes (JSONB), vector_embedding (array of floats).
- Search Overrides: id, query_pattern, pinned_product_ids (array), boosted_category_ids (array), redirect_url (nullable).
- Search Logs: id, raw_query, normalized_query, result_count, clicked_product_id (nullable), converted (boolean), created_at.

2. HYBRID SEARCH & MERCHANDISING LOGIC:
- Expose a POST /api/search endpoint accepting query string, page, limit, and active facet filters.
- First, check Search Overrides for an exact query match. If a redirect_url exists, return a redirect payload immediately.
- Send query to Typesense to perform hybrid search combining text BM25 matching on title/description and vector similarity via text-embedding-3-small.
- Apply any pinned_product_ids to force specified SKUs to positions 1-N of the returned array regardless of relevancy score.
- Exclude out-of-stock items or demote them to the bottom of the search results array.

3. STOREFRONT & UI COMPONENTS:
- Build an instant search bar modal component with keyboard shortcuts (Cmd+K), debounce (150ms), and auto-complete suggestions.
- Render faceted navigation sidebar (price slider, vendor checkboxes, category tree) updated dynamically from search response aggregation buckets.
- Track query execution time and render results grid under 100ms.

4. ANALYTICS & CLICK TRACKING:
- Log every execution to Search Logs asynchronously. Attach event listener on product card clicks to capture clicked_product_id against search session ID.

5. OUT OF SCOPE:
- Do not build full ERP catalog sync. Rely on standardized JSON product feed ingestion via API.

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