Search by Category

Workflow

1. Map the user's request to one or more categories. Examples:
   • "headlamps" → lighting, tactical
   • "DSLR camera bag" → photography, accessories
   • "gaming mouse" → gaming, peripherals
   • "UPS" → power_battery
2. Filter the store list:

   stores = [s for s in data
             if s["is_tech"] is not False
             and any(cat in (s.get("categories") or []) for cat in target_cats)]
   

3. Pick up to 8 stores to actually search. Prioritise:
   • Stores tagged is_tech: true over null
   • Stores whose description mentions the user's specific product type
   • When in doubt, prefer stores with multiple matching categories
4. Translate the query to Hebrew (keep brand/model names in Latin).
5. Search each store with the chosen backend:
   • Playwright: try https://{domain}/?s={query} or /search?q={query} — most Israeli sites use one of these
   • Tavily: site:{domain} <hebrew query>
6. Apply VAT rules from data/shopping-rules.md.
7. Rank low-to-high and report.

Output

## {product} — category search ({categories joined})
Hebrew query: {hebrew}
Stores searched: {count}
Backend: {Playwright | Tavily}

| # | Price (₪, inc. VAT) | Product | Store | Link |
|---|---------------------|---------|-------|------|

If a store's site has no working search, skip it and note in output. Never invent results.

Rules

• Only navigate to stores where is_tech is true or null. Never false.
• If the user's request doesn't map to any category, fall back to search-main-tech-stores and tell the user why.
• Cap at 3 results per store.