Search Suppliers

Files Used by This Skill

Firecrawl MCP Tool Selection

Always choose the cheapest tool that gets the job done:

Golden rule: firecrawl_search → firecrawl_scrape (2 credits) before escalating to firecrawl_extract (15 credits). Only use extract when the scraped markdown is genuinely too unstructured to parse.

Core Workflow

Step 1: Parse Intent

Before searching, determine:

Step 2: Search Phase

For each target supplier, execute a search using the patterns from supplier-registry.md:

For DB-integrated suppliers (Scoop, MiRO, Nology)

1. Check DB first:
   GET /api/admin/suppliers/products?supplier={code}&search={query}
   
   or via Supabase:
   SELECT * FROM supplier_products
   WHERE supplier = '{code}' AND name ILIKE '%{query}%'
   ORDER BY cost_price_zar ASC LIMIT 10;

2. If DB has results → show DB data (synced price + stock)
3. Then optionally enrich with live Firecrawl data (specs, datasheets)
4. If NOT in DB → proceed to Firecrawl live search

Important for Scoop: Never use Firecrawl for Scoop pricing — dealer cost prices require login. Use DB for all Scoop pricing data.

For Rectron (Firecrawl-only)

1. First time? Run firecrawl_map to discover site structure:
   firecrawl_map: { url: "https://www.rectronzone.co.za", search: "{query}" }
   → Saves discovered URL patterns to supplier-registry.md

2. Then search:
   firecrawl_search: { query: "site:rectronzone.co.za {query}" }
   → Returns product page URLs

Search execution per supplier

firecrawl_search:
  query: "site:{supplier_domain} {search_terms}"
  limit: 5
  scrapeOptions:
    formats: ["markdown"]
    onlyMainContent: true

Parse the returned markdown snippets to identify product URLs. If snippets contain enough pricing/spec data, you may not need to scrape individual pages.

Step 3: Extract Phase

For each product URL found, choose based on content quality:

If markdown snippet from search is sufficient:

• Parse directly — no additional Firecrawl call needed

If you need full page content:

firecrawl_scrape:
  url: "{product_url}"
  formats: ["markdown"]
  onlyMainContent: true

Parse the markdown for: product name, SKU, price (excl/incl VAT), stock status, key specs, datasheet links.

If page is too complex for manual parsing (JavaScript-heavy, spec tables, nested tabs):

firecrawl_extract:
  urls: ["{product_url}"]
  prompt: "Extract product name, SKU, pricing (excl and incl VAT), stock status, all specifications, and any datasheet PDF links"
  schema: <Hardware Product Schema from extraction-schemas.md>

For cross-supplier comparison of the same product:

firecrawl_extract:
  urls: ["{scoop_url}", "{miro_url}", "{nology_url}", "{rectron_url}"]
  prompt: "Extract pricing, stock status and warranty for this product from each supplier"
  schema: <Cross-Supplier Comparison Schema from extraction-schemas.md>

Step 4: Compare and Present

Format all results into a clear comparison. Always show:

1. Cross-supplier comparison table (if multiple suppliers searched)
2. Best value recommendation with reasoning
3. Margin potential — rough estimate based on standard CircleTel markup vs competitor retail

Margin estimate formula:

estimated_retail  = supplier_cost × 1.35  (35% target margin)
estimated_margin  = estimated_retail - supplier_cost
margin_pct        = estimated_margin / estimated_retail × 100

Flag if estimated retail would be uncompetitive vs SA ISP market benchmarks from references/competitor-benchmarks.md.

Datasheet handling:

• If a datasheet URL is found, display it prominently
• Note: products/research/hardware/datasheets/ is where PDFs should be saved for offline reference
• If the user says "download the datasheet", use firecrawl_scrape with parsers: ["pdf"] on the datasheet URL

Step 5: Store Results

Always store if:

• User used --save flag
• User explicitly asks to save
• Research took significant effort (multiple suppliers, 5+ products)

Store to: products/research/hardware/{YYYY-MM-DD}-{slugified-query}.md

Use the template at templates/research-result.md. Fill in all sections:

• Header: date, query, suppliers searched, credits used
• Cross-supplier comparison table
• Best candidate + rationale
• Per-product detail blocks
• Datasheets & document links
• Supplier notes (update registry if new patterns found)
• Recommended next actions

Search Modes (Detailed)

Mode 1: Quick Search

/product-search MikroTik hAP ax3
/product-search PoE switch 24 port gigabit
/product-search "access point" outdoor 5GHz

Searches all 4 suppliers. Returns comparison table. Does not auto-save unless --save is added.

Mode 2: Single Supplier

/product-search --supplier miro Reyee RG-RAP2260
/product-search --supplier rectron UPS 3000VA tower
/product-search --supplier nology "dell server" rack 1U

Targets one supplier only. Faster and cheaper (fewer Firecrawl calls).

Mode 3: Cross-Compare

/product-search --compare "MikroTik hAP ax3"
/product-search --compare "Ubiquiti UniFi AP U6 Pro"

Finds the exact product on all suppliers and extracts structured data from each for a side-by-side comparison. Uses firecrawl_extract with the Cross-Supplier Comparison Schema.

Mode 4: Site Discovery (First-Time Setup)

/product-search --map rectron
/product-search --map {new-supplier-domain}

Uses firecrawl_map to discover the site structure and URL patterns. After running, update supplier-registry.md with the discovered patterns so future searches are more precise.

Mode 5: Category Browse

/product-search --category "wireless access points" --supplier miro
/product-search --category routers --max-price 3000

Browses a category rather than searching for a specific product. Uses category URLs from the supplier registry.

Mode 6: Datasheet Hunt

/product-search --datasheet "Peraso PRS4000"
/product-search --datasheet "Tarana G1 Remote Node"

Specifically searches for product datasheets/spec sheets. Uses firecrawl_search targeting PDF documents: site:{domain} filetype:pdf {product_name} or searches product pages for linked PDFs.

DB vs Live Search Decision Tree

User asks for product search
         │
         ▼
Product in our DB (Scoop/MiRO/Nology synced)?
    │ YES                         │ NO (or Rectron/other)
    ▼                             ▼
Show DB results first         Use Firecrawl live search
    │                             │
    ▼                             ▼
User wants live pricing?      Find on supplier site(s)
    │ YES (not Scoop)             │
    ▼                             ▼
Firecrawl scrape for          Extract structured data
live price check              and present results
    │
    ▼
User wants datasheet?
    │ YES
    ▼
Search for PDF on product page
or use firecrawl_search with filetype:pdf

Supplier-Specific Tips

Scoop (scoop.co.za)

• Pricing: Use DB only (login wall). Firecrawl for specs/datasheets only.
• Search: site:scoop.co.za {query} works well
• Datasheets: Often linked directly on product pages as PDFs
• Tip: Scoop carries a lot of Ubiquiti and MikroTik — strong for infrastructure builds

MiRO (miro.co.za)

• Pricing: Public pricing available — Firecrawl is reliable
• Search: Good results from site:miro.co.za {query}
• Specialty: Best for ISP/FWB hardware (Tarana CPE, DUNE 60GHz Peraso units)
• Tip: Check their "ISP Hardware" and "Fixed Wireless" categories specifically for CircleTel-relevant CPE

Nology (nology.co.za)

• Pricing: Public pricing, VirtueMart CMS
• Search: Use firecrawl_extract with schema — VirtueMart pages can be messy as markdown
• Specialty: Enterprise IT, UPS, servers
• Tip: Good source for Managed IT Services hardware components (workstations, NAS, rack servers)

Rectron (rectronzone.co.za)

• Pricing: Public pricing available
• Search: Run --map rectron first to discover URL structure
• Specialty: Broadest range — networking, compute, peripherals, consumer electronics
• Tip: Check Rectron when MiRO/Scoop/Nology don't have a product — they cover many more brands

Integration with Other Skills

→ /product-analyze

After finding a product candidate, pass the extracted data to /product-analyze for a full market-fit analysis including margin calculation, competitive positioning, and GO/NO-GO recommendation.

After /product-search finds a candidate:
"Run /product-analyze on the {product name} at R{cost_price}"

→ /new-product

When designing a new product in /new-product Phase 2 (Architecture & Technology), use search results to fill in:

• CPE/hardware details (question 4)
• Installation cost (from supplier price × install hours)
• Hardware amortisation (supplier price ÷ 12 or 24 months)

→ /product-browse

/product-browse queries the Supabase DB. /product-search queries live websites. Run browse first for known products; search for discovery or when DB data is stale.

Research Storage Structure

products/research/hardware/
├── datasheets/                          ← PDFs downloaded from supplier sites
│   └── {manufacturer}-{model}-datasheet.pdf
├── {YYYY-MM-DD}-{slug}.md               ← Research result files
└── .gitkeep

File naming examples:

• 2026-04-03-mikrotik-hap-ax3-comparison.md
• 2026-04-03-outdoor-access-point-5ghz.md
• 2026-04-10-rectron-ups-3000va.md

Updating the Supplier Registry

When you discover new URL patterns, site behaviours, or pricing quirks during a search, update supplier-registry.md:

1. Open supplier-registry.md
2. Find the supplier block
3. Add/update URL patterns, notes, or pricing observations
4. Save — this knowledge persists for all future searches

This is especially important for Rectron since it starts with no URL patterns.

Version: 1.0.0 Last Updated: 2026-04-03 Maintained By: CircleTel Product Strategy