Pricing Tracker

Mission

Collecter prix actuel et disponibilité depuis multiples retailers (Amazon, sites spécialisés selon catégorie).

Outils de Scraping

Priorité 1: MCP Apify (Recommandé)

Outil	Usage
mcp__apify__call-actor avec apify/amazon-product-scraper	Prix Amazon structuré (inclut prix, stock, shipping)
mcp__apify__apify-slash-rag-web-browser	Prix autres retailers (Decathlon, Darty, etc.)

Priorité 2: Fallback

Outil	Usage
WebFetch	Si MCP échoue ou timeout (>2 min)

Avantages MCP Apify:

• Amazon: Prix exact, stock, shipping en JSON structuré
• RAG Browser: Recherche + extraction prix en une seule requête

Quick Summary

1. Amazon: MCP apify/amazon-product-scraper (ou WebFetch fallback)
2. Autres retailers: MCP rag-web-browser (ou WebFetch fallback)
3. Extract: prix, disponibilité, shipping, promos actives
4. Save comparison table JSON + cache 7j

Inputs

• product_name: Nom produit
• category: Catégorie (pour sélectionner retailers appropriés)
• amazon_url: URL Amazon (optionnel)

Outputs

• pricing.json: Tableau prix par retailer
• Cache 7j

Dependencies

• data/category_specs.yaml (retailers par catégorie)
• Load helpers/retailers.yaml when scraping for patterns

Workflow

1. Load retailers

retailers = category_specs.yaml[category].retailers
// Example velo: ["decathlon.fr", "alltricks.fr", "probikeshop.fr"]

2. Scrape Amazon (if in retailers)

**PRIMARY: MCP Apify**
mcp__apify__call-actor({
  actor: "apify/amazon-product-scraper",
  step: "call",
  input: {
    categoryOrProductUrls: [amazon_url],
    maxItems: 1
  }
})

mcp__apify__get-actor-output({
  datasetId: result.datasetId,
  fields: "title,price,availability,delivery"
})
→ Returns: { price: 549, availability: "In Stock", delivery: "Livraison gratuite" }

**FALLBACK: WebFetch** (if MCP fails)
WebFetch product page + extract with retailers.yaml selectors

3. Scrape other retailers (2-3)

For each retailer:
  query = "{product_name} prix site:{retailer}"

  **PRIMARY: MCP Apify RAG Web Browser**
  mcp__apify__apify-slash-rag-web-browser({
    query: query,
    maxResults: 1,
    outputFormats: ["markdown"]
  })

  Parse Markdown:
  "Extract pricing information:
   - Current price (number only)
   - Stock availability
   - Shipping cost
   - Active promotions
   Return as JSON."

  **FALLBACK: WebFetch**
  WebFetch product page + extract with retailers.yaml selectors

4. Build comparison table

{
  "product": product_name,
  "timestamp": now,
  "retailers": [
    {
      "name": "Amazon",
      "price": 549,
      "availability": "in_stock",
      "shipping": "Gratuit",
      "promo": null
    },
    {
      "name": "Decathlon",
      "price": 549,
      "availability": "in_stock",
      "shipping": "Gratuit en magasin",
      "promo": "-10% membres"
    }
  ],
  "best_price": 494.10,  // Decathlon with promo
  "best_retailer": "Decathlon"
}

5. Save + cache

• Save: data/research_{timestamp}/{product}/pricing.json
• Cache: data/cache/pricing/{cache_key}.json (7j TTL)

Error Handling

• Retailer unavailable → Skip, continue with others
• Price not found → Mark as "N/A"
• Out of stock → Mark availability = false
• MCP timeout → Fallback to WebFetch automatically