feat: Introduce AI extraction schema with validation, prompt generation, and a verification script to improve price extraction reliability.

app/ai_schema.py

@@ -141,7 +141,9 @@ EXTRACTION_PROMPT_TEMPLATE = """Extract product price and stock status from this
 - Look for: "PRIX DÉTECTÉ:", price tags, "Prix:", "€", numbers near "Ajouter au panier"
 - Extract as DECIMAL NUMBER: If you see "3.99", return 3.99
 - Ignore crossed-out/barré prices (old prices)
-- If multiple prices, take the current/main price (not the original)
+- **CRITICAL:** Ignore "Prix au litre", "Prix au kg", "P.U.", or unit prices usually shown in smaller text/parentheses (e.g., "4.60 € / L").
+- If multiple prices, take the current/main price (not the original, not the unit price)
+- **B&M STORES Specific:** The main price is often large and bold (e.g. "1.15€"), while unit price is small (e.g. "4.60 €/L"). ALWAYS take the main price.
 
 **CRITICAL - Common mistakes to avoid:**
 - "3.99 €" means 3.99 (NOT 399.00, NOT 3990.00)
@@ -161,7 +163,7 @@ EXTRACTION_PROMPT_TEMPLATE = """Extract product price and stock status from this
 - If unclear: set null and confidence < 0.5
 
 **STOCK:**
-- TRUE if: "Ajouter au panier", "Acheter", "En stock", "Disponible", "Add to Cart"
+- TRUE if: "Ajouter au panier", "Acheter", "En stock", "Disponible", "Add to Cart", "Retrait 2h", "Click & Collect"
 - FALSE if: "Rupture", "Indisponible", "Épuisé", "Out of Stock", "Notify Me"
 - NULL if unclear or not shown
 

app/services/tracking_scraper_service.py

@@ -416,9 +416,40 @@ class ScraperService:
 
         try:
             logger.info(f"Extracting text (limit: {text_length} chars)...")
-            raw_text = await page.inner_text("body")
-            page_text = raw_text[:text_length]
-            logger.info(f"Extracted {len(page_text)} characters")
+            
+            # Smart extraction: remove noise (nav, footer, scripts) before getting text
+            clean_text = await page.evaluate("""
+                () => {
+                    // Clone body to not affect the visual page
+                    const clone = document.body.cloneNode(true);
+                    
+                    // Remove noise selectors
+                    const noiseSelectors = [
+                        'nav', 'header', 'footer', 'script', 'style', 'noscript', 'iframe',
+                        '.cookie-banner', '.popup', '#menu', '.menu', '.sidebar',
+                        '[role="navigation"]', '[role="banner"]', '[role="contentinfo"]'
+                    ];
+                    
+                    noiseSelectors.forEach(selector => {
+                        const elements = clone.querySelectorAll(selector);
+                        elements.forEach(el => el.remove());
+                    });
+                    
+                    return clone.innerText;
+                }
+            """)
+            
+            # Fallback if cleaning removed everything (unlikely but safe)
+            if not clean_text or len(clean_text) < 100:
+                logger.warning("Cleaned text too short, falling back to full body text")
+                clean_text = await page.inner_text("body")
+
+            # Collapse whitespace
+            import re
+            clean_text = re.sub(r'\s+', ' ', clean_text).strip()
+
+            page_text = clean_text[:text_length]
+            logger.info(f"Extracted {len(page_text)} chars")
             return page_text
         except Exception as e:
             logger.error(f"Text extraction failed: {e}")