Implement SQLite-based product mapping persistence & scraped metadata

[ddvlamin]

Summary

Integrate a persistent SQLite database for resolving recipe ingredients to specific Colruyt product IDs using rapidfuzz for string similarity lookup, and scraping detailed product specifications from the Colruyt RTI portal.

Motivation

• Persist User Choices: Store associations between a cleaned ingredient name and a specific Colruyt product.
• Fuzzy Search Lookup: Query the database using fuzzy matching (rapidfuzz) before hitting the live APIs or asking the user.
• Scrape Rich Metadata: Dynamically fetch product details (description, content, conservation info, usage info) from rti.colruytgroup.com so the AI assistant can auto-resolve ambiguities before prompting the user.
• MCP API Refactoring: Keep the server interface clean and focused by exposing high-level resolution and mapping tools.

Proposed Solution

1. Database Storage and Schema:
   • Database File: resources/colruyt_mappings.db.
   • Table product_mappings: cleaned_ingredient (TEXT PRIMARY KEY), product_id (TEXT NOT NULL), product_name (TEXT NOT NULL), product_brand (TEXT), product_description (TEXT), conservation_info (TEXT), usage_info (TEXT), content (TEXT), created_at (TIMESTAMP DEFAULT CURRENT_TIMESTAMP).
2. Fuzzy Matching:
   • Install rapidfuzz as a dependency.
   • Use rapidfuzz.process.extractOne to check the database before executing search (threshold >= 80/100, token_sort_ratio).
3. Scraping Strategy:
   • Fetch specifications from rti.colruytgroup.com/nl/product-info/{technicalArticleNumber} matching the first GTIN or first available barcode block.
4. MCP Tools changes:
   • Expose new resolve_ingredient and save_product_mapping tools.
   • Keep add_recipe_to_list and add_items_to_list.
   • Remove get_most_bought_products and search_products from public tools.