E-Commerce Personal Shopper Agent

Write a production-ready system prompt for an e-commerce personal shopper AI agent for [STORE NAME], a [STORE DESCRIPTION] online retailer. The agent helps customers find products and complete purchases. Requirements: 1) Role: a knowledgeable personal shopper who knows the entire catalog, 2) Style profiling: on first interaction, ask 3 quick questions to understand the customer's style, size, and budget, 3) Product search: translate natural language queries into structured search parameters (category, size, color, price range, style keywords), 4) Recommendations: always suggest 3 options with different price points, explain why each matches their stated preferences, 5) Size guidance: proactively ask about sizing concerns, reference the brand's size chart, 6) Upsell/cross-sell: suggest complementary items naturally ("This jacket pairs well with..."), 7) Cart assistance: help customers complete checkout, answer shipping/return questions, 8) Prohibited: never pressure customers, never claim items are "almost sold out" unless inventory data confirms it. Include 3 few-shot examples.

Build a Node.js TypeScript product recommendation API for an e-commerce personal shopper agent. Endpoint: POST /api/recommendations. Input: { userId, query, styleProfile, cartItems[], budget }. Requirements: 1) Natural language parsing: extract category, color, size, price range, style keywords from the query, 2) Catalog search: full-text search against product catalog with facet filtering, 3) Personalization: boost products matching the user's style profile (past purchases, saved items, browsing history), 4) Diversity: ensure recommendations span different price points (budget, mid-range, premium), 5) Availability check: filter out out-of-stock items, 6) Response: return top 6 products with { productId, name, price, imageUrl, matchReason, sizeAvailability }, 7) Cache: cache results in Redis for 5 minutes per userId+query combination. TypeScript + Drizzle ORM. Include the product catalog schema.

Design a conversational cart recovery flow for an e-commerce AI agent. Trigger: customer has items in cart but hasn't checked out in 24 hours. The agent should: 1) Opening: acknowledge their saved cart naturally ("I noticed you left some items behind — want me to help you finish up?"), 2) Barrier identification: ask what stopped them (price concern, sizing question, shipping cost, just browsing), 3) Objection handling: for each barrier type — price concern (offer discount code if available, highlight value), sizing question (walk through size guide), shipping cost (mention free shipping threshold or current promotions), 4) Social proof: mention how many others bought the same item recently (if data available), 5) Urgency (honest only): mention if item is low stock (only if true), 6) Checkout assist: offer to walk them through checkout step by step, 7) Exit: if they decline, save their cart and offer to send a reminder email. Write as a TypeScript state machine with explicit states, transitions, and API calls.