Can I vibe code Sizekick?

Build a lightweight garment size recommendation microservice consisting of a React-based embeddable PDP widget, a Node.js/TypeScript REST API, and a PostgreSQL database.

1. Data Schemas: Implement tables for GarmentSpecs (id, SKU, brand_id, category, chest_cm, waist_cm, hip_cm, shoulder_cm, sleeve_cm, stretch_factor), SizeCharts (brand_id, size_name, chest_min, chest_max, waist_min, waist_max, hip_min, hip_max), and RecommendationLogs (session_id, sku, user_height, user_weight, body_shape, target_fit, recommended_size, confidence_score).

2. Recommendation Core Engine: Implement a sizing calculation module. Use standard anthropometric regression formulas (estimating chest, waist, and hip circumference from height, weight, age, and selected body shape) combined with user-provided direct measurements if available. Compare computed body bounds against the GarmentSpecs data adjusted by the garment stretch factor and user fit preference (tight, regular, loose).

3. Storefront PDP Widget: Build a client-side JavaScript bundle (<20KB) that injects a 'Find My Size' trigger adjacent to the native ecommerce size selector. Upon trigger, open a multi-step modal capturing user inputs. Provide instant size recommendations with detailed explanations (e.g., 'Size M will fit tightly across shoulders; Size L is recommended for a relaxed fit').

4. Return Feedback Calibration Engine: Implement a webhook handler for ecommerce return events. When an item is returned for fit reasons (e.g., 'too small'), adjust the dynamic offset variable for that specific SKU and size in the recommendation logic.

5. Edge & Failure Cases: Handle missing spec data gracefully by falling back to category-wide size averages. Manage metric (cm) vs imperial (in) conversions cleanly. Ensure no personally identifiable biometric data is logged.

6. Out of Scope: Real-time browser webcam video streaming, TensorFlow.js human body pose extraction, 3D body mesh generation, and biometric image data storage.

$ each button prefixes agent-specific run instructions · build your own product, never copy proprietary code, trademarks or designs