AI-native commerce platform that dynamically adapts the shopping interface based on detected user intent. Uses LangChain/LangGraph for agent orchestration and real-time personalization.
What I Learned
LangGraph is powerful but the learning curve is real
Intent detection is more useful than recommendation engines
The UI adaptation needs to be subtle or it feels uncanny
The Problem
Traditional e-commerce shows the same interface to everyone regardless of intent. Someone researching is different from someone ready to buy.
The Approach
Building a commerce experience that detects user intent in real-time and adapts the interface. Uses a multi-agent system where different agents specialize in different shopping modes.
Architecture Decisions
Intent detection via multi-model ensemble
LangChain/LangGraph agent orchestration
Product extraction pipeline with embeddings
Adaptive UI component system
Why I Built It This Way
Used LangGraph for complex agent workflows
Built intent-first architecture (detect intent before showing UI)
Implemented multi-model approach for accuracy
Technology Stack
Next.js 16React 19TypeScriptSupabaseLangChainLangGraphAnthropic APIOpenAI APIGoogle AI APITailwind CSS
Outcomes
Intent-first shopping experience
Multi-agent commerce orchestration
Adaptive UI component library
Started: 2025-11
Active development