Agentic Commerce Architecture

Unified Data Model & Agent Handoff Mechanism for AI-Ready Experience Booking

Complete System Architecture Diagram

Layer 1: Data Sources

Regular Source (Partner APIs)

Automated
  • • GetYourGuide API experiences
  • • Selected from partner catalog
  • • Real-time availability
  • • Automatic pricing updates
  • • Widget-based booking flow
  • • Instant confirmation

Premium Source (Manual Curation)

Curated
  • • Manually created experiences
  • • Hand-picked by CP team
  • • Quality verified partnerships
  • • Simple redirect to partner website
  • • No booking flow on platform
  • • Booking handled by partner

Layer 2: Data Normalizer Service

Transformation Engine

1. Extract
  • • Pull from both sources
  • • Parse different formats
  • • Validate data integrity
2. Transform
  • • Map to UOO schema
  • • Normalize pricing (EUR)
  • • Calculate agent_score
  • • Categorize uniformly
3. Load
  • • Cache in Redis/Memory
  • • Index for search
  • • Set expiry rules

Key Logic: Every offer gets a unique uoo_id, preserves source_specific_id, and is tagged withsource_tier for tracking.

Layer 3: Unified Offer Cache

In-Memory Cache / Redis

Cached Data
  • • All Unified Offer Objects
  • • Search indexes
  • • Category mappings
  • • Price distributions
Cache Strategy
  • • TTL: Regular = 15min, Premium = 24h
  • • LRU eviction policy
  • • Preload popular offers
  • • Background refresh

Layer 4: Agent API Gateway

AI Agent Interface

Endpoints
  • GET /api/agent/search
  • GET /api/agent/offers/:id
  • POST /api/agent/handoff
  • GET /api/agent/compare
Response Format
  • ✓ Source-agnostic UOO objects
  • ✓ Standardized pricing
  • ✓ Comparison scores
  • ✓ Deep-link ready

Agent Promise: The AI agent receives completely normalized data and doesn't need to handle source-specific logic. The source_tier field is visible for transparency but all technical complexity is abstracted.

Layer 5: Frontend & Handoff

Dynamic Deep-Link Generator

Based on source_tier:

Regular: Generate partner tracking URL with affiliate params
Premium: Generate internal redirect with tracking

Agent Delegation UI

  • • User grants agent permission
  • • Source transparency display
  • • Clear booking flow explanation
  • • One-click handoff to booking
  • • Tracking & analytics capture

Implementation Roadmap

Phase 1: Backend (2-3 weeks)

  • Create UOO schema in database
  • Build Data Normalizer service
  • Implement caching layer (Redis)
  • Create Agent API endpoints
  • Build deep-link generator logic

Phase 2: Frontend (1-2 weeks)

  • Agent permission dialog UI
  • Source transparency cards
  • Handoff confirmation screen
  • Tracking redirect logic

Phase 3: Testing & Launch (1 week)

  • Test with mock AI agent
  • Verify tracking & analytics
  • Load testing cache layer
  • Partner with AI platforms