Expedia's Custom GPT End-to-end architecture
1. System Architecture Overview
The System Architecture Overview diagram illustrates the end-to-end structure of the Expedia Travel Recommendation Service (ETRS). It shows how user or partner applications — including ExpediaGPT — connect securely to the ETRS API Gateway, which routes requests to Expedia’s verified Data Services layer. Each tier is modular: the top layer handles user intent and interfaces, the gateway layer enforces authentication and schema validation, and the data layer retrieves, filters, and enriches live content from Expedia’s core platforms (Hotels, Flights, Cars, and Activities). This architecture ensures high scalability, fault isolation, and consistent data accuracy across all integrations.
2. OAuth Authentication Flow
The OAuth Authentication Flow visual explains how partners and internal clients authenticate safely with Expedia’s APIs. Using the OAuth 2.0 Client Credentials Grant, an integration first requests an access token from Expedia’s OAuth server, receives a short-lived bearer token, and then uses it to call the ETRS endpoints. This model prevents hard-coded secrets in applications and allows centralized revocation, token rotation, and granular permission control. The diagram highlights the four key steps: token request, token issuance, authenticated request, and response delivery — ensuring that all API interactions remain encrypted, verifiable, and compliant with Expedia’s security standards.
3. Data Flow – ExpediaGPT → ETRS → Expedia Data
The Data Flow diagram traces how a natural language query from a user passes through ExpediaGPT, is structured into parameters, and then transmitted to the ETRS gateway. ETRS processes this structured request, fetches relevant listings from Expedia’s internal datasets, and returns a clean JSON payload. ExpediaGPT interprets this data and presents it conversationally — for example, turning “Find hotels in Tokyo” into verified, bookable hotel options with Expedia links. This visual demonstrates the seamless orchestration between AI-driven query understanding and Expedia’s real-time data infrastructure, ensuring both accuracy and personalization in responses.
4. Endpoint Relationships (Lodging, Flights, Activities, Cars)
The Endpoint Relationships diagram maps how the ETRS gateway connects to Expedia’s four main travel data domains — Lodging, Flights, Activities, and Cars. Each endpoint corresponds to a specific structured dataset, yet all are served through a single gateway that standardizes authentication, schema validation, and error handling. The diagram conveys ETRS’s hub-and-spoke design: a centralized access layer feeding multiple specialized data services. This unified approach enables consistent developer integration, cross-domain recommendations, and simplified maintenance of Expedia’s large-scale travel APIs.
5. Failover & Monitoring Topology
The Failover & Monitoring Topology diagram illustrates ETRS’s global reliability model. Expedia operates an active primary region (e.g., us-east-1) and a warm standby failover region (e.g., eu-central-1) with near real-time replication and an RTO of under five minutes. Should a primary outage occur, DNS-based routing automatically redirects traffic to the backup region. A centralized monitoring layer — powered by Datadog, Grafana, and PagerDuty — continuously tracks availability, latency percentiles, and error rates. This redundancy and observability framework guarantees service continuity, proactive detection, and compliance with Expedia’s 99.9% uptime SLA.
