Use Exa.AI for neighborhood discovery
Context
Users need to understand destination neighborhoods before booking accommodations. When planning a trip to Lisbon, you want to know that Bairro Alto is nightlife, Alfama is historic and walkable, Belém has museums — before searching for hotels.
We considered Wikipedia (free but generic, not traveler-focused), Google Places ($17/1000 requests, designed for businesses not neighborhood overviews), and manual curation (doesn't scale).
Decision
Use Exa.AI's /answer endpoint for neighborhood discovery. Exa searches travel-focused web content and its LLM generates structured neighborhood information with source citations.
The flow: user triggers neighborhood discovery for a destination → backend calls Exa → response parsed into structured Neighborhood records (name, description, characteristics, best-for traveler types) → stored as destination enrichment via effect → available in the trip.
Cost
$0.005 per query. Average trip (2-3 destinations) costs ~$0.01-$0.015. At 1000 trips/month that's $10-15, reducible to ~$5 with shared city-level caching since many users visit the same cities.
Consequences
Rich, current, traveler-focused content from real travel guides and blog posts. Structured output with source citations. Works for any city worldwide without manual curation.
The trade-off is API dependency, quality variance (LLM parsing may occasionally miss details), and the need for a caching strategy to avoid re-fetching on every trip view.