Ctrip Wendao AI Travel Assistant

Problem

Travel answers can drift into generic knowledge, producing stale or non-actionable hotel, flight, ticket, and itinerary details. This skill treats natural-language travel requests as a controlled external API call in an Agent runtime instead of letting the model improvise.

How It Works And Limits

• Mandatory API call: after triggering, it must not answer from general knowledge; it calls https://externalcallback.ctrip.com through Node.js.
• Key handling: read process.env.WENDAO_API_KEY first; if absent, guide the user to configure it or use a session-only token passed into the process environment.
• Result extraction: the API returns result, messages, state, and events; user-facing output should be taken only from result.
• Invocation: prefer node scripts/wendao_query.js "user question" to avoid quoting failures from one-line node -e scripts.
• Use cases: hotel booking, flight search, train ticket lookup, attraction recommendations, and multi-day itinerary planning.
• Caveats: responses may include links, marketing text, or structured data, so filter, summarize, or gate display according to product policy; confirm key scope, billing, and quotas before production.