Tongcheng Travel AI

Problem

When a user asks for “flights from Beijing to Shanghai,” “hotels near the Bund,” or “a three-day Beijing trip,” a generic answer can drop dates, price preferences, cabin classes, hotel star ratings, or attraction filters. This skill turns Tongcheng travel queries into explicit script-based intents, so the model preserves the full request before routing it to the right interface.

How It Works

• Intent routing: flights, trains, hotels, attractions, buses, general transport, and trip packages map to flight-query.js, train-query.js, hotel-query.js, scenery-query.js, bus-query.js, traffic-query.js, and travel-query.js.
• Parameter preservation: structured fields such as --departure, --destination, and --extra are extracted explicitly; modifiers like “earliest,” “cheapest,” “high-speed rail,” “breakfast included,” or “family-friendly” are passed through --extra rather than discarded.
• Output behavior: script results, including tables, cards, and booking links, are surfaced first; missing parameters, auth failures, or empty results trigger clarification or fallback guidance.

Boundaries

This skill is for querying and booking guidance, not for offline ticket rules or live airline check-in data. If destination, date, or party size is missing, the model should ask for the missing detail. For no-result flight queries, only script-returned nearby departure results and ground-transport connections may be shown, avoiding invented prices or links.