Tencent Mobility Ride Assistant

Problem Context

Ride-hailing requests look simple, but integrating them with an LLM often fails in two ways: the model answers addresses, prices, and order states from general knowledge, which may not match live service data; or first-time use is missing environment setup such as a token and resident city, causing later steps to break.

How It Works

This skill targets Tencent ride-hailing workflows and maps user intent to fixed processes, prioritizing tool calls for factual data. Common entry points include requesting a ride (“from A to B”), quick ride (“go home”, “go to work”), order lookup, driver location, cancellation, and token management. The flow starts with a preflight environment check. If it is not ready, the skill enters Quick Start to complete setup_token and setup_resident_city. Address, price, and order information must come from tools rather than model estimation. Responses also preserve the required template structure to avoid missing fields or formatting drift.

Boundaries and Notes

The skill fits authorized environments for immediate ride requests and live order-status checks. Contacting or urging the driver, proxy ride requests, airport transfers, invoices, carpooling, and ride-sharing are outside the current scope and should be directed to the corresponding Tencent service pages. In use, passenger capacity should be expressed as vehicle seats minus one; do not ask users to refine an address before the address tool is called; and if preflight returns non-zero, complete setup before continuing.