NPT Taxi
Book now

Developers · Example workflow

Fare-quote agent demo

A minimal agent that answers taxi fare questions end-to-end. Type a pickup and dropoff (and an optional natural-language question); the model decides to call the estimate_fare tool, receives the structured result, and writes a plain-English answer.

How it works

  1. 1. Request. The browser POSTs the form to /api/agent/fare-quote.
  2. 2. Model call. The server sends the question plus the estimate_fare tool schema to Lovable AI (google/gemini-2.5-flash).
  3. 3. Tool call. The model returns a tool call. The server runs the same estimatePublicFare function that powers the public MCP tool — geocoding + Argyll & Bute tariff engine — and appends the JSON result to the conversation.
  4. 4. Final answer. The model reads the tool result and writes a plain-English answer citing the fare, distance, drive time, and any surcharges.