Ask a travel voice agent about weather and money

First call · You are the traveler. For the opening weather and currency warm-up, say: “I land at Heathrow. What should I pack, and how much is $200 in pounds?” Heathrow is a warm-up. You’ll use Oslo–Bergen when you connect the flight boards to the forecast.

1. Meet the tools before the model

In a new workshop terminal, run:

bash desk data weather LHR
bash desk data fx USD GBP 200

Read the airport, forecast hour, temperature in Celsius and rain probability. Read the base and quote currencies, rate, converted amount and reference date. A forecast for the current hour helps with “what’s it like now?” To answer “when I land,” the tool needs the arrival time. A daily reference rate is useful for budgeting; the card issuer may charge a different rate.

Try a flight before minute 30. Ask for the remaining OSL → BGO schedule now:

bash desk data flights OSL BGO

Choose a flight number from the returned board and use bash desk data status FLIGHT_NUMBER with that actual number. Compare the flight’s estimated and scheduled arrival. If the provider returns no matches, keep the coverage note in your answer. The board needs no key.

Make one small change

Run bash desk data fx USD GBP 100 and predict the converted amount using the rate you just retrieved for $200. Compare your calculation with the tool’s answer, allowing for rounding. Which field would make you reconsider your prediction if the two calls crossed a rate update?

2. Build a working starting point

After the creation log says Travel factory dependencies ready, run:

bash desk doctor
bash desk starter
bash desk inspect builds/starter

The first command reports presence without printing values. The second creates a Mantle project from the supplied two-tool starter. The third trains it and starts Inspector. Leave this terminal running. If the starter already exists, skip its creation and inspect that folder again.

3. Ask a real question

Keep the server running in Terminal 1. Choose Terminal → New Terminal and run bash desk url. Open the printed HTTPS link in its own tab. This opens the private Rasa Inspector. If GitHub shows Continue, continue with the same signed-in GitHub account.

1. Connect your headset or turn on your microphone and speakers. In Inspector find the waveform button labeled Start voice conversation. Click it.
2. When your browser asks to use the microphone, choose Allow for this Inspector site. Speak into your own device: “What is the weather at London Heathrow right now?” Pause and listen for a spoken answer.
3. Stay in the same call. Ask: “And how much is two hundred US dollars in British pounds?” Listen for the amount and rate date. This second turn checks that the agent is still listening after its first reply.
4. Open Inspect and compare the recognized questions and the destination_weather and convert_currency tool results with what you heard. End the call with Stop voice conversation before closing the tab.

Minute 30 checkpoint. Explain which tool supplied each number. Show one tool result alongside what the traveler saw. If only the command-line tools ran, record that separately. The next chapter adds the flight information missing from this first agent.