Question

You are building a travel-planning application on Azure AI Foundry that uses an LLM with function calling. For a single user request such as 'Compare weather and flight prices for Rome and Athens next week,' you want the model to invoke your get_weather and get_flight_price functions for both cities as efficiently as possible before composing a final answer. Which approach best supports this multistep, tool-augmented reasoning while minimizing round-trip latency?

Accepted Answer

Enable parallel tool calling so the model can emit multiple function calls in one turn, then return all tool results together in a single follow-up message before the model synthesizes the answer.

Answer

Set the temperature to 0 so the model deterministically calls each function one at a time in strict sequence.

Answer

Force tool_choice to a single required function so the model only ever calls one tool per request.

Answer

Disable function calling and instruct the model in the system prompt to fabricate reasonable weather and price estimates from its training data.