How to Enable Function Calling in Mistral Agents Using the Standard JSON Schema Format

In this tutorial, we’ll demonstrate how to enable function calling in Mistral Agents using the standard JSON schema format. By defining your function’s input parameters with a clear schema, you can make your custom tools seamlessly callable by the agent—enabling powerful, dynamic interactions.

We will be using the AviationStack API to retrieve real-time flight status data, showcasing how external APIs can be integrated as callable functions within a Mistral Agent.

Step 1: Setting up dependencies

Installing the Mistral library

Loading the Mistral API Key

You can get an API key from https://console.mistral.ai/api-keys

from getpass import getpass
MISTRAL_API_KEY = getpass(‘Enter Mistral API Key: ‘)

Loading the Aviation Stack API Key

You can sign up for a free API key from their dashboard to get started.

AVIATIONSTACK_API_KEY = getpass(‘Enter Aviation Stack API: ‘)

Step 2: Defining the Custom Function

Next, we define a Python function get_flight_status() that calls the AviationStack API to retrieve the real-time status of a flight. The function accepts an optional flight_iata parameter and returns key details such as airline name, flight status, departure and arrival airports, and scheduled times. If no matching flight is found, it gracefully returns an error message.

import requests
from typing import Dict
def get_flight_status(flight_iata=None):
“””
Retrieve flight status using optional filters: dep_iata, arr_iata, flight_iata.
“””
params = {
“access_key”: AVIATIONSTACK_API_KEY,
“flight_iata”: flight_iata
}

response = requests.get(“http://api.aviationstack.com/v1/flights”, params=params)
data = response.json()

if “data” in data and data[“data”]:
flight = data[“data”][0]
return {
“airline”: flight[“airline”][“name”],
“flight_iata”: flight[“flight”][“iata”],
“status”: flight[“flight_status”],
“departure_airport”: flight[“departure”][“airport”],
“arrival_airport”: flight[“arrival”][“airport”],
“scheduled_departure”: flight[“departure”][“scheduled”],
“scheduled_arrival”: flight[“arrival”][“scheduled”],
}
else:
return {“error”: “No flight found for the provided parameters.”}

Step 3: Creating the Mistral client and Agent

In this step, we create a Mistral Agent that uses tool-calling to fetch real-time flight information. The agent, named Flight Status Agent, is configured to use the “mistral-medium-2505” model and is equipped with a custom function tool named get_flight_status. This tool is defined using a JSON schema that accepts a single required parameter: the flight’s IATA code (e.g., “AI101”). Once deployed, the agent can automatically invoke this function whenever it detects a relevant user query, enabling seamless integration between natural language inputs and structured API responses.

from mistralai import Mistral
client = Mistral(MISTRAL_API_KEY)

flight_status_agent = client.beta.agents.create(
model=”mistral-medium-2505″,
description=”Provides real-time flight status using aviationstack API.”,
name=”Flight Status Agent”,
tools=[
{
“type”: “function”,
“function”: {
“name”: “get_flight_status”,
“description”: “Retrieve the current status of a flight by its IATA code (e.g. AI101).”,
“parameters”: {
“type”: “object”,
“properties”: {
“flight_iata”: {
“type”: “string”,
“description”: “IATA code of the flight (e.g. AI101)”
},
},
“required”: [“flight_iata”]
}
}
}
]
)

Step 4: Starting the Conversation and handling Function Calling

In this step, we initiate a conversation with the Flight Status Agent by asking a natural language question: “What’s the current status of AI101?”. The Mistral model detects that it should invoke the get_flight_status function and returns a function call request. We parse the arguments, run the function locally using the AviationStack API, and return the result back to the agent using FunctionResultEntry. Finally, the model incorporates the API response and generates a natural language reply with the current flight status, which we print to the console.

from mistralai import FunctionResultEntry
import json

# User starts a conversation
response = client.beta.conversations.start(
agent_id=flight_status_agent.id,
inputs=[{“role”: “user”, “content”: “What’s the current status of AI101?”}]
)

# Check if model requested a function call
if response.outputs[-1].type == “function.call” and response.outputs[-1].name == “get_flight_status”:
args = json.loads(response.outputs[-1].arguments)

# Run the function
function_result = json.dumps(get_flight_status(**args))

# Create result entry
result_entry = FunctionResultEntry(
tool_call_id=response.outputs[-1].tool_call_id,
result=function_result
)

# Return result to agent
response = client.beta.conversations.append(
conversation_id=response.conversation_id,
inputs=[result_entry]
)

print(response.outputs[-1].content)
else:
print(response.outputs[-1].content)

Check out the Notebook on GitHub. All credit for this research goes to the researchers of this project. Also, feel free to follow us on Twitter and don’t forget to join our 95k+ ML SubReddit and Subscribe to our Newsletter.

I am a Civil Engineering Graduate (2022) from Jamia Millia Islamia, New Delhi, and I have a keen interest in Data Science, especially Neural Networks and their application in various areas.

Source link