Defining functions
This request is an example of function calling. You’re telling the AI model not only what the user said but also what tools (functions) are available for it to use.
| Field | Description |
|---|
model | Specifies which AI model will process the request. |
functions | A cookbook of tools. Each function has a name, description, and parameters. Here, get_weather is defined to fetch weather details. |
parameters | Describes the function inputs. Requires location; optionally unit (celsius or fahrenheit). |
messages | Represents the conversation history. The user asked, “What is the weather like in San Francisco?” |
Here is an example function definition for a get_weather function:
{
"model": "small-1",
"functions": [
{
"name": "get_weather",
"description": "Get current weather information for a specific location",
"parameters": {
"type": "object",
"properties": {
"location": {
"type": "string",
"description": "City name or coordinates"
},
"unit": {
"type": "string",
"enum": ["celsius", "fahrenheit"],
"description": "Temperature unit"
}
},
"required": ["location"]
}
}
],
"messages": [
{
"role": "user",
"content": "What is the weather like in San Francisco?"
}
]
}
Handling Function Calls
When the model calls a function, you must execute it and return the result. Since model responses can include zero, one, or multiple calls, it is best practice to assume there are several.
Sample response with multiple function calls:
{
"result": {
"response": [
{
"content": "\nI'll get the current weather information for San Francisco for you.\n",
"role": "assistant"
},
{
"type": "function_call"
"function_call_id": "664fdcb1-f26b-4a3c-a96d-beb733f321b1",
"function_calls": [
{
"name": "get_weather"
"input": {
"location": "San Francisco"
},
}
],
}
],
"thinking": "The user is asking about the weather in San Francisco. I need to use the `get_weather` function to get this information. Looking at the function description:\n\n- `location` (required): City name or coordinates (type: string)\n- `unit` (optional): Temperature unit (type: string), allowed values: celsius, fahrenheit\n\nThe user provided \"San Francisco\" as the location, which is a city name. They didn't specify a temperature unit, and since it's optional, I don't need to ask for it. I'll just call the function with the location parameter.\n\nI need to use the `run_function` tool to write code that calls the `get_weather` function."
}
}
function_call_id is a unique identifier for a specific function call. You will use this ID when you later submit the result of that call back to the model so it can associate the output with the correct request.
function_calls contains one or more function invocations in the order the model wants them executed. If multiple functions are requested, execute them and return their results in the same order so the model can reason correctly over each output.
Quick note: The model can also perform other types of server-side actions, and will include the results from those in the response. The full response needs to be added to future messages. Tip: copy all items from the response into your messages so the model can use them in subsequent steps.
After executing the function, send the results back in a follow-up message to let the agentic model continue.
{
"model": "small-1",
"functions": [ ...same as before ],
"messages": [
{
"role": "user",
"content": "What is the weather like in San Francisco?"
},
{
"content": "\nI'll get the current weather information for San Francisco for you.\n",
"role": "assistant"
},
{
"type": "function_call"
"function_call_id": "664fdcb1-f26b-4a3c-a96d-beb733f321b1",
"function_calls": [
{
"name": "get_weather"
"input": {
"location": "San Francisco"
},
}
],
},
{
"type": "function_call_result",
"function_call_id": "664fdcb1-f26b-4a3c-a96d-beb733f321b1",
"function_call_results": [
{ "degrees": 14 }
]
}
]
}
