Function Calling

Function Calling

Function calling allows the model to generate structured output that calls external functions.

How It Works

1. You define functions in your request using the tools parameter
2. The model decides when to call a function based on the conversation
3. The model returns a tool_calls object with function name and arguments
4. You execute the function and return the result using role: "tool"
5. The model continues with the function result

Defining Functions

{
  "tools": [
    {
      "type": "function",
      "function": {
        "name": "get_weather",
        "description": "Get current weather for a location",
        "parameters": {
          "type": "object",
          "properties": {
            "location": {
              "type": "string",
              "description": "City name, e.g. 'London'"
            },
            "unit": {
              "type": "string",
              "enum": ["celsius", "fahrenheit"],
              "description": "Temperature unit"
            }
          },
          "required": ["location"]
        }
      }
    }
  ]
}

Example Conversation

Step 1: User asks a question

{
  "model": "gemma-4-26b",
  "messages": [
    {"role": "user", "content": "What's the weather in London?"}
  ],
  "tools": [...]
}

Step 2: Model responds with tool call

{
  "choices": [{
    "message": {
      "role": "assistant",
      "content": null,
      "tool_calls": [{
        "id": "call_abc123",
        "type": "function",
        "function": {
          "name": "get_weather",
          "arguments": "{\"location\": \"London\", \"unit\": \"celsius\"}"
        }
      }]
    },
    "finish_reason": "tool_calls"
  }]
}

Step 3: You execute the function and return the result

{
  "model": "gemma-4-26b",
  "messages": [
    {"role": "user", "content": "What's the weather in London?"},
    {"role": "assistant", "content": null, "tool_calls": [...]},
    {"role": "tool", "tool_call_id": "call_abc123", "content": "{\"temperature\": 15, \"condition\": \"cloudy\"}"}
  ],
  "tools": [...]
}

Step 4: Model responds with natural language

{
  "choices": [{
    "message": {
      "role": "assistant",
      "content": "The current weather in London is 15°C and cloudy."
    },
    "finish_reason": "stop"
  }]
}

Tool Choice

Control when the model calls functions:

Value	Behavior
"auto"	Model decides (default)
"none"	Never call functions
{"type": "function", "function": {"name": "fn"}}	Force specific function

Python Example

from openai import OpenAI

client = OpenAI(
    api_key="your-api-key",
    base_url="https://cryptgpt.co/v1"
)

tools = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Get weather for a location",
        "parameters": {
            "type": "object",
            "properties": {
                "location": {"type": "string"}
            },
            "required": ["location"]
        }
    }
}]

response = client.chat.completions.create(
    model="gemma-4-26b",
    messages=[{"role": "user", "content": "Weather in Paris?"}],
    tools=tools,
    tool_choice="auto"
)

# Check if the model wants to call a function
if response.choices[0].message.tool_calls:
    tool_call = response.choices[0].message.tool_calls[0]
    print(f"Function: {tool_call.function.name}")
    print(f"Arguments: {tool_call.function.arguments}")