​
Function Calling

Function calling (also known as tool calling) allows AI models to interact with external tools and APIs, enabling them to perform actions beyond text generation. This powerful feature lets you build AI applications that can fetch real-time data, perform calculations, and integrate with your existing systems.

Model Performance: All chat models support function calling, but Qwen 2.5 72B offers exceptional tool calling capabilities and provides the most reliable results. While other models can handle function calling, Qwen is specifically optimized for complex tool workflows.

​
Basic Example

Here’s a simple example of how to implement function calling with a weather API:

from tinfoil import TinfoilAI
import json

# Initialize client with Qwen 2.5 72B (recommended for tool calling)
client = TinfoilAI(
    enclave="qwen2-5-72b.model.tinfoil.sh",
    repo="tinfoilsh/confidential-qwen2-5-72b",
    api_key="<YOUR_API_KEY>"
)

# Define the tool/function
tools = [
    {
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "Get current weather for a specific location",
            "parameters": {
                "type": "object",
                "properties": {
                    "location": {
                        "type": "string",
                        "description": "The city and state, e.g. San Francisco, CA"
                    }
                },
                "required": ["location"]
            }
        }
    }
]

# Mock weather function (replace with real API call)
def get_weather(location):
    return f"The weather in {location} is sunny, 22°C"

# Make the initial request
response = client.chat.completions.create(
    model="qwen2-5-72b",
    messages=[
        {"role": "user", "content": "What's the weather like in New York?"}
    ],
    tools=tools
)

# Check if the model wants to call a function
message = response.choices[0].message
if message.tool_calls:
    # Process each tool call
    for tool_call in message.tool_calls:
        if tool_call.function.name == "get_weather":
            # Parse function arguments
            args = json.loads(tool_call.function.arguments)
            location = args["location"]
            
            # Call the function
            weather_result = get_weather(location)
            
            # Send the function result back to the model
            messages = [
                {"role": "user", "content": "What's the weather like in New York?"},
                message,  # Assistant's message with tool call
                {
                    "role": "tool",
                    "content": weather_result,
                    "tool_call_id": tool_call.id
                }
            ]
            
            # Get the final response
            final_response = client.chat.completions.create(
                model="qwen2-5-72b",
                messages=messages,
                tools=tools
            )
            
            print(final_response.choices[0].message.content)
else:
    print(message.content)

​
Multiple Tools Example

You can define multiple tools for more complex workflows:

tools = [
    {
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "Get current weather for a location",
            "parameters": {
                "type": "object",
                "properties": {
                    "location": {"type": "string"}
                },
                "required": ["location"]
            }
        }
    },
    {
        "type": "function",
        "function": {
            "name": "calculate",
            "description": "Perform mathematical calculations",
            "parameters": {
                "type": "object",
                "properties": {
                    "expression": {
                        "type": "string",
                        "description": "Mathematical expression to evaluate"
                    }
                },
                "required": ["expression"]
            }
        }
    }
]

def calculate(expression):
    # Safe evaluation of mathematical expressions
    try:
        result = eval(expression)
        return str(result)
    except:
        return "Error: Invalid mathematical expression"

# The model can now choose between weather and calculation functions
response = client.chat.completions.create(
    model="qwen2-5-72b",
    messages=[
        {"role": "user", "content": "What's 15 * 23 + 45?"}
    ],
    tools=tools
)

​
Best Practices

  1. Choose the Right Model: Qwen 2.5 72B provides the most reliable function calling capabilities
  2. Clear Descriptions: Write detailed function descriptions to help the model understand when to use each tool
  3. Parameter Validation: Always validate function parameters before execution
  4. Error Handling: Implement proper error handling for function calls
  5. Security: Never execute untrusted code - validate all function arguments
  6. Testing: Test your functions independently before integrating with the AI model

​
Common Use Cases

  • API Integration: Fetch real-time data from external APIs
  • Database Queries: Retrieve information from your databases
  • Calculations: Perform complex mathematical operations
  • File Operations: Read, write, or process files
  • System Commands: Execute system operations (with proper security measures)
  • Third-party Services: Integrate with external services and platforms

​
Model Comparison

ModelFunction Calling SupportReliabilityBest For
Qwen 2.5 72B✅ ExcellentVery HighComplex tool workflows, multiple functions
Mistral Small 3.1 24B✅ GoodHighSimple to moderate tool calling
DeepSeek R1 70B✅ BasicModerateSimple function calls, reasoning tasks
Llama 3.3 70B✅ BasicModerateSimple function calls, conversational AI

Model Catalog

View all available models and their capabilities.

Python SDK

Complete Python SDK documentation with more examples.