Log tool calls

How to log tool calls and function calling interactions with Keywords AI

Keywords AI allows you to log tool calls and function calling interactions for monitoring, debugging, and analytics purposes.

Tool calls enable AI models to interact with external functions and APIs. Logging these interactions helps you:

• Monitor function usage and performance
• Debug tool calling issues and errors
• Analyze user interactions with AI-powered tools
• Track costs and usage across different functions

Simple Example

Here’s a minimal example focusing just on the tool-related fields:

import requests

url = "https://api.keywordsai.co/api/request-logs/create/"
headers = {
    "Authorization": "Bearer YOUR_KEYWORDS_AI_API_KEY",
    "Content-Type": "application/json"
}

payload = {
    "model": "gpt-4",
    "prompt_messages": [
        {"role": "user", "content": "Get weather for NYC"},
        {
            "role": "assistant",
            "tool_calls": [{
                "id": "call_123",
                "type": "function",
                "function": {
                    "name": "get_weather",
                    "arguments": '{"location": "New York, NY"}'
                }
            }]
        }
    ],
    "tools": [{
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "Get current weather",
            "parameters": {
                "type": "object",
                "properties": {
                    "location": {"type": "string"}
                },
                "required": ["location"]
            }
        }
    }]
}

response = requests.post(url, headers=headers, json=payload)

Key Tool Call Parameters

Tool Calls Structure

• tool_calls: Array of function calls made by the assistant
  • id: Unique identifier for the tool call
  • type: Always “function” for function calls
  • function.name: Name of the called function
  • function.arguments: JSON string of function arguments

Tool Choice

• tool_choice: Specifies which tool the model should use
  • type: “function” for function calling
  • function.name: Name of the preferred function

Tools Definition

• tools: Array of available functions
  • type: “function” for function definitions
  • function.name: Function name
  • function.description: What the function does
  • function.parameters: JSON schema for function parameters

Best Practices

1. Include complete tool definitions in the tools array
2. Use descriptive function names and descriptions
3. Log both successful and failed tool calls
4. Include customer identification for user tracking
5. Add relevant metadata for filtering and analysis

Viewing Tool Call Logs

Tool call logs appear in your Keywords AI dashboard with:

• Function names and arguments
• Execution results and errors
• Performance metrics (latency, tokens)
• User context and metadata

Use the dashboard filters to analyze tool usage patterns and optimize your function calling implementations.

Complete Example

Here’s a full example with all available parameters:

import requests

url = "https://api.keywordsai.co/api/request-logs/create/"
headers = {
    "Authorization": "Bearer YOUR_KEYWORDS_AI_API_KEY",
    "Content-Type": "application/json"
}
payload = {
    "model": "gpt-4",
    "prompt_messages": [
        {
          "role": "user",
          "content": "What's the weather like in Boston?"
        },
        {
          "role": "assistant",
          "content": None,
          "tool_calls": [
            {
              "id": "call_abc123",
              "type": "function",
              "function": {
                "name": "get_current_weather",
                "arguments": '{"location": "Boston, MA"}'
              }
            }
          ]
        }
    ],
    "completion_message": {
        "role": "assistant",
        "content": "The current weather in Boston is 72°F and sunny."
    },
    "tool_choice": {
        "type": "function",
        "function": {
            "name": "get_current_weather"
        }
    },
    "tools": [
      {
        "type": "function",
        "function": {
          "name": "get_current_weather",
          "description": "Get the current weather in a given location",
          "parameters": {
            "type": "object",
            "properties": {
              "location": {
                "type": "string",
                "description": "The city and state, e.g. San Francisco, CA"
              },
              "unit": {
                "type": "string", 
                "enum": ["celsius", "fahrenheit"]
              }
            },
            "required": ["location"]
          }
        }
      }
    ],
    "customer_params": {
        "customer_identifier": "customer_123"
    },
    "prompt_tokens": 25,
    "completion_tokens": 12,
    "cost": 0.00037,
    "timestamp": "2024-04-15T08:30:37.721313Z"
}

response = requests.post(url, headers=headers, json=payload)