agentxai

A lightweight, but effective autonomous agent framework powered by X.AI's Grok model.

Features

• AgentX: An autonomous agent that can execute complex missions using provided tools
  • tools: List of callable functions the agent can use to complete its mission
    • can be another agent, assuming the agent lives in a function
  • mission: String describing the goal or task to accomplish
  • exit_function: Optional callback function executed upon mission completion
    • default is a success bool
  • system_prompt: Customizable prompt to guide agent behavior (default: autonomous goal-focused agent)
  • model: AI model selection (default: "grok-beta")
  • enable_logging: Toggle detailed logging (default: False)
  • max_steps: Maximum execution steps before termination (default: 10)
• XAI: A robust client for interacting with X.AI's Grok model and other supported AI models
• Model Selection: Choose from three AI models to power your agents
  • grok-beta: X.AI's Grok model
  • gpt-4o: OpenAI's GPT-4 optimized for performance
  • gpt-4o-mini: A lightweight version of GPT-4 optimized for minimal resource usage
• Cost tracking and token usage monitoring
• Automatic retries and error handling
• JSON response validation
• Configurable logging

Model Selection

AgentXAI supports three AI models, each with different capabilities and cost structures. You can select the desired model when initializing AgentX or XAI.

Available Models

1. grok-beta

   • Provider: X.AI
   • Cost:
     • Input Tokens: $5.00 per 1M tokens
     • Output Tokens: $15.00 per 1M tokens

2. gpt-4o

   • Provider: OpenAI
   • Cost:
     • Input Tokens: $2.50 per 1M tokens
     • Output Tokens: $10.00 per 1M tokens

3. gpt-4o-mini

   • Provider: OpenAI
   • Cost:
     • Input Tokens: $0.150 per 1M tokens
     • Output Tokens: $0.600 per 1M tokens

Selecting a Model

When initializing AgentX or XAI, specify the model parameter to select the desired AI model.

Example: Initializing AgentX with gpt-4o-mini

example agent:

from agentX.agent_x import AgentX

def my_function(input: str) -> str:
    """
    A simple example function that can be passed to AgentX as a tool.

    Args:
        input (str): The input string to process

    Returns:
        str: The unmodified input string
    """
    # do something with input
    return input

def exit_function(message: str) -> int:
    """
    A simple example exit function that can be passed to AgentX.

    Args:
        message (str): A string containing a number to convert

    Returns:
        int: The string converted to an integer
    """
    try:
        return int(message)
    except ValueError:
        return -1

agent = AgentX(
    model="grok-beta",
    tools=[my_function],
    mission="""
    Do something with

    #1 setup
     - some useful details
    #2 do something
    - some details
    #3 last thing to do
    - some details

    success criteria:
      -  that something is done
      -  it is done well
    """,
    exit_function=exit_function,
)

Installation

pip install agentxai

ENV Variables

XAI_API_KEY=