TaskGen Function Usage Guide

This guide details how to define and utilize functions within TaskGen agents, enabling them to execute tasks efficiently.

Define a LLM Function

To enable your agent to interact with a large language model (LLM), define a function specifying how the agent should communicate with the LLM.

def llm(system_prompt: str, user_prompt: str) -> str:
    """
    Interacts with a specified LLM using system and user prompts to generate a response.

    :param system_prompt: A string that provides the context or instruction for the LLM.
    :param user_prompt: A string that represents the user's input or question.
    :return: A string response generated by the LLM.
    """
    from openai import ChatCompletion

    response = ChatCompletion.create(
        model="gpt-4o-mini",
        messages=[
            {"role": "system", "content": system_prompt},
            {"role": "user", "content": user_prompt}
        ]
    )
    return response.choices[0].message.content

Using Functions in Agents

Once functions are defined, they can be assigned to agents to carry out specific tasks.

1. Creating an Agent:

   Initialize an agent and assign it a function for interaction.

   from taskgen import Agent
   
   my_agent = Agent(name='HelperBot', description='A versatile helper agent.', llm=llm)
   

2. Assigning Tasks:

   Assign tasks to the agent using the defined functions, and specify the number of subtasks if necessary.

   task_output = my_agent.run(task="Schedule a meeting", num_subtasks=3)
   print(task_output)
   

3. Interacting with Tasks:

   Use the agent’s capabilities to interact dynamically with ongoing tasks.

   response = my_agent.interact(user_query="Reschedule the meeting to a later date.")
   print(response)
   

Advanced Function Usage

Define advanced behaviors and interactions by leveraging the full capabilities of TaskGen functions.

def advanced_interaction(system_prompt: str, user_input: str) -> str:
    """
    An advanced function to handle more complex interactions and provide detailed responses based on the context.

    :param system_prompt: Contextual prompt for the LLM.
    :param user_input: User's specific query or command.
    :return: Detailed response from the LLM based on the interaction.
    """
    response = llm(system_prompt, user_input)
    return response  # Modify or process the response as needed for the task

Conclusion

Utilizing these functions within your TaskGen agents allows for flexible, powerful task management and response generation tailored to specific user needs and contexts.