Skip to content

AI Agents

AI Agents are intelligent controllers that connect different tasks, process user input, and generate meaningful responses. They act like a human assistant.

They can operate in different environments, such as a command-line console, a web browser, a virtual gaming world or simulation, or even a real physical world like robots and autonomous vehicles.

Arkalos includes built-in AI Agents like DWHAgent, which allows users to interact with their data using natural language.

Let's create our own AI Agent.

Creating an AI Agent

To create an agent, implement the AIAgent contract and define the following methods:

  • NAME: A unique name for the agent.
  • DESCRIPTION: A brief explanation of what the agent does.
  • GREETING: Initial greeting message from the agent to the user.
  • ACTIONS: List of actions an agent can take.
  • processMessage(message): The method that processes user input and runs actions.

Note

The AI Agent's processMessage(message) method expects a markdown-formatted string and returns a response as plain text or markdown. Using markdown enables frontend formatting for user-agent communication.

Example: Simple Calculator Agent

Create a new file: app/ai/agents/my_agent.py

from arkalos.ai import AIAgent
from app.ai.tasks.calc_action import CalcAction

class MyAgent(AIAgent):

    NAME = 'MyAgent'

    DESCRIPTION = 'A calculator agent.'

    GREETING = 'Hi, I am a calculator. What do you want to calculate?'

    ACTIONS = [
        CalcAction
    ]

    def processMessage(self, message):
        output = self.runAction(CalcAction, message)
        return output

Running the Agent

Create a script and run the agent:

scripts/ai/my_agent.py
from app.ai.agents.my_agent import MyAgent

agent = MyAgent()
agent.runConsole()

Run the script:

uv run scripts/ai/my_agent.py

Multi-Action AI Agent

A multi-task agent can determine user intent and execute the appropriate action.

Arkalos includes a whichAction() method that uses AI to identify the correct task from the list of registered tasks.

Example: Multi-Task AI Agent

from arkalos.ai import AIAgent
from app.ai.tasks.calc_action import CalcAction
from app.ai.tasks.what_is_my_ip_action import WhatIsMyIpAction

class MultiAgent(AIAgent):

    NAME = 'MultiAgent'

    DESCRIPTION = 'An Agent that understands the intent, determines which task to perform and runs it.'

    GREETING = 'Hi, I am a MultiAgent. I can tell your IP address, do basic math calculations or transform text to SQL.'

    ACTIONS = [
        WhatIsMyIpAction, 
        CalcAction, 
        TextToSQLAction
    ]

    def processMessage(self, message):
        response = f"Determining the intent and which task to run...\n"
        which_action = self.whichAction(message)
        response += f"Based on your question, I determined this task: {which_action}\n"
        response += f"Running this task...\n"
        output = self.runAction(which_action, message)
        response += f"Task output: {output}\n"
        return response

Here, whichAction() determines the right task based on user input.

Now, you can create AI Agents that handle multiple tasks efficiently!