Getting Started

Basic Usage

Here's a simple example to get started with Agentic-AI:

from src.core.tool_enabled_ai import ToolEnabledAI
from src.config import configure, get_config

# Configure the framework (optional - uses defaults otherwise)
# Example: Use a specific model
# configure(model="claude-3-haiku")

# Create AI instance (uses the configured model or default)
ai = ToolEnabledAI()

# Make a simple request (uses the underlying provider)
# Note: For tool usage, use process_prompt() instead of request()
response = ai.request("What is the capital of France?")
print(response)

Using Tools

Tools allow the AI to perform actions. They are typically defined in src/config/tools.yml and loaded automatically by the ToolManager.

from src.core.tool_enabled_ai import ToolEnabledAI
from src.config import configure, UseCasePreset

# Configure the framework, maybe selecting a model good at tool use
configure(
    model="claude-3-5-sonnet",
    use_case=UseCasePreset.CHAT # Or another relevant use case
)

# Create AI instance. It automatically gets a ToolManager
# which loads tools defined in tools.yml.
ai = ToolEnabledAI()

# The AI will use tools loaded by ToolManager when appropriate.
# Use process_prompt() to enable the tool-calling loop.
response = ai.process_prompt("What's the weather like in Tokyo today?")
print(f"Weather Response: {response}")

response = ai.process_prompt("What time is it now?")
print(f"Time Response: {response}")

See the Tool Integration section for more details on defining and managing tools.

Using Agents (Coordinator)

For more complex interactions involving specific workflows or routing, you often use the Coordinator agent:

from src.agents import Coordinator
from src.config import configure, UseCasePreset

# Configure the framework (e.g., model, use case)
# configure(model="claude-3-haiku", use_case=UseCasePreset.CHAT)

# Create a Coordinator instance (uses default dependencies & config)
coordinator = Coordinator()

# Prepare a request for the coordinator
request_data = {
    "prompt": "Tell me a joke and then tell me the weather in London.",
}

# Process the request through the coordinator
response_coord = coordinator.process_request(request_data)

# Print the final content from the coordinator's response
print(f"Coordinator Response: {response_coord.get('content')}")

Refer to the Agent System documentation for more on agents.