Agents

Kader provides several agent types for different reasoning patterns and use cases.

Agent Types

BaseAgent

The foundation class for all agents. Provides core functionality for tool execution, conversation management, and session persistence.

from kader.agent import BaseAgent
from kader.providers import OllamaProvider, Message

agent = BaseAgent(
    name="MyAgent",
    system_prompt="You are a helpful assistant.",
    tools=[],
    provider=OllamaProvider(model="llama3.2"),
    use_persistence=True,
    interrupt_before_tool=True,
)

# Synchronous invocation
response = agent.invoke("What is Python?")
print(response.content)

# Streaming invocation
for chunk in agent.stream("Tell me a story"):
    print(chunk.content, end="", flush=True)

# Asynchronous invocation
import asyncio
response = await agent.ainvoke("What is asyncio?")

Key Parameters

Parameter	Type	Description
name	str	Agent name
system_prompt	str	System instructions
tools	list	List of tools
provider	BaseLLMProvider	LLM provider
use_persistence	bool	Enable session saving
interrupt_before_tool	bool	Confirm before tool execution
retry_attempts	int	Number of retry attempts

ReActAgent

ReAct (Reasoning + Acting) agents use a prompt strategy that interleaves reasoning traces with actions. Better for single-step tool usage.

from kader.agent import ReActAgent
from kader.providers import OllamaProvider
from kader.tools import ReadFileTool, TodoTool

react_agent = ReActAgent(
    name="ReActBot",
    tools=[ReadFileTool(), TodoTool()],
    provider=OllamaProvider(model="llama3.2"),
    interrupt_before_tool=True,
)

response = react_agent.invoke(
    "Read the file example.py and create a todo list for the tasks mentioned"
)

PlanningAgent

Planning agents break down complex tasks into plans and execute them step by step using the TodoTool.

from kader.agent import PlanningAgent
from kader.providers import OllamaProvider
from kader.tools import TodoTool

planner = PlanningAgent(
    name="PlannerBot",
    tools=[TodoTool()],
    provider=OllamaProvider(model="llama3.2"),
    use_persistence=True,
)

response = planner.invoke(
    "Create a Python web project with: "
    "1. Setup Flask app, "
    "2. Add routes, "
    "3. Write tests"
)

The PlanningAgent will: 1. Analyze the request 2. Create a plan using TodoTool 3. Execute each step 4. Update task status as it completes

YAML Configuration

BaseAgent supports loading configuration from YAML files:

# agent.yaml
name: ConfiguredAgent
system_prompt: "You are configured via YAML."
tools:
  - read_file
  - write_file
provider:
  model: llama3.2
  provider: ollama
persistence: true
retry_attempts: 3

agent = BaseAgent.from_yaml("agent.yaml")

Choosing the Right Agent

Agent Type	Use Case
BaseAgent	Simple tasks, custom workflows
ReActAgent	Single-step tool usage, reasoning + acting
PlanningAgent	Complex multi-step tasks, project creation