Quick Start
Installation Steps
Kagura Chatbot
Start Kagura AI Chatbot:
To start the Kagura chatbot, use the following command:
During the chat session, you can use the following commands:
/help: Show available commands and usage instructions./clear: Clear the current chat history./history: View the current chat history./system: Display system information and configuration settings./agents: List available agents and their descriptions./exit: Exit the chat session.
Creating an Agent
Step 1: Create Agent Directory
Create a directory for your agent configuration:
Step 2: Define Basic Agent Configuration
Create an agent.yml file in the newly created directory:
# ~/.config/kagura/agents/my_agent/agent.yml
cat > ~/.config/kagura/agents/my_agent/agent.yml <<EOF
type: atomic # atomic, tool, or composite
llm:
stream: true
skip_state_model: true # must be set to true if response_fields is not defined or there is no state_model.yml
description:
- language: en
text: This is the default agent.
- language: ja
text: デフォルトのエージェントです。
instructions:
- language: en
description: |
You will always respond in English.
- language: ja
description: |
あたなたは、必ず日本語で返答します。
prompt:
- language: en
template: |
{QUERY}
- language: ja
template: |
{QUERY}
EOF
Step 3: Use the Agent in Your Code
To interact with the agent programmatically, use the following Python code:
cat > my_chat.py <<EOF
from kagura.core.agent import Agent
async def chat():
# Assign the created agent by its directory name
agent = Agent.assigner("my_agent")
# Send a query and stream the responses
async for response in await agent.execute("Tell me about AI"):
print(response, end="")
if __name__ == "__main__":
import asyncio
asyncio.run(chat())
EOF
Step 4: Run the Python Script
Expected Output
When you run the Python script, the agent will respond to your query, for example: