AgentPlanck

AgentPlanck is a a lightweight production ready agentic framework to build your agents.

✨ Features

• Dynamic Pydantic model generation for structured outputs.
• Message managers for clean communication flow.
• MCP handlers for flexible tool interaction.
• Fault tolerance mechanisms to keep agents running reliably.
• And more to explore! 🚀

🔬 Why "AgentPlanck"?

The name comes from Max Planck and the Planck constant, which defines the smallest "action" in physics.
It’s fitting for a system that takes intelligent actions at minimal scales.

⚡ Quick Start

import asyncio
from agentplanck import Agent, ToolsController, ToolResult
from pydantic import BaseModel
from langchain_openai import ChatOpenAI
from dotenv import load_dotenv
load_dotenv()

llm = ChatOpenAI(model="gpt-4o")

class CheckWeather(BaseModel):
    city: str

class ExecDb(BaseModel):
    query: str

async def main():
    tools_controller = ToolsController()

    @tools_controller.tool("Check weather", param_model=CheckWeather)
    async def check_weather(params: CheckWeather) -> ToolResult:
        return ToolResult(content="Weather is hot, about 35°C")

    @tools_controller.tool("Interact with database", param_model=ExecDb)
    async def sql(params: ExecDb) -> ToolResult:
        return ToolResult(content="Successfully executed query")
    
    agent = Agent(
        "Check weather of San Francisco and insert to my db", 
        llm=llm, 
        tools_controller=tools_controller
    )
    res = await agent.run()
    print(res)

if __name__ == "__main__":
    asyncio.run(main())

📜 License

MIT License. Free to use and modify.