推理

本身是推理模型

from utils import new_openai_model

from agno.agent import Agent

def test():
    reasoning_agent = Agent(
        model=new_openai_model("o3-mini"),
        reasoning=True,
        markdown=True,
    )
    reasoning_agent.print_response(
        "Solve the trolley problem. Evaluate multiple ethical frameworks. "
        "Include an ASCII diagram of your solution.",
        stream=True,
        show_full_reasoning=True,
    )

使用推理工具

from utils import new_openai_model

from agno.agent import Agent
from agno.tools.thinking import ThinkingTools

def test():
    agent = Agent(
        model=new_openai_model("gpt-3.5-turbo"),
        tools=[ThinkingTools(add_instructions=True)],
        instructions="Use tables where possible",
        markdown=True,
    )

    agent.print_response(
        "请你使用think工具, 思考1+1为什么等于2, 并给出完整证明",
        stream=True,
        show_full_reasoning=True,
        stream_intermediate_steps=True
    )

思考中疑似不能调用其它工具.

推理Agent

直接启用reasoning=True

from utils import new_openai_model

from agno.agent import Agent

def test():
    agent = Agent(model=new_openai_model("gpt-3.5-turbo"), reasoning=True)

    agent.print_response("你是什么模型?", stream=True, show_full_reasoning=True)

会有两个智能体, 一个用来推理. 这个推理智能体也可以调用工具. 推理完后把结果交给原始智能体.

也可以让另一个模型来推理

from utils import new_openai_model

from agno.agent import Agent

def test():
    agent = Agent(model=new_openai_model("gpt-3.5-turbo"), reasoning_model=new_openai_model("o3-mini"))
    agent.print_response(
        "你是什么模型?",
        stream=True,
        show_full_reasoning=True,
    )

在下面的例子, gpt3.5会说自己是gpt4. 上面的会说自己是gpt3.