手搓Agno智能体使用Ollama模型（ollama本地智能体）

轻量化智能体开源框架Agno 有兴趣了解 What is Agno - Agno

安装python3.12

安装UV

pip install uv

创建本地目录

uv init agno_agents
cd agno_agents
# 创建虚拟环境
uv venv
uv add  agno  ollama duckduckgo-search yfinance 

agno_agents.py

# -*- coding: utf-8 -*-
from textwrap import dedent
from agno.agent import Agent 
from agno.models.ollama.chat import Ollama
from agno.tools.duckduckgo import DuckDuckGoTools
from agno.tools.yfinance import YFinanceTools

ollama_model = Ollama(id="qwq:latest",
                      show_tool_calls=True,
                      timeout=120,
                      host= "http://106.14.140.127:11434")

# Create our News Reporter with a fun personality

web_agent = Agent(
    name="Web Agent",
    role="Search the web for information",
    model=ollama_model,
    tools=[DuckDuckGoTools()],
    instructions=dedent("""\
        You are an experienced web researcher and news analyst! 

        Follow these steps when searching for information:
        1. Start with the most recent and relevant sources
        2. Cross-reference information from multiple sources
        3. Prioritize reputable news outlets and official sources
        4. Always cite your sources with links
        5. Focus on market-moving news and significant developments

        Your style guide:
        - Present information in a clear, journalistic style
        - Use bullet points for key takeaways
        - Include relevant quotes when available
        - Specify the date and time for each piece of news
        - Highlight market sentiment and industry trends
        - End with a brief analysis of the overall narrative
        - Pay special attention to regulatory news, earnings reports, and strategic announcements\
    """),
    show_tool_calls=True,
    markdown=True,
)

finance_agent = Agent(
    name="Finance Agent",
    role="Get financial data",
    model=ollama_model,
    tools=[
        YFinanceTools(stock_price=True, analyst_recommendations=True, company_info=True)
    ],
    instructions=dedent("""\
        You are a skilled financial analyst with expertise in market data! 

        Follow these steps when analyzing financial data:
        1. Start with the latest stock price, trading volume, and daily range
        2. Present detailed analyst recommendations and consensus target prices
        3. Include key metrics: P/E ratio, market cap, 52-week range
        4. Analyze trading patterns and volume trends
        5. Compare performance against relevant sector indices

        Your style guide:
        - Use tables for structured data presentation
        - Include clear headers for each data section
        - Add brief explanations for technical terms
        - Highlight notable changes with emojis ( )
        - Use bullet points for quick insights
        - Compare current values with historical averages
        - End with a data-driven financial outlook\
    """),
    show_tool_calls=True,
    markdown=True,
)

# Example usage
# web_agent.print_response(
#     "Tell me about a breaking news story happening in Times Square.", stream=True
# )

agent_team = Agent(
    team=[web_agent, finance_agent],
    model=ollama_model,
    instructions=dedent("""\
        You are the lead editor of a prestigious financial news desk! 

        Your role:
        1. Coordinate between the web researcher and financial analyst
        2. Combine their findings into a compelling narrative
        3. Ensure all information is properly sourced and verified
        4. Present a balanced view of both news and data
        5. Highlight key risks and opportunities

        Your style guide:
        - Start with an attention-grabbing headline
        - Begin with a powerful executive summary
        - Present financial data first, followed by news context
        - Use clear section breaks between different types of information
        - Include relevant charts or tables when available
        - Add 'Market Sentiment' section with current mood
        - Include a 'Key Takeaways' section at the end
        - End with 'Risk Factors' when appropriate
        - Sign off with 'Market Watch Team' and the current date\
    """),
    add_datetime_to_instructions=True,
    show_tool_calls=True,
    markdown=True,
)

agent_team.print_response(
    "Summarize analyst recommendations and share the latest news for NVDA", stream=True
)