from agno.agent import Agent
from agno.models.anthropic import Claude
from agno.tools.reasoning import ReasoningTools
from agno.tools.yfinance import YFinanceTools

# 創(chuàng)建具有推理能力的Agent
reasoning_agent = Agent(
    model=Claude(id="claude-3-7-sonnet-latest"),
    tools=[
        ReasoningTools(
            think=True,           # 啟用思考工具
            analyze=True,         # 啟用分析工具
            add_instructions=True,  # 添加默認(rèn)推理指令
            add_few_shot=True      # 添加少樣本示例
        ),
        YFinanceTools(stock_price=True, company_info=True)
    ],
    instructions=[
        "Break down complex problems into component parts",
        "Clearly state your assumptions",
        "Develop a structured reasoning path",
        "Consider multiple perspectives",
        "Evaluate evidence and counter-arguments",
        "Draw well-justified conclusions"
    ],
    show_tool_calls=True,  # 顯示推理過程
    show_full_reasoning=True,  # 顯示完整推理鏈
    stream_intermediate_steps=True# 流式輸出中間步驟
)

# 模型接入示例
from agno.models.openai import OpenAIChat
from agno.models.anthropic import Claude
from agno.models.groq import Groq

# 靈活切換不同的模型提供商
agent = Agent(
    model=Claude(id="claude-3-7-sonnet-latest"),  # 可隨時切換
    # model=OpenAIChat(id="gpt-4o"),
    # model=Groq(id="llama-3.3-70b-versatile"),
)

from agno.tools.duckduckgo import DuckDuckGoTools
from agno.tools.yfinance import YFinanceTools
from agno.tools.github import GithubTools

agent = Agent(
    tools=[
        DuckDuckGoTools(),          # 網(wǎng)絡(luò)搜索能力
        YFinanceTools(              # 金融數(shù)據(jù)訪問
            stock_price=True,
            analyst_recommendations=True
        ),
        GithubTools(search_repositories=True)  # 代碼倉庫搜索
    ]
)

from agno.agent import Agent
from agno.models.openai import OpenAIChat
from agno.tools.duckduckgo import DuckDuckGoTools
from agno.tools.yfinance import YFinanceTools
from textwrap import dedent

# 創(chuàng)建具有復(fù)雜指令系統(tǒng)的高級Agent
advanced_agent = Agent(
    model=OpenAIChat(id="gpt-4o"),
    name="Market Intelligence Agent",
    role="Financial Market Analyst and Research Expert",
    
    # 全局指令 - 定義Agent的核心行為準(zhǔn)則
    description=dedent("""
        You are an elite financial analyst with expertise in market research,
        technical analysis, and investment strategy. Your communication style
        is professional yet accessible, always backing claims with data.
    """),
    
    # 多層級的任務(wù)指令
    instructions=[
        # === 通信風(fēng)格指令 ===
        "Always maintain a professional but approachable tone",
        "Use industry terminology appropriately, but explain complex concepts clearly",
        "Be confident in your analysis but acknowledge uncertainties",
        
        # === 數(shù)據(jù)處理指令 ===
        "When presenting numerical data:",
        "  - Use tables for comparing multiple metrics",
        "  - Round percentages to 2 decimal places",
        "  - Always include data sources and timestamps",
        "  - Highlight significant changes (>5%) in bold",
        
        # === 分析方法指令 ===
        "For market analysis requests:",
        "  1. Start with macro economic context",
        "  2. Analyze sector-specific trends",
        "  3. Examine company fundamentals",
        "  4. Consider technical indicators",
        "  5. Provide risk assessment",
        "  6. End with actionable recommendations",
        
        # === 工具使用策略 ===
        "Tool usage priorities:",
        "  - For real-time prices: Use YFinanceTools first",
        "  - For news and context: Use DuckDuckGoTools",
        "  - For historical data: Combine both tools",
        "  - Always verify data from multiple sources when possible",
        
        # === 錯誤處理指令 ===
        "When encountering data issues:",
        "  - Explicitly state if data is unavailable",
        "  - Provide alternative data sources",
        "  - Suggest manual verification steps",
        "  - Never make up or estimate critical figures",
        
        # === 合規(guī)和免責(zé)聲明 ===
        "Include appropriate disclaimers:",
        "  - 'This is not financial advice'",
        "  - 'Past performance does not guarantee future results'",
        "  - 'Please consult with a qualified financial advisor'",
        
        # === 響應(yīng)結(jié)構(gòu)指令 ===
        "Structure your responses as follows:",
        "  1. Executive Summary (2-3 sentences)",
        "  2. Detailed Analysis (with subsections)",
        "  3. Key Metrics Table",
        "  4. Risk Factors",
        "  5. Recommendations",
        "  6. Disclaimers",
        
        # === 個性化指令 ===
        "Adapt communication based on user expertise:",
        "  - For beginners: Use more explanations and examples",
        "  - For professionals: Focus on technical details",
        "  - For executives: Emphasize strategic implications"
    ],
    
    # 工具配置
    tools=[
        YFinanceTools(
            stock_price=True,
            analyst_recommendations=True,
            company_info=True,
            company_news=True,
            stock_fundamentals=True
        ),
        DuckDuckGoTools()
    ],
    
    # 上下文相關(guān)配置
    add_datetime_to_instructions=True,  # 自動添加時間上下文
    add_history_to_messages=True,       # 包含歷史對話
    num_history_runs=5,                 # 包含最近5輪對話
    
    # 調(diào)試和監(jiān)控
    show_tool_calls=True,
    markdown=True,
    debug_mode=False
)

# 使用示例 - Agent會根據(jù)指令層級處理請求
response = advanced_agent.run(
    "Analyze NVDA's investment potential for Q1 2025",
    user_context={
        "expertise_level": "professional",
        "investment_horizon": "6-12 months",
        "risk_tolerance": "moderate"
    }
)

# 根據(jù)市場條件動態(tài)調(diào)整指令
if market_volatility > 0.3:
    agent.update_instructions([
        "Emphasize risk management strategies",
        "Include volatility analysis in all recommendations",
        "Suggest hedging strategies when appropriate"
    ])

# 根據(jù)用戶偏好個性化
if user_preferences.get("technical_analysis"):
    agent.add_instructions([
        "Include RSI, MACD, and Bollinger Bands analysis",
        "Provide chart pattern recognition insights"
    ])

from agno.storage.sqlite import SqliteStorage

agent = Agent(
    storage=SqliteStorage(
        table_name="agent_sessions",
        db_file="sessions.db"
    ),
    add_history_to_messages=True,
    num_history_runs=5  # 包含最近5輪對話
)

from agno.memory.v2.memory import Memory
from agno.memory.v2.db.sqlite import SqliteMemoryDb

agent = Agent(
    memory=Memory(
        db=SqliteMemoryDb(db_file="memories.db"),
        enable_user_memories=True
    )
)

from agno.knowledge import AgentKnowledge
from agno.vectordb.pgvector import PgVector

agent = Agent(
    knowledge=AgentKnowledge(
        vector_db=PgVector(
            search_type="hybrid",  # 混合搜索
            collection="domain_knowledge"
        )
    )
)

# 完整的初始化示例
agent = Agent(
    # 模型配置
    model=OpenAIChat(id="gpt-4o"),
    
    # 身份配置
    name="DataAnalyst",
    role="Analyze and visualize data",
    
    # 工具配置
    tools=[...],
    
    # 內(nèi)存配置
    storage=SqliteStorage(...),
    memory=Memory(...),
    
    # 知識配置
    knowledge=AgentKnowledge(...),
    
    # 行為配置
    instructions=[...],
    
    # 運(yùn)行配置
    show_tool_calls=True,
    markdown=True
)

agent.print_response(
    "Analyze the market",
    stream=True,
    stream_intermediate_steps=True
)

from pydantic import BaseModel

class AnalysisResult(BaseModel):
    summary: str
    metrics: dict
    recommendations: list

agent = Agent(
    response_model=AnalysisResult,
    structured_output=True
)

# 僅需模型即可創(chuàng)建功能完整的Agent
agent = Agent(
    model=OpenAIChat(id="gpt-4o")
)

# 動態(tài)添加工具
agent.add_tool(NewTool())

# 更新指令
agent.update_instructions([
    "Focus on technical analysis",
    "Include risk assessment"
])

# 切換模型
agent.set_model(Claude(id="claude-3-opus"))

# 基礎(chǔ)配置
base_config = {
    "model": OpenAIChat(id="gpt-4o"),
    "instructions": ["Be helpful"]
}

# 專門化配置
specialized_agent = Agent(
    **base_config,
    role="Financial Analyst",
    tools=[YFinanceTools()],
    instructions=base_config["instructions"] + ["Focus on financial metrics"]
)

import os
from dotenv import load_dotenv

load_dotenv()

agent = Agent(
    model=OpenAIChat(
        api_key=os.getenv("OPENAI_API_KEY"),
        id=os.getenv("MODEL_ID", "gpt-4o")  # 帶默認(rèn)值
    )
)

# 上下文組成
context = {
    "user_message": current_input,
    "chat_history": last_n_messages,
    "user_memories": relevant_memories,
    "knowledge": retrieved_knowledge,
    "session_state": current_state,
    "tools_available": registered_tools
}

# 高級調(diào)用配置
agent = Agent(
    model=OpenAIChat(
        id="gpt-4o",
        temperature=0.7,
        max_tokens=2000,
        retry_count=3,
        timeout=30
    )
)

# 流式輸出配置
for chunk in agent.stream_response(query):
    print(chunk, end="", flush=True)
    
# 帶中間步驟的流式輸出
agent.print_response(
    query,
    stream=True,
    show_full_reasoning=True,
    stream_intermediate_steps=True
)

# Team協(xié)作示例
team = Agent(
    team=[web_agent, finance_agent, analyst_agent],
    model=OpenAIChat(id="gpt-4o"),
    instructions=["Coordinate team efforts", "Synthesize findings"]
)

agent = Agent(
    model=OpenAIChat(id="gpt-4o"),
    show_tool_calls=True,      # 顯示工具調(diào)用
    debug_mode=True,           # 啟用調(diào)試模式
    log_level="DEBUG"          # 詳細(xì)日志
)

# 監(jiān)控會話
session_metrics = agent.get_session_metrics()
performance_stats = agent.get_performance_stats()