CrewAI — Multi-Agent Collaboration

from crewai import Agent, Task, Crew, Process
from crewai_tools import SerperDevTool, ScrapeWebsiteTool
from langchain_anthropic import ChatAnthropic

# Initialize the LLM (CrewAI works with any LangChain-compatible LLM)
llm = ChatAnthropic(model="claude-sonnet-4-6", temperature=0)

# Search and scrape tools
search_tool = SerperDevTool()
scrape_tool = ScrapeWebsiteTool()

# -------------------------------------------------------
# AGENT 1: The Researcher
# -------------------------------------------------------
researcher = Agent(
    role="Senior Research Analyst",
    goal=(
        "Find comprehensive, accurate, and up-to-date information on the given topic. "
        "Focus on primary sources, statistics, and expert opinions."
    ),
    backstory=(
        "You are a veteran research analyst with 15 years of experience in tech journalism. "
        "You are known for your thorough fact-checking and ability to synthesize complex topics. "
        "You always cite sources and flag uncertain information."
    ),
    tools=[search_tool, scrape_tool],
    llm=llm,
    verbose=True,
    allow_delegation=False,  # this agent does its own work
)

# -------------------------------------------------------
# AGENT 2: The Writer
# -------------------------------------------------------
writer = Agent(
    role="Technical Content Writer",
    goal=(
        "Transform research findings into clear, engaging, well-structured articles "
        "that are accessible to a technical audience."
    ),
    backstory=(
        "You are a technical writer who has authored documentation for major open-source projects. "
        "You excel at structuring complex information hierarchically and making it scannable. "
        "You always write in an active voice with concrete examples."
    ),
    tools=[],  # the writer works from research, no external tools needed
    llm=llm,
    verbose=True,
    allow_delegation=False,
)

# -------------------------------------------------------
# AGENT 3: The Editor
# -------------------------------------------------------
editor = Agent(
    role="Senior Editor",
    goal=(
        "Review and refine the draft article for accuracy, clarity, structure, "
        "and consistency. Catch factual errors and improve readability."
    ),
    backstory=(
        "You have edited publications for The Verge and Wired. "
        "Your editorial instincts are sharp: you immediately spot vague claims, "
        "logical gaps, and jargon that needs explaining."
    ),
    tools=[search_tool],  # can verify facts if needed
    llm=llm,
    verbose=True,
    allow_delegation=True,  # can reassign to researcher if facts need checking
)

# -------------------------------------------------------
# TASK 1: Research
# -------------------------------------------------------
research_task = Task(
    description=(
        "Research the current state of agentic AI frameworks in 2024. "
        "Cover: (1) the top 5 frameworks and their key features, "
        "(2) adoption trends, (3) production use cases, "
        "(4) technical limitations and how teams are working around them. "
        "Produce a structured research document with citations."
    ),
    expected_output=(
        "A detailed research report (800-1000 words) with sections for each framework, "
        "adoption data with sources, and a summary of key trade-offs."
    ),
    agent=researcher,
)

# -------------------------------------------------------
# TASK 2: Write (depends on research_task)
# -------------------------------------------------------
write_task = Task(
    description=(
        "Using the research report, write a comprehensive article titled "
        "'The State of Agentic AI Frameworks in 2024'. "
        "The article should: introduce the topic for a developer audience, "
        "compare the frameworks with a feature matrix, discuss real use cases, "
        "and end with recommendations for choosing a framework."
    ),
    expected_output=(
        "A polished article of 1200-1500 words with a clear structure: "
        "intro, framework comparison, use cases, recommendations, and conclusion."
    ),
    agent=writer,
    context=[research_task],  # writer receives researcher's output
)

# -------------------------------------------------------
# TASK 3: Edit (depends on write_task)
# -------------------------------------------------------
edit_task = Task(
    description=(
        "Review the article draft. Check for: factual accuracy, "
        "clarity of explanations, logical flow, and appropriate tone. "
        "Make improvements directly in the text and add an editor's note "
        "summarizing what was changed and why."
    ),
    expected_output=(
        "The final, polished article with all improvements applied, "
        "followed by a brief editor's note explaining key changes."
    ),
    agent=editor,
    context=[write_task],
)

# -------------------------------------------------------
# CREW: Assemble and run
# -------------------------------------------------------
crew = Crew(
    agents=[researcher, writer, editor],
    tasks=[research_task, write_task, edit_task],
    process=Process.sequential,  # tasks run in order
    verbose=True,
)

# Kick off the crew
result = crew.kickoff()
print(result)

from crewai_tools import BaseTool
from pydantic import BaseModel, Field
import requests

class StockPriceInput(BaseModel):
    ticker: str = Field(description="Stock ticker symbol e.g. AAPL")

class StockPriceTool(BaseTool):
    name: str = "get_stock_price"
    description: str = (
        "Get the current stock price and daily change for a given ticker symbol. "
        "Use when you need real-time market data."
    )
    args_schema: type[BaseModel] = StockPriceInput

    def _run(self, ticker: str) -> str:
        # Replace with a real API like Yahoo Finance or Alpha Vantage
        response = requests.get(
            f"https://query1.finance.yahoo.com/v8/finance/chart/{ticker}",
            headers={"User-Agent": "Mozilla/5.0"}
        )
        data = response.json()
        price = data["chart"]["result"][0]["meta"]["regularMarketPrice"]
        prev_close = data["chart"]["result"][0]["meta"]["chartPreviousClose"]
        change_pct = ((price - prev_close) / prev_close) * 100
        return f"{ticker}: ${price:.2f} ({change_pct:+.2f}%)"

# Attach to an agent
market_analyst = Agent(
    role="Market Analyst",
    goal="Analyze stock market data and provide investment insights",
    backstory="You are a CFA with 20 years of equity research experience.",
    tools=[StockPriceTool()],
    llm=llm,
)