Pipeline RAG con CrewAI

from crewai import Agent, Crew, Task
from langchain_openai import ChatOpenAI
from langchain_core.tools import tool

@tool
def search_documents(query: str, top_k: int = 5) -> str:
    """Search uploaded documents for relevant content."""
    results = vector_store.similarity_search(query, k=top_k)
    return "\n\n".join(
        f"[{i+1}] {r.metadata.get('source', 'unknown')}: {r.page_content}"
        for i, r in enumerate(results)
    )

agent = Agent(
    role="Specialist",
    goal="You are a document Q&A assistant. Search documents before answering. Cite sources using [1], [2] format.",
    tools=[search_documents],
    llm=ChatOpenAI(model="gpt-4o"),
)

task = Task(
    description="What does the refund policy say about digital products?",
    expected_output="Detailed response",
    agent=agent,
)

crew = Crew(agents=[agent], tasks=[task])
result = crew.kickoff()
print(result.raw)