Kundensupport-Chatbot mit Deep Agents

from deepagents import create_deep_agent
from langchain_core.tools import tool

@tool
def search_faq(query: str) -> str:
    """Search the FAQ knowledge base for relevant answers."""
    results = faq_store.similarity_search(query, k=3)
    return "\n\n".join(
        f"Q: {r.metadata['question']}\nA: {r.page_content}"
        for r in results
    )

agent = create_deep_agent(
    model="anthropic:claude-sonnet-4-5-20250929",
    tools=[search_faq],
    system_prompt="You are a helpful customer support agent. Use search_faq to find answers from the knowledge base before responding. Always cite the source FAQ.",
)

result = agent.invoke({
    "messages": [("user", "How do I reset my password?")]
})
print(result["messages"][-1].content)