from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent
from langchain_openai import ChatOpenAI

# Conéctate a la base de datos unificada a través de su punto de acceso MCP nativo
async with MultiServerMCPClient({
    "unified_db": {
        "url": "http://localhost:8765/mcp",
        "transport": "streamable_http"
    }
}) as client:
    tools = client.get_tools()
    agent = create_react_agent(ChatOpenAI(model="gpt-4o"), tools)

    # En lugar de tres llamadas separadas, el agente ejecuta una sola 
    # potente consulta que combina transacciones, vectores y analíticas.
    await agent.ainvoke({"messages": [{
        "role": "user",
        "content": """
        Ejecuta una operación unificada:
        1. Inserta un nuevo ticket de soporte para 'acme-corp'.
        2. Encuentra los 3 tickets pasados más similares mediante búsqueda vectorial.
        3. Calcula el tiempo de resolución promedio de esos tickets similares.
        """
    }]})

    # Bajo el capó, la herramienta MCP ejecuta este SQL unificado:
    """
    WITH new_ticket AS (
      INSERT INTO tickets (customer_id, content, embedding)
      VALUES ('acme-corp', 'Problema de facturación...', '[0.1, 0.4, ...]')
      RETURNING id, embedding
    ),
    similar_tickets AS (
      SELECT t.id, t.resolution_time_hrs,
             vector_distance(t.embedding, (SELECT embedding FROM new_ticket)) as dist
      FROM tickets t
      WHERE t.id != (SELECT id FROM new_ticket)
      ORDER BY dist ASC
      LIMIT 3
    )
    SELECT 
      (SELECT id FROM new_ticket) as inserted_id,
      count(*) as similar_count,
      avg(resolution_time_hrs) as avg_resolution_time
    FROM similar_tickets;
    """

from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent
from langchain_openai import ChatOpenAI

# Verbinden Sie sich mit der einheitlichen Datenbank über ihren nativen MCP-Endpunkt
async with MultiServerMCPClient({
    "unified_db": {
        "url": "http://localhost:8765/mcp",
        "transport": "streamable_http"
    }
}) as client:
    tools = client.get_tools()
    agent = create_react_agent(ChatOpenAI(model="gpt-4o"), tools)

    # Anstelle von drei separaten Aufrufen führt der Agent eine einzige 
    # leistungsstarke Abfrage aus, die Transaktionen, Vektoren und Analysen kombiniert.
    await agent.ainvoke({"messages": [{
        "role": "user",
        "content": """
        Führe eine einheitliche Operation aus:
        1. Füge ein neues Support-Ticket für 'acme-corp' ein.
        2. Finde die 3 ähnlichsten vergangenen Tickets über Vektorsuche.
        3. Berechne die durchschnittliche Lösungszeit für diese ähnlichen Tickets.
        """
    }]})

    # Unter der Haube führt das MCP-Tool dieses einheitliche SQL aus:
    """
    WITH new_ticket AS (
      INSERT INTO tickets (customer_id, content, embedding)
      VALUES ('acme-corp', 'Abrechnungsproblem...', '[0.1, 0.4, ...]')
      RETURNING id, embedding
    ),
    similar_tickets AS (
      SELECT t.id, t.resolution_time_hrs,
             vector_distance(t.embedding, (SELECT embedding FROM new_ticket)) as dist
      FROM tickets t
      WHERE t.id != (SELECT id FROM new_ticket)
      ORDER BY dist ASC
      LIMIT 3
    )
    SELECT 
      (SELECT id FROM new_ticket) as inserted_id,
      count(*) as similar_count,
      avg(resolution_time_hrs) as avg_resolution_time
    FROM similar_tickets;
    """

from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent
from langchain_openai import ChatOpenAI

# ネイティブMCPエンドポイントを介して統合データベースに接続します
async with MultiServerMCPClient({
    "unified_db": {
        "url": "http://localhost:8765/mcp",
        "transport": "streamable_http"
    }
}) as client:
    tools = client.get_tools()
    agent = create_react_agent(ChatOpenAI(model="gpt-4o"), tools)

    # エージェントは3つの別々の呼び出しの代わりに、トランザクション、
    # ベクトル、分析を組み合わせた1つの強力なクエリを実行します。
    await agent.ainvoke({"messages": [{
        "role": "user",
        "content": """
        統合操作を実行してください:
        1. 'acme-corp'の新しいサポートチケットを挿入する。
        2. ベクトル検索を使用して上位3つの類似した過去のチケットを見つける。
        3. それらの類似チケットの平均解決時間を計算する。
        """
    }]})

    # 裏側では、MCPツールはこの統合SQLを実行しています：
    """
    WITH new_ticket AS (
      INSERT INTO tickets (customer_id, content, embedding)
      VALUES ('acme-corp', 'Billing issue...', '[0.1, 0.4, ...]')
      RETURNING id, embedding
    ),
    similar_tickets AS (
      SELECT t.id, t.resolution_time_hrs,
             vector_distance(t.embedding, (SELECT embedding FROM new_ticket)) as dist
      FROM tickets t
      WHERE t.id != (SELECT id FROM new_ticket)
      ORDER BY dist ASC
      LIMIT 3
    )
    SELECT 
      (SELECT id FROM new_ticket) as inserted_id,
      count(*) as similar_count,
      avg(resolution_time_hrs) as avg_resolution_time
    FROM similar_tickets;
    """