backend/api/mcp_clients/mcp_client.py

import httpx
from dataclasses import dataclass, field


@dataclass
class MCPClient:
    rag_url: str
    web_url: str
    admin_url: str
    client: httpx.AsyncClient = field(default_factory=lambda: httpx.AsyncClient(timeout=30))


    async def call_rag(self, tenant_id: str, query: str, threshold: float = 0.3):
        """
        Calls the RAG search endpoint and returns the unwrapped results.
        The MCP server wraps responses in a 'data' field, so we extract it.
        
        Uses a lower threshold (0.3) by default to ensure we find relevant results
        even if semantic similarity is moderate.
        """
        r = await self.client.post(
            f"{self.rag_url}/search", 
            json={
                "tenant_id": tenant_id,
                "query": query,
                "threshold": threshold  # Lower threshold for better recall
            }
        )
        if r.status_code != 200:
            return {"results": [], "error": f"HTTP {r.status_code}"}
        data = r.json()
        # MCP server wraps response in a 'data' field
        # Extract the actual result data
        if isinstance(data, dict) and "data" in data:
            return data["data"]
        # If not wrapped, return as-is (backward compatibility)
        return data


    async def call_web(self, tenant_id: str, query: str):
        r = await self.client.post(f"{self.web_url}/search", json={"tenant_id":tenant_id,"query":query})
        return r.json()


    async def call_admin(self, tenant_id: str, query: str):
        r = await self.client.post(f"{self.admin_url}/eval", json={"tenant_id":tenant_id,"query":query})
        return r.json()