import requests
import pandas as pd
from smolagents import MCPClient



def extracthackathontools():
    """Scrape all MCP tools from hackathon Spaces and return as a pandas DataFrame to allow download as a bonus feature of the MCP Explorer."""
    HF_API = "https://huggingface.co/api/spaces?author=Agents-MCP-Hackathon"

    try:
        resp = requests.get(HF_API)
        resp.raise_for_status()
    except Exception as e:
        raise RuntimeError(f"Failed to fetch Spaces list: {e}")

    spaces = resp.json()
    rows = []

    for space in spaces:
        tags = space.get("tags", [])
        if "mcp-server" not in tags:
            continue

        space_id = space["id"]  
        space_slug = space_id.replace("/", "-").replace("_", "-").lower()
        mcp_base_url = f"https://{space_slug}.hf.space/gradio_api/mcp/sse"
            

        try:
            print(mcp_base_url)
            #mcp_status = requests.get(mcp_base_url, timeout=5)
            #print(mcp_status.status_code)
            #if mcp_status.status_code != 200:
            #    print(f"[Skip] Space in error: {mcp_base_url}")
            #    continue
            
            mcp_client = MCPClient({"url": mcp_base_url,"transport": "sse"}) # Might be deprecated soon but didnt find out the clean way
            tools = mcp_client.get_tools()
            print(len(tools))
            mcp_client.disconnect()
        except Exception as e:
            print(f"[Warning] Could not fetch tools from {mcp_base_url}: {e}")
            continue

        # Infos générales du Space
        author, name = space_id.split("/")
        hf_url = f"https://huggingface.co/spaces/{space_id}"
        created_at = space.get("createdAt", "")
        n_likes = space.get("likes", 0)

        for tool in tools:
            input_fields = ", ".join(param for param in tool.inputs)
            rows.append({
                "Gradio MCP endpoint": f"{mcp_base_url}",
                "Tool name": tool.name,
                "Tool description": tool.description,
                "Tool inputs": input_fields,
                "Space name": name,
                "HF Space URL": hf_url,
                "Likes": n_likes,
                "Created at": created_at,
                "Tags": ", ".join(tags)
            })

    df = pd.DataFrame(rows)
    return df

df=extracthackathontools()

df.to_parquet("./data/hackathon_mcp_tools.parquet", index=False)