jxta/jupyter-sage-mcp

git clone https://github.com/jxta/jupyter-sage-mcp.git
cd jupyter-sage-mcp
conda env create -f environment.yml
conda activate jupyter-mcp-sage
chmod +x postBuild
./postBuild
jupyter lab

conda env create -f environment.yml
conda activate jupyter-mcp-sage

# SageMath
conda install -c conda-forge sagemath

# Jupyter環境
pip install jupyterlab==4.4.1 jupyter-collaboration==4.0.2
pip install datalayer_pycrdt==0.12.17

# MCP関連
pip install websockets requests nest-asyncio

# SageMathカーネル登録
sage -python -m ipykernel install --user --name sagemath --display-name "SageMath"

# 確認
jupyter kernelspec list

{
  "mcpServers": {
    "jupyter-sage": {
      "command": "python",
      "args": ["path/to/src/mcp_sage_helper.py"],
      "env": {
        "JUPYTER_URL": "http://localhost:8888",
        "JUPYTER_TOKEN": "your_token"
      }
    }
  }
}

# MCPクライアントのセットアップ
from src.mcp_sage_helper import setup_sage_mcp
mcp = setup_sage_mcp()

# SageMathコードの実行
sage_result = mcp.execute_code("""
var('x y')
f = x^2 + y^2
solve(f == 25, x)
""", kernel_type="sagemath")

print(sage_result)

# 現在のカーネル確認
print("Current kernel:", mcp.current_kernel)

# SageMathに切り替え
mcp.switch_kernel_context("sagemath")

# Pythonに切り替え
mcp.switch_kernel_context("python")

# SageMathで数学計算
sage_result = mcp.execute_code("""
# 楕円曲線暗号
p = 23
E = EllipticCurve(GF(p), [1, 0])
print(f"Elliptic curve points: {E.order()}")
""", kernel_type="sagemath")

# Pythonでデータ可視化
python_result = mcp.execute_code("""
import matplotlib.pyplot as plt
import numpy as np

x = np.linspace(0, 2*np.pi, 100)
plt.plot(x, np.sin(x))
plt.title('Sine Wave')
plt.show()
""", kernel_type="python")

# 数論的関数の解析
sage_code = """
# 完全数の研究
def is_perfect(n):
    return sum(divisors(n)[:-1]) == n

perfects = [n for n in range(2, 10000) if is_perfect(n)]
print("Perfect numbers up to 10000:", perfects)

# メルセンヌ素数
mersenne_primes = []
for p in primes(20):
    mp = 2^p - 1
    if mp.is_prime():
        mersenne_primes.append((p, mp))
        
print("Mersenne primes:", mersenne_primes)
"""

# 統計分析とMCPの組み合わせ
analysis_code = """
import pandas as pd
import numpy as np
from scipy import stats

# データ生成
np.random.seed(42)
data = {
    'treatment': np.random.choice(['A', 'B'], 1000),
    'outcome': np.random.normal(50, 10, 1000)
}
df = pd.DataFrame(data)

# A/Bテスト
group_a = df[df['treatment'] == 'A']['outcome']
group_b = df[df['treatment'] == 'B']['outcome']

t_stat, p_value = stats.ttest_ind(group_a, group_b)
print(f"t-statistic: {t_stat:.4f}")
print(f"p-value: {p_value:.4f}")
"""

# 楕円曲線暗号の実装
crypto_code = """
# ECDH鍵交換のシミュレーション
p = 2^255 - 19  # Curve25519のp
F = GF(p)
E = EllipticCurve(F, [-1, 0])

# Alice の秘密鍵と公開鍵
alice_private = randint(1, E.order()-1)
alice_public = alice_private * E.random_point()

# Bob の秘密鍵と公開鍵  
bob_private = randint(1, E.order()-1)
bob_public = bob_private * E.random_point()

# 共有秘密の計算
shared_secret_alice = alice_private * bob_public
shared_secret_bob = bob_private * alice_public

print("ECDH key exchange successful:", shared_secret_alice == shared_secret_bob)
"""

import asyncio

async def connect_external():
    # 外部MCPサーバーへの接続
    connection = await mcp.connect_external_mcp(
        "wss://math-mcp-server.example.com/ws",
        auth_token="your_token_here"
    )
    
    if connection["status"] == "success":
        print("✅ External MCP connected!")
        # 外部ツールの使用
        result = await mcp.call_external_tool(
            connection["connection_id"],
            "solve_equation",
            {"equation": "x^2 + 2*x + 1 = 0"}
        )
        print("External result:", result)
    
# 実行
await connect_external()

# SSH接続後
module load miniconda3
conda env create -f environment.yml
conda activate jupyter-mcp-sage

# ポートフォワーディングでJupyterLab起動
jupyter lab --port 8888 --no-browser \
    --ServerApp.token='nii_mcp_token' \
    --ip=0.0.0.0

{
  "inputs": [
    {
      "type": "promptString",
      "id": "jupyter_token",
      "description": "Jupyter Token",
      "password": true
    }
  ],
  "servers": {
    "jupyter-sage": {
      "command": "python",
      "args": ["src/mcp_sage_helper.py"],
      "env": {
        "JUPYTER_TOKEN": "${input:jupyter_token}"
      }
    }
  }
}

# 強制再インストール
sage -python -m ipykernel install --user --name sagemath --display-name "SageMath" --force
jupyter kernelspec list

# デバッグモード有効化
import logging
logging.basicConfig(level=logging.DEBUG)

# 接続状態確認
print("System info:", mcp.get_system_info())

# 追加パッケージのインストール
pip install package_name

# 環境の再構築
conda env update -f environment.yml

# テストスイートの実行
python -m pytest tests/

# MCPクライアントのテスト
python src/mcp_sage_helper.py

# セキュアなMCP設定例
secure_config = {
    "max_execution_time": 30,  # 実行時間制限
    "allowed_modules": ["numpy", "matplotlib"],  # 許可モジュール
    "sandbox_mode": True,  # サンドボックス有効
    "log_all_executions": True  # 実行ログ記録
}