CSV数据分析工具

mcp-csv-analyzer/
├── mcp_csv_analyzer/       # MCP server implementation
│   ├── __init__.py
│   └── server.py          # Main server with all tools
├── client/                # Azure OpenAI client
│   ├── __init__.py
│   └── app.py            # Interactive client application
├── generate_token/        # Gmail OAuth token generator
│   ├── __init__.py
│   └── generate_oauth_token.py
├── scripts/              # Convenience runner scripts
│   ├── __init__.py
│   ├── run_mcp.py       # Complete system runner
│   ├── run_server.py    # Server runner
│   ├── run_client.py    # Client runner
│   └── run_token_generator.py
├── pyproject.toml        # Project configuration and dependencies
├── test_data.csv         # Sample data file
└── README.md            # This file

# If using uv (recommended)
uv init mcp-csv-analyzer
cd mcp-csv-analyzer

uv sync

AZURE_OPENAI_ENDPOINT=https://your-resource-name.openai.azure.com/
AZURE_OPENAI_DEPLOYMENT=your-deployment-name
AZURE_OPENAI_API_KEY=your-api-key
AZURE_OPENAI_API_VERSION=2024-02-01

uv run mcp

uv run mcp

uv run mcp_server

uv run mcp_client

uv run generate_token

uv run generate_token

uv run generate_token credentials.json

You: Load the test_data.csv file and tell me what columns it has
Assistant: I'll help you load the test_data.csv file and show you its columns...

You: Create a histogram of the numeric columns
Assistant: I'll create histograms for the numeric columns in your dataset...

You: Calculate the correlation between all numeric variables
Assistant: I'll calculate and visualize the correlation matrix for all numeric variables...

You: Send me an email with the analysis results
Assistant: I'll prepare and send an email with the analysis results...

# Tool call: load_csv
{
    "file_path": "data/sales_data.csv",
    "encoding": "utf-8",
    "separator": ","
}

# Tool call: get_dataset_info
{
    "dataset_name": "sales_data"
}

# Tool call: execute_analysis_code
{
    "code": """
# Basic data exploration
print("Dataset shape:", df.shape)
print("\\nColumn info:")
print(df.info())

# Statistical summary
print("\\nStatistical summary:")
print(df.describe())

# Check for missing values
print("\\nMissing values:")
print(df.isnull().sum())

# Create a simple visualization
plt.figure(figsize=(10, 6))
if 'sales' in df.columns:
    plt.hist(df['sales'], bins=30, alpha=0.7)
    plt.title('Sales Distribution')
    plt.xlabel('Sales Amount')
    plt.ylabel('Frequency')
    plt.show()
"""
}

# Tool call: generate_basic_insights
{
    "dataset_name": "sales_data"
}

# Tool call: execute_analysis_code
{
    "code": """
# Correlation heatmap
import seaborn as sns
plt.figure(figsize=(12, 8))
correlation_matrix = df.select_dtypes(include=[np.number]).corr()
sns.heatmap(correlation_matrix, annot=True, cmap='coolwarm', center=0)
plt.title('Correlation Heatmap')
plt.tight_layout()
plt.show()
"""
}

# Tool call: execute_analysis_code
{
    "code": """
# Convert date column and create time series plot
df['date'] = pd.to_datetime(df['date'])
df_sorted = df.sort_values('date')

plt.figure(figsize=(15, 6))
plt.plot(df_sorted['date'], df_sorted['sales'])
plt.title('Sales Over Time')
plt.xlabel('Date')
plt.ylabel('Sales')
plt.xticks(rotation=45)
plt.tight_layout()
plt.show()
"""
}

# Tool call: execute_analysis_code
{
    "code": """
from sklearn.cluster import KMeans
from sklearn.preprocessing import StandardScaler

# Prepare numeric data for clustering
numeric_cols = df.select_dtypes(include=[np.number]).columns
X = df[numeric_cols].dropna()

# Standardize the data
scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)

# Perform K-means clustering
kmeans = KMeans(n_clusters=3, random_state=42)
clusters = kmeans.fit_predict(X_scaled)

# Visualize clusters (first two features)
plt.figure(figsize=(10, 6))
plt.scatter(X_scaled[:, 0], X_scaled[:, 1], c=clusters, cmap='viridis')
plt.title('K-means Clustering')
plt.xlabel(f'{numeric_cols[0]} (standardized)')
plt.ylabel(f'{numeric_cols[1]} (standardized)')
plt.colorbar()
plt.show()

print(f"Cluster distribution: {np.bincount(clusters)}")
"""
}

# Tool call: send_email
{
    "to_email": "recipient@example.com",
    "subject": "Data Analysis Report - Sales Data",
    "body": "Please find attached the analysis results for the sales data.\n\nKey findings:\n- Total revenue: $1.2M\n- Growth rate: 15%\n- Top performing category: Electronics",
    "attachments": ["sales_analysis.pdf", "charts.png"]
}

# Tool call: send_email
{
    "to_email": "team@example.com",
    "subject": "Weekly Data Insights",
    "body": "<h2>Weekly Report</h2><p>This week's <b>key metrics</b>:</p><ul><li>Orders: 1,234</li><li>Revenue: $45,678</li></ul>",
    "body_type": "html",
    "cc_emails": "manager@example.com",
    "attachments": ["weekly_report.xlsx"]
}

# First, generate analysis and export
# Tool call: execute_analysis_code
{
    "code": """
# Generate comprehensive report
report_data = []
for dataset_name, df in loaded_datasets.items():
    summary = {
        'dataset': dataset_name,
        'rows': len(df),
        'columns': len(df.columns),
        'numeric_mean': df.select_dtypes(include=[np.number]).mean().to_dict()
    }
    report_data.append(summary)

# Create visualization
plt.figure(figsize=(12, 6))
# ... create charts ...
plt.savefig('analysis_charts.png')

# Export summary
import json
with open('analysis_summary.json', 'w') as f:
    json.dump(report_data, f, indent=2)
"""
}

# Then send the results
# Tool call: send_email
{
    "to_email": "stakeholders@example.com",
    "subject": "Automated Analysis Report",
    "body": "Attached are the automated analysis results generated on " + datetime.now().strftime('%Y-%m-%d'),
    "attachments": ["analysis_summary.json", "analysis_charts.png"]
}