Data Analysis

Data Analysis Skill

Description

Load data files (CSV, XLSX, JSON, Parquet) into the AG Grid viewer, run pandas queries, save results, and generate visualizations.

Tools Used

Primary (Data Grid workflow)

• data_list - List available data files in /workspace/data/
• data_load - Load a data file into AG Grid (returns markdown preview for context)
• data_query - Execute pandas operations on loaded data (filter, aggregate, transform)
• data_save - Save the current DataFrame to a file

Secondary (Jupyter workflow for visualization)

• jupyter_execute - Execute Python code in Jupyter kernel (for plots and complex analysis)
• update_notebook - Add cells to Jupyter notebook
• update_gallery - Display generated plots in the gallery

Workflow

Recommended: Data Grid Workflow

For tabular data exploration, use the data tools which provide a spreadsheet-like experience:

1. List files: data_list to see what's in /workspace/data/
2. Load data: data_load to read a file and display in AG Grid
   • You'll receive a markdown preview to understand columns and types
3. Query/Filter: data_query to run pandas operations
   • The df variable contains the loaded data
   • Set result = ... to define output
4. Save results: data_save to export to CSV/XLSX

Alternative: Jupyter Workflow

For visualization, statistical analysis, or ML, use Jupyter tools:

1. Load data with jupyter_execute running pandas code
2. Create visualizations with matplotlib/seaborn
3. Display plots with update_gallery

Usage Patterns

Load and Explore Data

When user says: "Analyze this dataset" or "Show me the data"

1. data_list to find available files
2. data_load with the target file
3. Review the markdown preview to understand structure
4. data_query with result = df.describe() for statistics
5. Offer filtering, sorting, or visualization

Filter and Transform

When user says: "Show only rows where X > Y" or "Group by category"

1. data_query with pandas filter/groupby code
2. Grid updates automatically with filtered results
3. Inform user of result count and preview

Save Processed Data

When user says: "Export this" or "Save as Excel"

1. data_save with desired filename and format
2. Report file location and size

Visualize Data

When user says: "Create a chart" or "Plot the distribution"

1. Use jupyter_execute with matplotlib/seaborn code
2. Save plot and display via update_gallery

Code Snippets for data_query

Filter rows

result = df[df['score'] > 90]

Group and aggregate

result = df.groupby('category').agg({'value': ['mean', 'sum', 'count']}).reset_index()

Sort by column

result = df.sort_values('date', ascending=False)

Add computed column

df['ratio'] = df['value_a'] / df['value_b']
result = df

Summary statistics

result = df.describe()

Handle missing values

result = df.dropna(subset=['important_column'])

Best Practices

1. Start with data_list: Always check what files are available first
2. Use data_load first: Load data to get markdown preview before querying
3. Keep queries simple: One operation per data_query call for clarity
4. Save intermediate results: Use data_save for important filtered datasets
5. Switch to Jupyter for plots: AG Grid is for tabular data, use Jupyter for visualizations