Stats Compass

Get a URL for the user to upload a CSV or Excel file in their browser. When sharing the upload_url with the user, always include the raw URL as plain text (not as a markdown hyperlink) so the user can see and copy it directly. You may add brief instructions around it. After they confirm the upload is done, call register_uploaded_file() with no arguments. Args: filename: Optional suggested filename — leave blank, the user picks the file content_type: MIME type (default: text/csv) Returns: Upload info with upload_url and instructions.

Return the Stats Compass usage guide covering correct tool workflows, hypothesis test patterns, and common pitfalls. Call this if you are unsure how to structure a statistical test, which tool to use for a task, or have received a validation error.

List all DataFrames in your session. Returns: List of DataFrames with name, shape, columns, and active status.

List files in a directory. Useful for finding data files. Args: directory: Directory path. Supports ~ expansion (e.g., ~/Downloads). Returns: List of files in the directory.

Load a CSV file from a local path. If you do not know the exact path, call list_files first (e.g. list_files(directory="~/Downloads")) to find the file. Do NOT use bash or shell commands — they cannot access the user's machine. Args: path: Absolute path to the CSV file. Supports ~ expansion. name: Name for the DataFrame (default: filename without extension) delimiter: Field delimiter (default: comma) encoding: File encoding (default: utf-8, retry with latin-1 on errors) set_active: Whether to set as active DataFrame (default: True) Returns: DataFrame info with name, …

Load a built-in sample dataset. Available datasets: TATASTEEL, Housing, Bukayo_Saka_7322 Args: name: Dataset name set_active: Whether to set as active DataFrame (default: True) Returns: DataFrame info with name, shape, columns.

Load an Excel file from a local path. If you do not know the exact path, call list_files first (e.g. list_files(directory="~/Downloads")) to find the file. Do NOT use bash or shell commands — they cannot access the user's machine. Args: path: Absolute path to the Excel file. Supports ~ expansion. name: Name for the DataFrame (default: filename without extension) sheet_name: Sheet to load (default: first sheet) set_active: Whether to set as active DataFrame (default: True) Returns: DataFrame info with name, shape, columns, dtypes.

Health check - verify server is running.

Load an uploaded file as a DataFrame. Call this after the user confirms they have uploaded their file. If file_key is omitted, automatically loads the most recently uploaded file. File type is auto-detected from the file extension (csv, xlsx, xls). Args: file_key: Filename to load (optional — omit to auto-load the latest upload) dataframe_name: Name for the DataFrame (default: filename without extension) file_type: File type override - "csv" or "excel" (auto-detected from extension if omitted) encoding: File encoding (default: utf-8). Try "latin-1" if you get codec errors. R…

Run classification workflow: train model, evaluate performance, generate confusion matrix, ROC curve, and feature importance plots. Args: target_column: Column with class labels to predict dataframe_name: Name of DataFrame (default: active) feature_columns: Feature columns (default: all numeric except target) config: Optional classification configuration dict Returns: Workflow result with metrics, model ID, and diagnostic charts.

Run comprehensive EDA report: descriptive stats, correlations, missing data analysis, and auto-generated visualizations. Args: dataframe_name: Name of DataFrame to analyze (default: active) config: Optional EDA configuration dict Returns: Workflow result with steps, metrics, and charts.

Run data preprocessing pipeline: analyze missing data, apply imputation, handle outliers, and remove duplicates. Args: dataframe_name: Name of DataFrame to preprocess (default: active) save_as: Name for the cleaned DataFrame (default: auto-generated) config: Optional preprocessing configuration dict Returns: Workflow result with steps and cleaned DataFrame name.

Run regression workflow: train model, evaluate with RMSE/MAE/R², generate feature importance plots. Args: target_column: Column with continuous values to predict dataframe_name: Name of DataFrame (default: active) feature_columns: Feature columns (default: all numeric except target) config: Optional regression configuration dict Returns: Workflow result with metrics, model ID, and charts.

Run time series forecasting: fit ARIMA model, generate forecasts and visualization. Datasets larger than 500 rows are fitted on their most recent 500 rows (by date_column when given), to keep the fit fast; the forecast DataFrame is then named <name>_last500_forecast. Args: target_column: Column with values to forecast dataframe_name: Name of DataFrame (default: active) date_column: Column containing dates (default: uses row index) forecast_periods: Number of future periods to forecast (default: 30) check_stationarity: Run ADF stationarity test before fitting (default: Fals…

Save a DataFrame to a CSV file. Args: dataframe_name: Name of the DataFrame to save filepath: Path where the CSV file will be saved. For local mode: can be absolute path (e.g., ~/Downloads/data.csv) For remote mode: filename only, saved to session exports index: Whether to write row index (default: False) Returns: Save result with filepath and download_url (if remote).

Save a trained model to a file. Args: model_id: ID of the model to save filepath: Path where the model will be saved. For local mode: can be absolute path (e.g., ~/Downloads/model.joblib) For remote mode: filename only, saved to session exports Returns: Save result with filepath and download_url (if remote).

Get server statistics (admin tool). Returns: Active sessions count, configuration, and session details.

Get information about your current session. Returns: Session info including created_at, dataframes, models and exports.