Spaces:
Sleeping
Sleeping
Nolan Zandi
Claude Sonnet 4.6
feat: add rolling stats, K-Means clustering, and hypothesis testing
7afc8ec | # Shared parameter snippets reused across chart tool schemas. | |
| # Update here to change the description everywhere at once. | |
| _LAYOUT_PARAM = { | |
| "type": "array", | |
| "description": ( | |
| "Optional. An array containing a single JSON-formatted Plotly layout dictionary. " | |
| "Use to set chart title, axis labels, colours, fonts, and other layout properties. " | |
| "Example: [{\"title\": \"Monthly Sales\", \"xaxis\": {\"title\": \"Month\"}}]" | |
| ), | |
| "items": {"type": "string"}, | |
| } | |
| _TRACE_STYLE_PARAM = { | |
| "type": "array", | |
| "description": ( | |
| "Optional. An array containing a single JSON-formatted Plotly trace styling dictionary. " | |
| "Use to control visual properties such as line colour, opacity, and marker style. " | |
| "Do NOT include 'x', 'y', or 'type' keys — those are set automatically from query.csv." | |
| ), | |
| "items": {"type": "string"}, | |
| } | |
| chart_tool_schemas = [ | |
| { | |
| "name": "scatter_chart_generation_func", | |
| "description": ( | |
| "Generates a Plotly scatter plot from query.csv data. " | |
| "Use when the user wants to visualise the relationship between two numeric columns, " | |
| "create a bubble chart (via the size parameter), or overlay a trendline. " | |
| "Returns an HTML iframe — display it verbatim in the chat." | |
| ), | |
| "parameters": { | |
| "type": "object", | |
| "properties": { | |
| "x_column": { | |
| "type": "array", | |
| "description": ( | |
| "One or more column names from query.csv to plot on the x-axis. " | |
| "Multiple columns produce multiple series, each plotted against y_column." | |
| ), | |
| "items": {"type": "string"}, | |
| }, | |
| "y_column": { | |
| "type": "string", | |
| "description": "Column name from query.csv to plot on the y-axis.", | |
| }, | |
| "category": { | |
| "type": "string", | |
| "description": "Optional column name used to colour-code points by a categorical grouping.", | |
| }, | |
| "trendline": { | |
| "type": "string", | |
| "description": ( | |
| "Optional trendline type. One of: 'ols' (linear regression), " | |
| "'lowess' (local smoothing), 'rolling', 'ewm', 'expanding'. " | |
| "Requires trendline_options when using 'lowess', 'rolling', or 'ewm'." | |
| ), | |
| }, | |
| "trendline_options": { | |
| "type": "array", | |
| "description": ( | |
| "Required when trendline is 'lowess', 'rolling', or 'ewm'. " | |
| "An array containing a single JSON-formatted dict of trendline options " | |
| "(e.g. [{\"window\": 7}] for a 7-point rolling average)." | |
| ), | |
| "items": {"type": "string"}, | |
| }, | |
| "marginal_x": { | |
| "type": "string", | |
| "description": "Optional marginal distribution plot along the x-axis. One of: 'histogram', 'rug', 'box', 'violin'.", | |
| }, | |
| "marginal_y": { | |
| "type": "string", | |
| "description": "Optional marginal distribution plot along the y-axis. One of: 'histogram', 'rug', 'box', 'violin'.", | |
| }, | |
| "size": { | |
| "type": "string", | |
| "description": "Optional column name whose values control the size of each point (bubble chart). Negative values are clamped to zero.", | |
| }, | |
| "data": _TRACE_STYLE_PARAM, | |
| "layout": _LAYOUT_PARAM, | |
| }, | |
| "required": ["x_column", "y_column"], | |
| }, | |
| }, | |
| { | |
| "name": "line_chart_generation_func", | |
| "description": ( | |
| "Generates a Plotly line chart from query.csv data. " | |
| "Use for trends over time or any ordered sequence. " | |
| "Returns an HTML iframe — display it verbatim in the chat." | |
| ), | |
| "parameters": { | |
| "type": "object", | |
| "properties": { | |
| "x_column": { | |
| "type": "string", | |
| "description": "Column name from query.csv for the x-axis (typically a date or ordered index).", | |
| }, | |
| "y_column": { | |
| "type": "string", | |
| "description": "Column name from query.csv for the y-axis (numeric values).", | |
| }, | |
| "category": { | |
| "type": "string", | |
| "description": "Optional column name used to split the data into multiple colour-coded lines.", | |
| }, | |
| "data": _TRACE_STYLE_PARAM, | |
| "layout": _LAYOUT_PARAM, | |
| }, | |
| "required": ["x_column", "y_column"], | |
| }, | |
| }, | |
| { | |
| "name": "bar_chart_generation_func", | |
| "description": ( | |
| "Generates a Plotly bar chart from query.csv data. " | |
| "Use for comparing values across categories. Supports grouped/stacked bars via category, " | |
| "and faceted subplots via facet_row or facet_col. " | |
| "Returns an HTML iframe — display it verbatim in the chat." | |
| ), | |
| "parameters": { | |
| "type": "object", | |
| "properties": { | |
| "x_column": { | |
| "type": "string", | |
| "description": "Column name from query.csv for the x-axis (category labels).", | |
| }, | |
| "y_column": { | |
| "type": "string", | |
| "description": "Column name from query.csv for the y-axis (numeric values).", | |
| }, | |
| "category": { | |
| "type": "string", | |
| "description": "Optional column name used to colour-code bars into grouped or stacked series.", | |
| }, | |
| "facet_row": { | |
| "type": "string", | |
| "description": "Optional column name. Creates one subplot row per unique value — useful for comparing distributions across a second dimension.", | |
| }, | |
| "facet_col": { | |
| "type": "string", | |
| "description": "Optional column name. Creates one subplot column per unique value.", | |
| }, | |
| "data": _TRACE_STYLE_PARAM, | |
| "layout": _LAYOUT_PARAM, | |
| }, | |
| "required": ["x_column", "y_column"], | |
| }, | |
| }, | |
| { | |
| "name": "pie_chart_generation_func", | |
| "description": ( | |
| "Generates a Plotly pie chart from query.csv data. " | |
| "Use when the user wants to show part-to-whole proportions. " | |
| "Returns an HTML iframe — display it verbatim in the chat." | |
| ), | |
| "parameters": { | |
| "type": "object", | |
| "properties": { | |
| "values": { | |
| "type": "string", | |
| "description": "Column name from query.csv containing the numeric value for each slice.", | |
| }, | |
| "names": { | |
| "type": "string", | |
| "description": "Column name from query.csv containing the label for each slice.", | |
| }, | |
| "data": _TRACE_STYLE_PARAM, | |
| "layout": _LAYOUT_PARAM, | |
| }, | |
| "required": ["values", "names"], | |
| }, | |
| }, | |
| { | |
| "name": "histogram_generation_func", | |
| "description": ( | |
| "Generates a Plotly histogram from query.csv data. " | |
| "Use to show the frequency distribution of a numeric column. " | |
| "Supports normalisation (percent, probability, density) and aggregation functions per bin. " | |
| "Returns an HTML iframe — display it verbatim in the chat." | |
| ), | |
| "parameters": { | |
| "type": "object", | |
| "properties": { | |
| "x_column": { | |
| "type": "string", | |
| "description": "Column name from query.csv whose values are binned on the x-axis.", | |
| }, | |
| "y_column": { | |
| "type": "string", | |
| "description": "Optional column name aggregated per bin via histfunc (e.g. sum of sales per price bucket).", | |
| }, | |
| "histnorm": { | |
| "type": "string", | |
| "description": "Optional normalisation. One of: 'percent', 'probability', 'density', 'probability density'.", | |
| }, | |
| "category": { | |
| "type": "string", | |
| "description": "Optional column name used to overlay multiple colour-coded histograms.", | |
| }, | |
| "histfunc": { | |
| "type": "string", | |
| "description": "Optional aggregation function applied to y_column per bin. One of: 'avg', 'sum', 'count'.", | |
| }, | |
| "data": _TRACE_STYLE_PARAM, | |
| "layout": _LAYOUT_PARAM, | |
| }, | |
| "required": ["x_column"], | |
| }, | |
| }, | |
| { | |
| "name": "box_chart_generation_func", | |
| "description": ( | |
| "Generates a Plotly box plot from query.csv data. " | |
| "Use to visualise the distribution of a numeric column and identify outliers. " | |
| "Especially useful for comparing distributions across categories. " | |
| "Returns an HTML iframe — display it verbatim in the chat." | |
| ), | |
| "parameters": { | |
| "type": "object", | |
| "properties": { | |
| "y_column": { | |
| "type": "string", | |
| "description": "Column name from query.csv containing the numeric values to distribute on the y-axis.", | |
| }, | |
| "x_column": { | |
| "type": "string", | |
| "description": "Optional column name. Groups data into one box per unique value on the x-axis.", | |
| }, | |
| "category": { | |
| "type": "string", | |
| "description": "Optional column name used to colour-code boxes by a secondary grouping.", | |
| }, | |
| "layout": _LAYOUT_PARAM, | |
| }, | |
| "required": ["y_column"], | |
| }, | |
| }, | |
| { | |
| "name": "correlation_heatmap_func", | |
| "description": ( | |
| "Computes pairwise Pearson correlations between numeric columns in query.csv and renders " | |
| "the result as a colour-coded heatmap (blue = positive, red = negative). " | |
| "Use when the user asks which variables are related, correlated, or associated with each other. " | |
| "Returns an HTML iframe — display it verbatim in the chat." | |
| ), | |
| "parameters": { | |
| "type": "object", | |
| "properties": { | |
| "columns": { | |
| "type": "array", | |
| "description": "Optional list of numeric column names to include in the matrix. If omitted, all numeric columns from query.csv are used. Avoid ID or index columns.", | |
| "items": {"type": "string"}, | |
| }, | |
| }, | |
| "required": [], | |
| }, | |
| }, | |
| { | |
| "name": "rolling_stats_func", | |
| "description": ( | |
| "Generates a rolling statistics / moving average chart from query.csv data. " | |
| "Overlays rolling aggregations (mean, std, min, max) on top of the original series. " | |
| "Use when the user asks for a moving average, rolling average, rolling statistics, or wants to smooth a time series. " | |
| "Returns an HTML iframe — display it verbatim in the chat." | |
| ), | |
| "parameters": { | |
| "type": "object", | |
| "properties": { | |
| "x_column": { | |
| "type": "string", | |
| "description": "Column name from query.csv for the x-axis — typically a date or sequential index.", | |
| }, | |
| "y_column": { | |
| "type": "string", | |
| "description": "Column name from query.csv containing the numeric values to compute rolling stats on.", | |
| }, | |
| "window": { | |
| "type": "integer", | |
| "description": "Rolling window size in number of rows. Default 7. Infer from the user's request.", | |
| }, | |
| "stats": { | |
| "type": "array", | |
| "description": "Statistics to overlay. Valid values: 'mean', 'std', 'min', 'max'. Defaults to ['mean'] if omitted.", | |
| "items": {"type": "string"}, | |
| }, | |
| "category": { | |
| "type": "string", | |
| "description": "Optional column name to group the data, producing separate rolling stat lines per group.", | |
| }, | |
| "layout": _LAYOUT_PARAM, | |
| }, | |
| "required": ["x_column", "y_column"], | |
| }, | |
| }, | |
| { | |
| "name": "table_generation_func", | |
| "description": ( | |
| "Formats query.csv results as a styled HTML table. " | |
| "Use when the user wants to view raw query results in a readable format, " | |
| "or when result data is too large to describe in text. Displays up to 200 rows. " | |
| "Returns an HTML table — display it verbatim in the chat." | |
| ), | |
| "parameters": {"type": "object", "properties": {}}, | |
| }, | |
| ] | |