1 Second. 1 Python Script. 1 Fully Formatted, Fully Functional Tableau Dashboard
We proved that cwtwb could conquer complex nested containers and underlying XML logic. But every frontline data developer knows the brutal truth: building the chart is only 20% of the work. The remaining 80% of your life is drained by tweaking margins, hiding headers, aligning fonts, and clicking the format painter a thousand times. If code can only generate a rough wireframe, it hasn't truly liberated productivity. cwtwb underwent a massive visual-layer refactoring, pushing the styling replication fidelity of human designs to over 90% now.
To be clear, this isn't about AI suddenly having design "taste"—this is about handing the absolute execution power of visual parameters to the underlying SDK. Core Engineering Breakthroughs: Advanced Worksheet Styling Engine: We broke the limits of coarse-grained rendering.
Through the refactored configure_worksheet_style (exposed across both SDK and MCP interfaces), you can now execute fine-grained XML manipulation.
You can hide row field labels, reference lines, and tooltips with a single command.
It includes pane-level styling for cells and data labels, and supports passing dictionary lists for "per-field" formatting to control specific header widths, heights, and custom alignments.
Axis styles can now be fine-tuned globally or targeted to specific measure fields. Calculated Fields & Dual Axis Precision: When creating calculated fields via add_calculated_field, you can now directly inject the default_format (like currency $#,##0 or percentages) into the XML column nodes.
For dual-axis charts (configure_dual_axis), we added mark_color_1 to control the primary axis mark, alongside full support for custom Palette Color Mapping targeting specific :Measure Names. Responsive Dashboard Layout & Rendering Hardening: We removed rigid, hardcoded weight constraints in the layout engine. By introducing Spacers, fluid layout logic, and automatically hiding redundant titles, the generated dashboards are truly adaptive. What does this mean? The hundreds of mouse clicks previously required deep within Tableau Desktop menus are now structured, declarative code configurations. Whether it is a human developer executing a script, or an AI Agent utilizing the MCP to output data, as long as clear visual rules are provided, the code outputs a high-fidelity, presentation-ready dashboard in exactly one second. Moving from "functional replication" to "pixel-perfect delivery" gives that 80% of formatting time back to actual business thinking. This is the true meaning of Workbook as Code.
Southard Jones Elif Tutuk Matthew Miller Adam Mico Allan Folting Roy Raviv Would love your thoughts!
#Tableau #DataVisualization #ArtificialIntelligence #BusinessIntelligence #OpenSource #DataAnalytics #MCP #AIAgents #DataEngineering #datafam #gemini #google #ai #bi #agent #cwtwb #codex #chatgpt #salesforce