Welcome to Data360-MCP Visualization Engine Explorer
An interactive interface for evaluating the Data360 MCP visualization engine vs. Direct LLM renders.
Surprise Me! Action
Randomly generates a data profile (indicator counts, start/end years, country cardinalities, sex/age/urbanisation filters, and custom dimensions) directly from the Data360 MCP, then prompts GPT-4o-mini to frame a natural question matching those parameters. This ensures testing is free of prompt bias.
Run Visual Critique
Captures high-fidelity PNG screenshots of both chart renders and triggers a GPT-4o Vision audit to check for layout issues (axis overlapping, contrast, flatlining). Critique score and improvement lists are displayed under each chart and cached instantly.
How it works
The central view allows side-by-side assessment. On the left is the Data360 Engine Spec, applying the repository's strict visualization routing and scaling rules. On the right is the Direct LLM Spec (GPT-4o), which generates free-form Chart.js graphs.