dsh-antv-ava

When you ask a model about KPI shifts, trends, or category shares, the reply is often a wall of numbers or an isolated chart with no narrative context. dsh-antv-ava lets the model decide whether to tell, show, or do both: narrative text streams phrase by phrase while AntV charts mount the moment usable data rows arrive and keep updating through generation. Both live inside a single conversation bubble instead of splitting into separate cards. The plugin adds semantic data expression—highlighted metrics, direction arrows, absolute changes, rates, contributions, inline proportion micro-charts—and inherits DSH typography and spacing, with chart axes, labels, and grid lines adapting in real time when you switch light or dark themes.

After generation completes, the AVA module scans the stable tabular result and surfaces significant trends, outliers, shifts, shares, and correlations, extending interpretation beyond the model single pass. The GPT-Vis chart engine loads lazily only when an answer contains a chart, and the parsing layer enforces source-size limits while blocking executable markup, dangerous URLs, and private-network resources.

Ideal for analysts, product managers, and anyone who needs model data output to be readable, visual, and trustworthy: business analysis, trend interpretation, rankings, comparisons, and composition breakdowns. It never invents values for visualization and never executes arbitrary HTML, SVG, or scripts emitted by the model.