Limits & errors

Protect the host. Keep the message intact.

VSLZ uses bounded inputs so a model cannot accidentally freeze the host with pathological data.

Default guardrails

  • Charts accept up to 5,000 rows.
  • Networks accept up to 1,000 nodes and 2,000 edges.
  • Mermaid source is limited to 20,000 characters.
  • Labels, metadata, and annotations have bounded lengths.
  • Invalid input returns a structured, non-destructive error.

Icons, glyphs and inline images

Each visual accepts up to 64 asset definitions, 256 bindings and 24 images. Supply PNG, JPEG or static WebP as inline base64 data URLs: at most 512 KiB per image, 4096 pixels per side, four megapixels and 2 MiB combined per visual. Infographics allow 5 MiB combined across blocks. Alt text is required. Unknown targets, duplicate bindings, remote URLs, file paths, SVG and HTML are rejected.

There is no server-side photo storage, upload registration or image proxy. Corrupt images retain text labels and a neutral badge fallback. Resize images before supplying them.

What VSLZ will not do

VSLZ does not silently drop rows, run arbitrary JavaScript, fetch remote data from a render tool, persist visualization payloads, or render images on the server.

When a visual is too dense

Filter, cluster, or split the request. For networks, ask for the critical path or a subset of nodes. For charts, reduce the categories or choose a chart type that supports the analytical question.