Four renderers. One optional guide.

One quantitative visual → render_chart. One structured explanation → render_diagram. One exploratory network → render_network. Several coordinated pieces of evidence → render_infographic.

Explore eight composition examples with full JSON and prompts. Composition blocks retain the existing child specs. Use unique id values and kind for each block; connectors reference those IDs.

get_visual_guidance

Use this read-only helper when the model needs current deployment capabilities, selection guidance, or valid example inputs. Pass task as chart, diagram, network, infographic, or all. It does not render or fetch evidence. Clear requests can go directly to a renderer.

The same deployment-aware guide is available as the MCP resource vslz://guidance/visual-tools. Connection instructions explain when a visual helps, how to preserve evidence, and how to repair validation errors. Examples contain synthetic data and must be adapted to the user's evidence.

render_chart

Use for quantitative stories: ranking, comparison, trends, composition, contribution, before/after, ranges, and relationships. Include dataContext.unit and dataContext.period when known; use options.percent, options.currency, and options.compactNumbers so axes, labels, and hover values stay interpretable.

{
  "type": "bar",
  "insight": "Enterprise is the growth engine",
  "title": "Net retention by segment",
  "data": [
    { "segment": "Enterprise", "retention": 1.18 },
    { "segment": "SMB", "retention": 0.96 }
  ],
  "encoding": { "category": "segment", "value": "retention" },
  "dataContext": { "unit": "percent", "period": "FY2025" },
  "options": { "percent": true, "showValues": true }
}

Chart selection and input guide

Choose the reader job first, then provide the encoding and context that make the measurement unambiguous. Values must be expressed in the declared unit. For scatter and bubble charts, use dataContext.units when x, y, size, start, or end do not share one unit.

Editorial chart templates · enabled by default

Basic charts remain the default. Enabled deployments accept an optional template alongside the matching chart type. The four templates add small-multiple lines, histograms, connected observations and radar comparisons. Disabled deployments return a clear error for template requests.

Keep the existing encodings and context. Supply numeric measurements as numbers. Small multiples require at least two panel series. Histograms count observations in equal-width bins; connected scatter follows supplied row order. Use note for context. Exact values remain available in the data table.

Icons, glyphs and inline raster images

Add optional visualAssets to chart, diagram or network specs, including nested infographic blocks. get_visual_guidance lists curated icon/glyph keys, placements and limits. Charts show a separate identity key; networks reserve badge space beside labels; compact diagrams use the key when a badge would crowd text.

{
  "visualAssets": {
    "version": 1,
    "definitions": [
      {
        "id": "enterprise",
        "kind": "icon",
        "key": "building"
      }
    ],
    "bindings": [
      {
        "target": {
          "type": "category",
          "field": "segment",
          "value": "Enterprise"
        },
        "assetRef": "enterprise",
        "placement": "label"
      }
    ]
  }
}

Bindings target an existing encoded category/series value, or an entity with field: "id" and an existing node ID. Mermaid identity keys require explicit nodes. Use kind: "glyph" with circle, square, diamond, triangle or star for font-independent identity. Glyphs do not encode quantitative size.

For a supplied image, use kind: "image", src containing an inline PNG/JPEG/static WebP data URL, required alt text, and optional fit: "contain" | "cover". No server-side photo storage, upload registration, remote URL or proxy is supported. Each image is limited to 512 KiB and four megapixels; see all limits. Do not invent image bytes or replace data marks with photos. Browse the identity fixtures for live examples.

render_diagram

Use for explicit structure: processes, systems, sequences, state transitions, entity relationships, timelines, and conceptual maps. Mermaid runs with strict security settings.

{
  "insight": "The decision path has three gates",
  "mermaid": "flowchart LR\n A[Input] --> B[Decision] --> C[Outcome]"
}

For an authored editorial diagram, provide a structured template, nodes, and edges. Mermaid and a structured template cannot be mixed. Editorial diagrams are independently feature-flagged, so Mermaid remains the basic path.

{
  "template": "process_flow",
  "insight": "Every handoff has an owner",
  "nodes": [{ "id": "a", "label": "Intake" }, { "id": "b", "label": "Review" }],
  "edges": [{ "source": "a", "target": "b", "label": "qualify" }]
}

render_network

Use for relationships that are too dense or exploratory for a hand-authored diagram. Layout, pan, zoom, fit, selection, and highlight behavior are client-side.

{
  "title": "Service dependencies",
  "nodes": [{ "id": "api", "label": "API" }, { "id": "db", "label": "Database" }],
  "edges": [{ "source": "api", "target": "db" }],
  "layout": "dagre"
}

Composition templates

Choose evidence_network, hero_metric_support, evidence_board, charticle, small_multiples_story, timeline_with_evidence, comparison_story, or cause_effect_story. Maximum 12 blocks, six visual blocks, one network and 12 connectors. Sources remain visible. Comparison and small-multiple charts with matching units share a numeric domain by default.

Declare scaleGroups to coordinate specific chart blocks, or set independent to true to label separate scales explicitly. Optional linking uses a group and an entityField or timeField; tapping persists selection until cleared. Compact layouts stack evidence and retain connector meanings as text.

render_infographic

Use for a compact authored story that combines metrics, text, diagrams, networks, and up to six visual blocks. Infographics require an explicit template and are enabled by default unless disabled by deployment configuration.

{
  "template": "hero_metric_story",
  "insight": "Retention improved while acquisition held steady",
  "blocks": [
    { "kind": "metric", "label": "Retention", "value": 84, "unit": "%", "change": 6, "changeUnit": "pp" },
    { "kind": "text", "text": "The improvement is concentrated in returning customers." }
  ]
}