Inline Widget logo

Inline Widget

OrganizationPopular
ginlix-ai
inline-widget

Inline HTML widgets: charts, dashboards, data tables rendered directly in the chat via ShowWidget

Overview

Publisherginlix-ai
RepositoryLangAlpha
Skill nameinline-widget
Stars
1.8K
Forks
288
Bundled files
Instructions only
LicenseApache-2.0
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by ginlix-ai on GitHub. Read the source before you install it.

Installation

Install the Inline Widget AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/ginlix-ai/LangAlpha.git /tmp/LangAlpha
mkdir -p .claude/skills
cp -r /tmp/LangAlpha/plugins/langalpha_deliverables/skills/inline-widget .claude/skills/inline-widget
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Inline Widget in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Inline Widget on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Inline Widget is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Inline Widget

Render interactive HTML/SVG widgets directly inside the chat conversation using ShowWidget. Widgets appear inline between text — no sandbox, no preview URL, no side panel. They run JavaScript, so lean into making them interactive and explorable where it helps — something the user can sort, filter, toggle, and hover over, not just a static picture.

When to Use

  • User wants a quick visualization embedded in the conversation (chart, metric card, data table)
  • The visualization is self-contained — all data is embedded in the HTML, no server needed
  • User wants interactivity within the chat: buttons, toggles, hover effects, animated charts
  • The output is a single view — not a multi-page app or dashboard that needs routing

Use interactive-dashboard instead if: User needs a multi-page web app, server-side data, live data refresh, or complex interactivity requiring React/FastAPI.

Read .agents/skills/ui-design/SKILL.md for design quality — its color discipline, chart restraint, and anti-slop principles apply here too. Its font pairings and type scale, though, are for full documents; a widget sits on the chat surface, so use the host-font typography rules below instead.

ShowWidget API

ShowWidget(html: str, title: str | None = None, data_files: list[str] | None = None)
  • html: Raw HTML fragment — no <!DOCTYPE>, <html>, <head>, or <body> tags
  • title: Optional metadata (not displayed to user)
  • data_files: Optional list of sandbox file paths to make available as window.__WIDGET_DATA__

The HTML is rendered in a sandboxed iframe with:

  • CDN libraries: cdnjs.cloudflare.com, cdn.jsdelivr.net, unpkg.com, esm.sh
  • CSS theme variables: automatically injected (see Theme section)
  • sendPrompt('text'): global function to trigger follow-up chat messages
  • window.__WIDGET_DATA__: dict of filename→content for files passed via data_files
  • No network to non-CDN origins: fetch() / XMLHttpRequest to arbitrary URLs are blocked by CSP — only CDN domains (cdnjs, jsdelivr, unpkg, esm.sh) are allowed. Use data_files for sandbox files, or embed small data directly in HTML

Layout Rules (CRITICAL)

The widget sits directly on the chat surface inside a transparent iframe. Follow these rules for seamless integration:

Outer Element — Transparent Shell

The outermost HTML element must have:

  • NO background (or background: transparent)
  • NO border
  • NO border-radius
  • NO box-shadow
  • NO padding — add padding on inner sections only
html
<!-- CORRECT: transparent outer shell -->
<div>
  <div style="background: var(--color-bg-card, #ffffff); border-radius: 8px; padding: 16px; ...">
    ...inner card content...
  </div>
</div>

<!-- WRONG: styled outer wrapper — will be rejected -->
<div style="background: var(--color-bg-page); border: 1px solid ...; border-radius: 8px; padding: 20px;">
  ...content...
</div>

Inner Elements — Use Theme Variables

Inner cards, sections, and components should use CSS variables for styling, always in the fallback form:

css
/* Card */
background: var(--color-bg-card, #ffffff);
border: 0.5px solid var(--color-border-muted, #e4e1dc);
border-radius: 8px;
padding: 16px;

/* Metric card */
background: var(--color-bg-subtle, #f4f2ee);
border: 0.5px solid var(--color-border-muted, #e4e1dc);
border-radius: 8px;

In chat the variables are always injected, so the fallback never shows there — it is what renders when the widget is saved as a workspace .html file and previewed before theme injection or opened standalone. Use the light literals from the html-report skill's table.

Positioning

  • NO position: fixed — breaks iframe auto-sizing (elements collapse to 0 height)
  • Use position: relative for chart containers
  • No nested scrolling — the iframe auto-sizes to fit all content

Theme Variables

These CSS variables are automatically injected and resolve correctly in both light and dark mode:

VariablePurpose
--color-bg-pagePage background
--color-bg-cardCard/panel background
--color-bg-elevatedElevated surface
--color-bg-subtleSubtle/muted background
--color-bg-hoverHover state background
--color-text-primaryPrimary text
--color-text-secondarySecondary/muted text
--color-text-tertiaryHint/label text
--color-border-mutedDefault border (use with 0.5px)
--color-accent-primaryBrand/accent color
--color-profitPositive/gain (green)
--color-lossNegative/loss (red)
--color-warningWarning (amber)
--color-infoInfo (blue)
--color-successSuccess (green)

Never hardcode colors like #333 or rgb(...) for text, backgrounds, or borders — they break in dark mode. Author every color as var(--color-x, #lightLiteral): the injected variable always wins in chat, and the fallback keeps the widget legible on its secondary surfaces. The only exception is chart canvas colors — Chart.js canvas cannot read CSS variables; resolve them via getComputedStyle with a literal fallback.

Charts (Chart.js)

Load Chart.js from CDN and follow these rules:

html
<!-- Wrapper div with explicit height — REQUIRED -->
<div style="position: relative; height: 200px;">
  <canvas id="myChart"></canvas>
</div>

<script src="https://cdn.jsdelivr.net/npm/chart.js"></script>
<script>
  // Read CSS variables for chart colors (canvas can't use var());
  // the literal fallback covers the widget opened outside the chat surface.
  var cs = getComputedStyle(document.documentElement);
  function pick(name, fallback) { return cs.getPropertyValue(name).trim() || fallback; }
  var accent = pick('--color-accent-primary', '#1f5fb4');
  var border = pick('--color-border-muted', '#e4e1dc');

  new Chart(document.getElementById('myChart'), {
    type: 'line',
    data: {
      labels: [...],
      datasets: [{
        data: [...],
        borderColor: accent,
        backgroundColor: accent + '20',
        tension: 0.4,
        fill: true
      }]
    },
    options: {
      responsive: true,
      maintainAspectRatio: false,
      plugins: { legend: { display: true } },
      scales: {
        y: { grid: { color: border } },
        x: { grid: { display: false } }
      }
    }
  });
</script>

Key rules:

  • Set height on the wrapper div, never on the canvas
  • Always use responsive: true, maintainAspectRatio: false
  • Use UMD build from CDN (sets window.Chart global)
  • Read CSS variables via getComputedStyle with a literal fallback (pick()) for chart colors

Typography

  • Font: inherited from host (-apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif)
  • Weight: favor 400 (regular) and 500 (medium); use 600 sparingly for a key figure or heading. Avoid 700 — it reads heavy on the chat surface
  • Heading sizes: h1 = 22px, h2 = 18px, h3 = 16px (all weight 500)
  • Body: 14-16px, weight 400
  • Use sentence case — no Title Case or ALL CAPS (except short metric labels)

Interactivity

Make it explorable

A widget runs JavaScript, so prefer something the user can poke at, not just read. When the data supports it, reach for:

  • Sortable / filterable tables — click a header to sort by P&L or weight; filter to a sector or watchlist.
  • Series & metric toggles — show/hide chart series, switch price ↔ % change, flip timeframe (1M / 6M / 1Y).
  • Hover detail — tooltips on chart points and table rows that surface the underlying numbers.
  • What-if inputs — a slider or field that recomputes a figure live (drag a growth rate, watch the projection update).
  • Tabs / segmented views — split a dense widget (Overview / Holdings / Performance) so the reader drills in.

It all runs client-side over the embedded data — never fetch() a non-CDN origin (CSP blocks it; use data_files for anything large). Keep a meaningful default state so the widget reads correctly before any interaction. Use sendPrompt() only when the next step genuinely belongs back in the chat (a new query, a deeper analysis) — handle exploration the widget can do itself in-place.

sendPrompt()

Call sendPrompt('text') from buttons to trigger a follow-up chat message:

html
<button onclick="sendPrompt('Show detailed revenue breakdown')"
        style="padding: 8px 16px; background: var(--color-accent-primary); color: white;
               border: none; border-radius: 6px; cursor: pointer;">
  Revenue Details ↗
</button>

Add a ↗ arrow on buttons that call sendPrompt() to signal they trigger a chat action.

Refresh / Animation

setInterval and requestAnimationFrame work normally for animations and live tickers:

javascript
setInterval(function() {
  // Update prices, rotate data, animate
  updateDisplay();
}, 3000);

File Data

Use data_files to load data from sandbox files instead of inlining everything in the HTML string. This is especially useful for larger datasets.

Workflow

  1. Generate data files via Python
  2. Pass file paths to ShowWidget via data_files
  3. Access data in the widget via window.__WIDGET_DATA__["filename"]
python
# Step 1: Generate data
import json
data = {"labels": ["Q1", "Q2", "Q3"], "values": [100, 150, 200]}
with open("work/<task_name>/chart_data.json", "w") as f:
    json.dump(data, f)

# Step 2: Agent calls ShowWidget with data_files
ShowWidget(
    html='<div id="chart">...</div><script>var d = JSON.parse(__WIDGET_DATA__["chart_data.json"]); ...</script>',
    data_files=["work/<task_name>/chart_data.json"]
)

Widget access

javascript
// Text files (json, csv, txt, etc.) — returned as strings
var data = JSON.parse(window.__WIDGET_DATA__["chart_data.json"]);
var csvText = window.__WIDGET_DATA__["results.csv"];

// Binary files (png, jpg, etc.) — returned as data URLs
document.getElementById("img").src = window.__WIDGET_DATA__["chart.png"];

Supported file types

  • Text (returned as strings): .json, .csv, .txt, .html, .xml, .svg, .md, .yaml, .yml, .tsv, .geojson, .topojson
  • Binary (returned as data URLs): .png, .jpg, .jpeg, .gif, .webp, .ico

Size limits

Total inline data is capped at 500KB across all files. Keep datasets concise — aggregate or sample large files before passing them.

Blocked Patterns

The following will cause ShowWidget to reject your HTML with an error. Fix and retry:

PatternWhy blocked
new ResizeObserver(...)Host handles iframe sizing — your observer creates infinite resize loops
parent.postMessage(...)Use sendPrompt() instead — direct postMessage bypasses the bridge
window.top.* / window.parent.*Sandboxed iframe — parent access is blocked
position: fixedBreaks iframe auto-sizing
Background/border on outermost elementBreaks seamless integration with chat surface

Design Patterns

Metric Cards Row

html
<div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(150px, 1fr)); gap: 12px; margin-bottom: 16px;">
  <div style="background: var(--color-bg-subtle); padding: 14px 16px; border-radius: 8px; border: 0.5px solid var(--color-border-muted);">
    <div style="font-size: 11px; color: var(--color-text-tertiary); text-transform: uppercase; letter-spacing: 0.06em; margin-bottom: 6px;">Revenue</div>
    <div style="font-size: 24px; font-weight: 500;">$2.4M</div>
    <div style="font-size: 12px; color: var(--color-profit);">+12.5%</div>
  </div>
  <!-- more cards... -->
</div>

Data Table

html
<table style="width: 100%; border-collapse: collapse; font-size: 14px;">
  <thead>
    <tr style="border-bottom: 0.5px solid var(--color-border-muted);">
      <th style="text-align: left; padding: 8px; color: var(--color-text-secondary); font-weight: 500; font-size: 12px;">Symbol</th>
      <th style="text-align: right; padding: 8px; color: var(--color-text-secondary); font-weight: 500; font-size: 12px;">Price</th>
    </tr>
  </thead>
  <tbody>
    <tr style="border-bottom: 0.5px solid var(--color-border-muted);">
      <td style="padding: 8px; font-weight: 500;">AAPL</td>
      <td style="text-align: right; padding: 8px;">$213.18</td>
    </tr>
  </tbody>
</table>

Section with Chart

html
<div style="background: var(--color-bg-card); border-radius: 8px; border: 0.5px solid var(--color-border-muted); padding: 16px; margin-bottom: 16px;">
  <div style="font-size: 16px; font-weight: 500; margin-bottom: 12px;">Performance</div>
  <div style="position: relative; height: 200px;">
    <canvas id="perfChart"></canvas>
  </div>
</div>

Frequently asked questions

What does the Inline Widget AI skill do?

Inline HTML widgets: charts, dashboards, data tables rendered directly in the chat via ShowWidget

Why use Inline Widget on TypingMind?

Because you install it once and use it with any model. Inline Widget is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Inline Widget in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/ginlix-ai/LangAlpha/tree/main/plugins/langalpha_deliverables/skills/inline-widget. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Inline Widget?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Inline Widget?

As many as you like. As long as a model supports skills, you can use Inline Widget with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Inline Widget AI skill free?

Yes. It is published on GitHub by ginlix-ai under the Apache-2.0 license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇