Engineers live telemetry views: pick SVG/Canvas2D/WebGL by mark count, cap a ring buffer, ingest WebSockets off the React render path, downsample with LTTB, scale for devicePixelRatio, and flag stale streams. Pick this chair when shipping live charts, incident canvases, brush-linked analytics, or bandwidth-tight field dashboards. Not for static quarterly BI decks or batch ETL with no socket.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Engineers live telemetry views: pick SVG/Canvas2D/WebGL by mark count, cap a ring buffer, ingest WebSockets off the React render path, downsample with LTTB, scale for devicePixelRatio, and flag stale streams. Pick this chair when shipping live charts, incident canvases, brush-linked analytics, or bandwidth-tight field dashboards. Not for static quarterly BI decks or batch ETL with no socket.
This skill covers the design, engineering, and performance optimization of real-time streaming dashboards and telemetry visualization interfaces. In operational environments, a dashboard is a complex system subject to hardware, network, and layout constraints. This skill provides instructions on managing high-frequency updates, optimizing layout scanning paths, choosing rendering engines, and implementing responsive interactions.
WebSocket-Driven Telemetry Visualizers — Live charts consuming high-frequency data streams (e.g., server resource usage, IoT sensors, trade feeds).
Incident Response Dashboards — Interfaces built for rapid anomaly detection, historical comparisons, and incident triage.
Cross-Filtered Analytics Views — Coordinated dashboards where selecting a region on one chart dynamically filters three other charts.
Resource-Constrained Mobile Views — Dashboards optimized for operators in the field with low bandwidth, high battery drain constraints, and touch interactions.
Node.js 18+ and a modern browser (Chrome 110+, Firefox 110+, Safari 16+) for Canvas/WebSocket APIs.
TypeScript 5+ recommended for type-safe buffer and streaming implementations.
Familiarity with requestAnimationFrame, Canvas2D API, and WebSocket lifecycle events.
For Windows hosts (PowerShell primary): ensure line endings are LF in source files to avoid git checkout issues — run git config core.autocrlf false in the repo root.
Procedure
1. Choose the Rendering Engine
The rendering engine is the most critical decision for dashboard stability. Match the engine to the expected data volume and interaction model.
Parameter
SVG (D3, Recharts)
Canvas2D (Chart.js)
WebGL (deck.gl, Three.js)
Max Marks (Stable)
~1,000 DOM elements
~50,000 pixels/shapes
500,000+ coordinates
CPU Overhead
High (Repaints DOM tree)
Medium (Redraws static pixels)
Low (Delegates to GPU)
GPU Acceleration
Minimal (Browser dependent)
Moderate (Hardware accelerated raster)
Maximum (Direct shader pipeline)
Event Listeners
Native (onClick directly)
Manual (Calculate coordinates on tap)
Manual (Raycasting)
Accessibility (DOM)
Excellent (Screen readers scan tags)
None (Requires hidden ARIA fallback table)
None (Requires hidden ARIA fallback table)
Best Operational Use
Clean, interactive metrics tables
High-frequency CPU timeline charts
Spatial maps, network graphs, IoT nodes
Decision rule:
≤ 1,000 marks and native accessibility required → SVG.
[!WARNING]
Do not use SVG paths to render datasets exceeding 2,000 points. The browser's layout recalculation and DOM rendering system will choke, dropping the viewport frame rate below 10 frames per second.
2. Implement a Ring Buffer for Memory Management
Never append data to arrays infinitely. Establish a maximum window size (e.g., last 500 points) and use a circular queue (ring buffer) to drop old points.
Keep incoming data streams in raw queues. Do not trigger React state updates for every single packet. Pull data from the queue at a throttled interval (e.g., 60fps / 16.6ms) for rendering.
Pattern:
WebSocket onmessage → push to bufferRef.current (no setState).
requestAnimationFrame loop → read bufferRef.current.toArray() → draw to canvas.
Only call setState for lightweight UI metadata (connection status, last value).
4. Implement the WebSocket Streaming Component
Below is a Canvas-based real-time line chart in React. It renders a grid, coordinates drawing tasks, and handles disconnection and stale data indicators.
If visualizing long histories, use the Largest Triangle Three Buckets (LTTB) downsampling algorithm to reduce a 100,000 point series to 1,000 points before passing it to the drawing engine.
LTTB preserves visual shape better than naive every-Nth sampling.
Downsample on the worker thread if the series exceeds 10,000 points to avoid blocking the render loop.
6. Batch Canvas Drawing Calls
[!TIP]
Group multiple canvas drawing operations into a single path call (beginPath(), loop lineTo(), stroke()) instead of calling stroke() inside a loop. Drawing lines as a single batch operation is up to 50 times faster.
7. Handle High-DPI Displays
HTML5 canvas elements look blurry on Retina displays if they are not scaled by the system's window.devicePixelRatio. Always adjust the canvas backing store size to match pixel ratios.
Browsers throttle or freeze requestAnimationFrame loops when a tab is hidden in the background. Sockets can buffer data or overflow. Use the Page Visibility API (document.hidden) to pause WebSocket feeds or clear buffers when tabs are inactive.
document.addEventListener('visibilitychange', () => {
if (document.hidden) {
ws.close(); // or pause feeding
bufferRef.current.clear();
} else {
// reconnect
}
});
9. Sort Out-of-Order Packets
Network packets from WebSockets can arrive out of chronological order. Verify and sort timestamps on insertion to the circular buffer to prevent lines from drawing backward.
10. Implement Coordinated Views and Progressive Disclosure
Overview First, Detail on Demand — Display general statuses as micro-charts (sparklines). Allow the user to click to expand a full-resolution time chart.
Brush and Link — Selecting a specific time range in a master timeline must automatically apply identical time-range filters to all secondary metrics on the page.
Active Annotation — Overlay vertical marker lines on charts corresponding to system events (e.g., "Build #104 Deployed" or "Node Restarted") to give context to telemetry spikes.
11. Indicate Stale Data
[!IMPORTANT]
When rendering data that has stopped streaming (e.g., server offline), do not leave the chart flatlining as if values are normal. Change the line color to neutral grey and overlay a prominent "DATA STREAM STALE" alert box on the canvas.
Pitfalls
Anti-Patterns to Avoid
The Garbage Collector Avalanche: Allocating new JavaScript objects or arrays for every incoming WebSocket packet. This causes frequent garbage collection pauses, making charts stutter. Always reuse objects or write to fixed buffers.
The Equal-Weight Grid: Layouts that present 20 metrics tiles in identical square shapes. Users cannot focus on critical telemetry if everything carries equal visual weight. Create a primary focal timeline.
Hover-Dependent Actions: Hiding critical details (like warning text or exact values) behind hover states. Touch-screen mobile operators cannot hover over elements.
Ambient Glow Clutter: Adding heavy CSS animations, drop shadows, or blinking effects to normal status indicators. Limit motion and highlights strictly to warnings and critical errors.
Edge Cases
Tab Backgrounding: requestAnimationFrame freezes when tabs are hidden. Sockets may buffer and overflow. Use Page Visibility API to pause/clear.
Out-of-Order Packets: WebSocket packets can arrive out of chronological order. Sort timestamps on insertion to prevent backward-drawing lines.
High-DPI Canvas Blur: Canvas elements blur on Retina displays if not scaled by window.devicePixelRatio. Always adjust the backing store size.
SVG Overload: SVG paths exceeding 2,000 points will drop frame rate below 10 FPS due to DOM layout recalculation.
Verification
Architecture Verification Checklist
Has the telemetry buffer capacity been capped to prevent browser tab out-of-memory (OOM) crashes?
Does the connection manager implement an exponential backoff reconnect algorithm when WebSocket connections drop?
Are raw metrics downsampled before rendering charts containing over 5,000 data points?
Is the repaint budget strictly checked? Ensure canvas drawing calls take less than 12ms inside requestAnimationFrame.
Usability & Scanning Checklist
Are labels, status values, and unit tags readable without interactive hovers?
Does the interface clearly indicate when data is stale (e.g., if no socket packet has been received for 10 seconds)?
Are the critical alert thresholds drawn as static horizontal marker lines directly on the chart plane?
Runtime Checks
Memory leak check — Open DevTools → Memory tab → take heap snapshot, let dashboard run for 5 minutes, take another snapshot. Heap should not grow beyond the ring buffer capacity.
Frame rate check — Open DevTools → Performance → Record 10 seconds while streaming. Confirm requestAnimationFrame callbacks stay under 16.6ms and no long tasks exceed 50ms.
Stale data check — Disconnect the WebSocket server. Confirm the chart line turns grey and a "DATA STREAM STALE" alert appears within the expected timeout window (e.g., 10 seconds).
Background tab check — Switch to another browser tab for 30 seconds, return. Confirm no buffer overflow or backward-drawing lines. Buffer should have been cleared or paused.
High-DPI check — Open on a Retina/4K display. Confirm canvas lines and text are crisp, not blurry.
Related Skills
[[frontend-design]] — layout styling and minimalism guidelines
[[scroll-experience]] — canvas rendering techniques and animations
[[ui-ux-pro-max]] — data visualization accessibility standards
[[react-best-practices]] — React hooks and performance optimization guidelines
