Esta página se muestra en inglés. Una traducción al español está en curso.
ConfigsJul 3, 2026·3 min de lectura

Speedscope — Interactive Flame Chart Viewer for Performance Profiles

A fast, web-based viewer for performance profiles that renders flame charts, left-heavy views, and sandwich views from many profiler formats.

Listo para agents

Instalación lista para agent

Este activo puede instalarse después de elegir el runtime, revisar el plan y ejecutar el comando correspondiente.

Native · 98/100Política: permitir
Superficie agent
Cualquier agent MCP/CLI
Tipo
Skill
Instalación
Single
Confianza
Confianza: Established
Entrada
Speedscope
Comando de instalación directa
npx -y tokrepo@latest install 75a7642f-76db-11f1-9bc6-00163e2b0d79 --target codex

Ejecutar después de confirmar el plan con dry-run.

Introduction

Speedscope is a web-based performance profile viewer that renders interactive flame charts. It supports profiles from dozens of tools across many languages, making it a universal viewer for CPU and memory profiling data. It runs entirely in the browser with no server-side processing.

What Speedscope Does

  • Renders flame charts, left-heavy (aggregated) views, and sandwich views
  • Imports profiles from pprof, perf, Chrome DevTools, Instruments, and many more
  • Provides smooth zooming and panning through large profile datasets
  • Highlights individual functions across all call stacks (sandwich view)
  • Runs entirely client-side with no data uploaded to any server

Architecture Overview

Speedscope is built with TypeScript and WebGL for high-performance rendering. Profile data is parsed client-side from various formats into a normalized internal model. The flame chart renderer uses WebGL to draw thousands of rectangles efficiently, enabling smooth interaction even with millions of samples. All processing happens in the browser.

Self-Hosting & Configuration

  • Use the hosted version at speedscope.app (no install needed)
  • Install the CLI globally with npm install -g speedscope
  • Run speedscope profile.json to open a profile in your default browser
  • Self-host by serving the static build output from any web server
  • Pipe profiler output directly: py-spy record -o profile.json && speedscope profile.json

Key Features

  • Supports 20+ profile formats across Python, Ruby, Go, Node.js, Java, and native code
  • WebGL-based rendering handles profiles with millions of samples smoothly
  • Three complementary views: time-ordered flame chart, left-heavy, and sandwich
  • No data leaves your machine; all processing is client-side
  • Keyboard shortcuts for quick navigation and function search

Comparison with Similar Tools

  • Chrome DevTools — Built into Chrome but limited to Chrome profiles; Speedscope supports many formats
  • Perfetto UI — Powerful trace viewer; Speedscope is lighter and focused on flame charts
  • FlameGraph (Brendan Gregg) — Generates static SVGs; Speedscope provides interactive exploration
  • pprof web UI — Go-specific; Speedscope supports profiles from any language

FAQ

Q: Is my profile data uploaded anywhere? A: No. All processing happens in the browser. Your data never leaves your machine.

Q: What profile formats are supported? A: pprof, Chrome DevTools, Instruments, perf, collapsed stacks, Pyinstrument, VizTracer, and many more.

Q: Can I use it offline? A: Yes. Install via npm and run locally, or download the static build for fully offline use.

Q: How does the sandwich view work? A: It shows a selected function in the center, with callers above and callees below, aggregated across all occurrences.

Sources

Discusión

Inicia sesión para unirte a la discusión.
Aún no hay comentarios. Sé el primero en compartir tus ideas.

Activos relacionados