kedge — docs — instant sandboxes & scale-out

A fresh sandbox is ready in 0.6 ms. App scale-out from a ready warm-pool instance takes 1.5 ms. When the warm pool is empty, cold scale-out is ready to accept connections in about 16 ms.

These are production-host measurements from local requests. They exclude client network latency and application response time.

ready before demand

Kedge keeps clean sandbox runtimes and initialized app instances ready in warm pools. They are parked in a paused state until needed. A sandbox request gets a fresh, hardware-isolated runtime. App scale-out resumes an instance whose app and network are already initialized.

Neither path repeats a boot or starts the app again.

when the warm pool is empty

Kedge restores an initialized instance from the deploy snapshot, so cold scale-out does not repeat app startup.

see it on a request

App responses report whether that request waited for an instance:

curl -si https://myapp.kedge.run | grep x-kedge-restore

x-kedge-restore: warm
x-kedge-restore-ms: 1.5

x-kedge-restore is warm or cold when the request triggered a restore, and none when an instance was already running. x-kedge-restore-ms appears only for a restore and reports how long the instance took to become ready.

See runtime & scaling for the snapshot lifecycle or sandboxes for one-shot and interactive use.

ask the docsabout this page

This panel is itself an HTML app. Here is its whole source.

<!doctype html>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Ask the Kedge docs</title>
<style>
  body { font: 14px/1.5 system-ui, sans-serif; margin: 0; padding: .75rem 1rem; color: #1a1a1a; background: #fff; }
  form { display: flex; gap: .5rem; }
  input { flex: 1; font: inherit; padding: .4rem .6rem; border: 1px solid #bbb; border-radius: 4px; }
  button { font: inherit; padding: .4rem .8rem; border: 1px solid #888; border-radius: 4px; background: #f4f4f4; cursor: pointer; }
  ol { list-style: none; padding: 0; margin: .75rem 0 0; }
  li { border-top: 1px solid #e5e5e5; padding: .6rem 0; }
  li p { margin: 0 0 .35rem; }
  .answer { color: #333; }
  .answer p { margin: .35rem 0; }
  .answer:empty::before, [data-kedge-pending] .answer::before { content: "\2026"; color: #777; }
  li button { font-size: 12px; padding: .15rem .5rem; margin-top: .4rem; }
  .empty { color: #777; }
</style>

<form data-kedge="me/questions">
  <input type="hidden" name="page" value="{{$url.page}}" pattern="[a-z0-9-]{1,40}" required>
  <input name="q" required maxlength="300" placeholder="Ask about this page" autocomplete="off" autofocus>
  <button>ask</button>
</form>

<ol data-kedge="me/questions?$order=-created_at&$limit=5">
  <template>
    <li>
      <p><b>{{q}}</b></p>
      <div class="answer">{{answer | markdown}}</div>
      <button data-kedge="me/helpful/self?question={{id}}&$toggle">helpful</button>
    </li>
  </template>
  <li data-when=":empty" class="empty">Answers come from the documentation page you are reading. Your questions stay in this browser.</li>
</ol>

<template data-kedge="me/questions" data-agent="Docs"
  data-reads="https://kedge.dev/docs/{{page}}.md https://kedge.dev/llms.txt me/questions?$limit=5"
  data-writes="answer">
  Answer the question from the current documentation page (the first source),
  in Markdown, under 120 words unless a command or example is needed. If the
  page does not cover it, say so in one sentence and link the best page from
  the index (the second source) as https://kedge.dev/docs/&lt;slug&gt;. Earlier
  questions from the same visitor are in me/questions for context.
  Question: {{q}}
</template>