Files
twenty/packages
Félix Malfait 2c5da39dc5 perf(front): load Front chat during browser idle time (#21533)
## Context

The Front support chat bundle
(`chat-assets.frontapp.com/v1/chat.bundle.js`, ~2.3s in a profiling
trace) was being injected only `500ms` after auth + client-config +
workspace-member resolved (`useInstantiateSupportChat.ts`). Because the
effect's gating conditions are themselves network-bound, that `500ms`
still lands the fetch + execute **inside the critical boot window**
(metadata load + first render), where the bundle competes for bandwidth
and main-thread time.

Two pre-existing issues:
- The `500ms` delay was too short to clear the critical window.
- The injected `<script>` already had `defer = true`, but `defer` is a
no-op on dynamically-inserted scripts (they're `async` by default), so
it contributed nothing.
- The `setTimeout` was never cleared, so an effect re-run within the
delay could schedule duplicate loads.

## Change

- Add a small `scheduleIdleCallback(callback, { timeout })` helper
(`src/utils/`) that runs work during a browser idle period via
`requestIdleCallback`, capped by `timeout`, and returns a canceller.
- Use it in `useInstantiateSupportChat` with a `2000ms` cap, and
**return the canceller from the effect** so a pending load is cancelled
on re-run/unmount.

### Why not gate on first interaction?
The launcher must appear proactively to surface an unread-reply badge,
so it has to load without user action. `requestIdleCallback` keeps it
proactive while yielding to the critical path.

### Safari / iOS
`requestIdleCallback` is disabled by default in all shipping Safari/iOS
versions (not Baseline). The helper falls back to a plain `setTimeout`
of the same duration there. Because `requestIdleCallback`'s `timeout` is
a *maximum* (it fires earlier at the first idle gap) while `setTimeout`
fires *at* that value, a single `2000ms` value gives:
- **Chrome/Firefox/Edge/Android**: loads at first idle, guaranteed
within 2s.
- **Safari/iOS**: loads at 2s (a fixed, longer delay — 4× the old
500ms).

Both paths are strictly better than the previous behavior.

## Testing

- `scheduleIdleCallback` unit tests (both the `requestIdleCallback` and
the fallback path, plus cancellation) — 4/4 pass.
- `npx nx typecheck twenty-front` — passes.
- `oxlint --type-aware` + `oxfmt` on changed files — clean.

https://claude.ai/code/session_013YXr5yNGFH1NYUe4ysmiEy

---
_Generated by [Claude
Code](https://claude.ai/code/session_013YXr5yNGFH1NYUe4ysmiEy)_

<!-- This is an auto-generated description by cubic. -->
<a
href="https://cubic.dev/pr/twentyhq/twenty/pull/21533?utm_source=github"
target="_blank" rel="noopener noreferrer"
data-no-image-dialog="true"><picture><source
media="(prefers-color-scheme: dark)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"><source
media="(prefers-color-scheme: light)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-light.svg"><img
alt="Review in cubic"
src="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"></picture></a>
<!-- End of auto-generated description by cubic. -->

---------

Co-authored-by: Claude <noreply@anthropic.com>
2026-06-13 17:53:09 +02:00
..