Files
5chan/docs/agent-playbooks/low-spec-verification.md
T
Tommaso Casaburi 1a33f7dc88 chore(agents): add a machine-wide Playwright browser resource budget
Playwright disables normal background throttling, so a hidden 5chan page keeps
doing P2P and rendering work after a check finishes. Agents verifying in
parallel across worktrees stacked whole browser engines on one machine.

Add scripts/pw-session.sh, a wrapper that permits one active Playwright browser
at a time and records who holds it:

- The lock is machine-wide, not per-repository, because the contended resource
  is RAM and CPU. Every worktree and checkout shares one slot.
- Acquisition is an atomic mkdir. Stale locks clear themselves: `open` reclaims
  any slot whose recorded browser is no longer `status: open` in
  `playwright-cli list --all`, so an interrupted workflow cannot strand the
  budget. When that list cannot be read the lock is left alone, so a broken CLI
  never silently disables the budget.
- `open` exits 75 when the slot is busy; `--wait[=SECONDS]` blocks instead.
- `close` always stops the browser, even when the lock was already lost, and
  never releases a slot held by a different session.
- `status` reports the holder and whether its browser is still alive.

Agent policy now runs browser engines and profiler batches sequentially, uses
Chrome/Blink during iteration and the full engine matrix only for final
verification, and never uses `close-all` or `kill-all` while other agents may
own sessions.

Covered by scripts/pw-session.test.js.
2026-08-01 19:21:01 +02:00

65 lines
3.1 KiB
Markdown

# Low-Spec Device Verification
Use this playbook when you need to know whether loading, navigation, or interaction speed is a real user problem or just an artifact of a fast development machine. It adds CPU + network throttling on top of the normal `playwright-cli` verification flow.
## When to run
- A change touches navigation, data loading, list/feed rendering, images/media, or anything where "is this slow?" matters.
- You cannot tell whether jank or lag is real because the dev machine is high-spec.
- Before claiming a perf-sensitive change "feels fast".
This is an extra pass, not a replacement for the standard Chrome/Firefox/WebKit verification.
## How it works
`playwright-cli` (`@playwright/cli`) has no throttle command, so throttling is applied through CDP via `run-code`. `scripts/pw-throttle.sh` wraps that call. The throttle is set on the page target and persists for the life of the `-s=<session>`, so apply it once, then run the usual `goto` / `snapshot` / `screenshot` commands on the same session.
Throttling is **Chromium-only** (CDP). It does not work on `firefox` or `webkit` sessions — run those engines unthrottled.
## Profiles
| Profile | CPU | Network (down / up / latency) | Models |
|---|---|---|---|
| `mid` | 4x | ~1.6 Mbps / 750 Kbps / 150 ms | mid-tier phone (Lighthouse mobile default) |
| `low` | 6x | ~400 Kbps / 400 Kbps / 400 ms | low-end phone / poor connection |
| `cpu4` | 4x | full speed | isolate JS cost from network |
| `cpu6` | 6x | full speed | isolate JS cost from network |
| `off` | 1x | full speed | reset to normal |
## Usage
```bash
# open a Chromium session, throttle it to a mid-tier phone, then verify as usual
./scripts/pw-session.sh open lowspec https://5chan.localhost --browser=chrome
./scripts/pw-throttle.sh lowspec mid
playwright-cli -s=lowspec snapshot
playwright-cli -s=lowspec screenshot --filename=lowspec-mid.png
# tighten to a low-end device
./scripts/pw-throttle.sh lowspec low
# isolate JS cost (CPU throttle, full-speed network)
./scripts/pw-throttle.sh lowspec cpu4
# reset and finish
./scripts/pw-throttle.sh lowspec off
./scripts/pw-session.sh close lowspec
```
## Measuring, not guessing
Throttled numbers are approximations (CDP throttling, not a real device), but the relative difference is the signal. For the initial page load:
```bash
playwright-cli -s=lowspec eval "() => Math.round(performance.getEntriesByType('navigation')[0].duration)"
```
5chan is a React Router SPA, so in-app navigation does not create a new navigation entry. Time those manually: read `performance.now()` before triggering the route change, then again once the target content appears in the snapshot.
## Caveats
- Chromium-only. Skip on Firefox/WebKit sessions; keep those checks unthrottled.
- Low-spec emulation is a measurement pass, not a machine-resource control. Hold the machine-wide browser slot for the whole pass and close it immediately afterward.
- The `low` latency is intentionally aggressive; if requests time out, fall back to `mid`.
- For render/rerender hotspots after a slow result, use the `profile-browsing` skill (it drives `playwright-cli` + react-scan) on the already-throttled session.