Settings

Test Hub v2.21.0

System check is this machine healthy? — the same checks as `testhub doctor` on the command line

ok	Python	3.9.25 at /opt/pw-offline/venv/bin/python
ok	Test Hub version	2.21.0
ok	Playwright library	version unknown
ok	Bundled browsers	6 browser build(s) in /opt/pw-offline/browsers
ok	Database	SQLite 3.26.0, 484 run(s) recorded
ok	Database schema	up to date
ok	Tests folder	/opt/pw-testhub/data/tests — 25 test file(s)
ok	Results folder	/opt/pw-testhub/data/results
ok	Disk space	17.9 GB free where runs are stored
ok	Schedule timezone	America/Los_Angeles
ok	File permissions	secret key and config are not world-readable
warn	Sleep prevention	this machine cannot take a sleep-inhibitor lock → Run 'caffeinate --status' to see whether it can suspend at all. On a server that must run overnight: sudo systemctl mask sleep.target suspend.target hibernate.target hybrid-sleep.target
ok	OpenCode (AI agent)	installed at /opt/pw-offline/bin/opencode — shells see it after: source /opt/pw-offline/env.sh (add to ~/.bashrc, copied exactly)
ok	Runner	active — tests can run and schedules can fire
ok	Target reachable	HTTP 200 from http://127.0.0.1:8000/demo/

This machine's config /opt/pw-testhub/config.json

Target under test

Name URL — the website Playwright drives (empty = this hub's built-in /demo/ page) Current version of the target — recorded on every run, powers the per-version charts

While the URL is empty the hub tests its built-in demo site, which reports its OWN version (now 3.0.0) — switch it in the demo's Release console. This field applies once you point the hub at your own site.

Runner

Browser Parallel workers (restart to apply)

headed (browser windows visible — needs a display or xvfb on the hub machine)

What to record each one costs disk on every run — turn off what you do not need

Video of each run (biggest item: ~1–5 MB per minute)
Screenshots for the activity replay (~30–60 KB per captured frame; idle time costs nothing)
Playwright trace (~0.5–3 MB per run; the deep-debug artifact)
Keep this machine awake while tests run (holds a sleep inhibitor; a machine that suspends at 01:50 misses the 02:00 schedule and it looks like a broken scheduler. Released as soon as nothing is running. Check with caffeinate --status)
Watch security configuration during runs (headers, cookie flags and requests to other hosts — costs nothing, see the Security page)
Live view while running (one small file, overwritten — negligible)

Your own ctx.screenshot("name") calls inside tests are always saved; these switches control what the hub records automatically.

Scheduler

Timezone for weekly schedules (restart to apply)

Maintenance

Auto-prune run artifacts older than N days at startup (0 = never; database history and all charts are always kept)

Where things live

Hub URL (site.host / site.port)	`https://test-hub.us`
Target under test	`http://127.0.0.1:8000/demo/`
Data directory	`/opt/pw-testhub/data`
Tests folder (source of truth)	`/opt/pw-testhub/data/tests`
Results / artifacts	`/opt/pw-testhub/data/results`
Database	`/opt/pw-testhub/data/db.sqlite3`

To point this hub at a different machine's target (local vs office network vs AWS), edit target.url above — nothing else changes. To move the hub itself, copy the data directory and start manage.py serve there.

Machine clock air-gapped boxes have no NTP — a wrong clock fires schedules at surreal hours and scrambles run history

This machine thinks it is	`2026-09-25 08:58:51`
Your device says	`…`
Fix it	or pick one:

Needs the hub to be running as root (the systemd service does). The timezone name used for schedules is separate — set it under Scheduler above.

Disk & connectivity

Run artifacts on disk	0 bytes across 0 run folder(s)
Target reachability
Free space now	days

Free up space deleting artifacts keeps every chart — only the videos/screenshots/traces go

Delete artifacts, keep history reclaims essentially all the space; pass rates, durations and timings survive	keep newest run(s) per test
Delete only passed runs' artifacts failures are usually the ones worth keeping
Delete run history too removes the rows as well — charts lose those runs

Per-test and per-run delete buttons live on those pages. Automatic cleanup: set retention days above.

Artifact storage where videos/traces/screenshots live

Local disk is the default and the only option offline. On AWS you can push finished runs to a bucket and (optionally) drop the local copy — the database always stays local. Credentials come from the instance role or the environment, never from config.json.

Backend Bucket

Prefix Region Endpoint (S3-compatible only)

delete the local copy after a successful upload (that is the point — it frees the disk)

Backup & transfer upgrades are painful here — take a backup first

Backup everything tests + run history + config, as one zip	⬇ Download backup restore with: `testhub restore <file>`
Export tests test definitions only — the format to carry between machines	⬇ Download tests
Import tests upload a tests zip from another machine; run history here is never touched

A safety backup is taken automatically before an import or a restore, and anything replaced is moved aside — never deleted.

Archived tests

Nothing archived.

Credentials for tests that have to log in

Test files travel — laptop, zip, disc, git. A password written into one travels with it. So credentials live only on this machine in a 0600 file, and a test refers to them by name: ctx.secret("target_password"). Move the test anywhere and it works as soon as that machine has its own value for the name — which is also right, because the lab password should not be the laptop password.

Values are never shown again after saving, never included in a backup or a tests export, and are redacted from run logs, error messages and results even if a test prints one by accident.

Name	Use it as
No credentials stored on this machine.