The studio's hub
Signal Lab talks to one service of the studio's: ProAnima Hub (https://hub.proanima.net, its own repository ProAnima/pas-Hub), which serves every app of ProAnimaStudio. Signal Lab is its project signal-lab. The hub does two things for it:
- Updates. The desktop app's updater asks the hub which release to install.
- Feedback. The ✉ form goes to the hub, which mails it to the developers.
The app holds no secret for either. Its code is open, so a secret could not be kept there; the hub keeps the mailbox's password and decides who gets which release.
Updates
plugins.updater.endpoints in src-tauri/tauri.conf.json, asked in turn:
https://hub.proanima.net/v1/signal-lab/update/{{target}}/{{arch}}/{{current_version}}— the release promoted in the hub'sstablechannel, if it is newer and this install is within its rollout:200with that release's ownlatest.json(its signature untouched; its files are links through the hub to GitHub, so downloads are counted).204says there is nothing, and the updater looks no further — so a paused channel or a partial rollout holds.https://github.com/ProAnima/SignalLab/releases/latest/download/latest.json— only when the hub cannot be reached or answers an error: the latest published release, for everyone.
Either way the updater installs nothing whose signature does not match the public key built into the app (plugins.updater.pubkey), and only the release workflow signs. The check sends X-Install-Id, a random UUID made once (installId in src/lib/platform.ts, kept with the app's settings), so a rollout reaches the same installs as it widens; the hub keeps it with the version, system and channel, and no address. Check once a day in About says so in its tip, and turns the automatic look off.
After a release is published (and only then — drafts never reach the hub): in the hub's console, Signal Lab → Releases, Read from GitHub (or wait for the hourly read), choose the version in stable, set the rollout (say 20 %, then 100 %) and save. Until a version is promoted, the hub offers nothing and the app stays as it is.
Feedback
POST https://hub.proanima.net/v1/signal-lab/feedback, multipart/form-data (engine/src/feedback.rs, the command feedback_send):
| Field | Limit | |
|---|---|---|
message | required, the text | 20 000 characters |
email | optional, becomes Reply-To | one address |
meta | JSON: version, os, arch, mode, lang, screen | 16 KB |
screenshot | files; PNG, JPEG, WebP or GIF, known by their bytes | 6, 8 MB each |
log | files; UTF-8 text | 4, 2 MB each |
| everything together | 15 MB |
The engine checks these limits before it uploads anything (they are the hub's, src/feedback/form.rs there: change both together). The hub answers 202 {"id"} — the id is in the mail too — or {"error": {"code", "params"}}, which the app shows in the reader's language: feedback.message_required, message_too_long, email_invalid, too_many_files, file_type, file_too_large, too_large, rate_limited (retry_after_s), send_failed (id), invalid, disabled (feedback switched off for the project). Anything else the hub says is feedback.failed with its status; a network failure is a transport.* cause about the hub. Limits per sender: 5 in ten minutes, 20 a day.
The project's mail is set in the hub's console, Signal Lab → Settings: to info@proanima.net, from Signal Lab <signal-labs@proanima.net>, SMTP smtp.timeweb.ru, port 465, TLS, login signal-labs@proanima.net; the password is typed there once (stored sealed, never shown again), and Send a test mail proves it. Feedback is on only when all of that is set.
Another hub
SIGNALLAB_HUB_URL points the feedback form elsewhere: set when the engine is compiled, or in a running engine's environment (the end-to-end tour points it at its stand-in, scripts/e2e/fixtures.mjs; engine/tests/feedback.rs at one of its own). The updater's addresses are the app's configuration: another hub means another endpoints list, which tests/delivery.test.mjs keeps equal to engine/src/hub.rs.