Settings
Two separate screens, both under “Settings” in the sidebar: Security Settings (/admin/settings) and Host Settings (/admin/hostsettings).
Table of contents
Security Settings
Platform-wide security policy, grouped into six cards. Every setting here is event-driven: saving invalidates the in-memory cache immediately, and the next request sees the new value — no restart needed, unlike several Host Settings below.
| Group | Settings |
|---|---|
| JWT Configuration | Issuer, Audience, Token Expiry (min), Signing Algorithm (HS256/384/512), Require HTTPS, Key Retention (hours). |
| API Key Policy | Default Expiry (days; 0 = never expires by default), Max Keys Per Tenant (0 = unlimited). |
| Audit Configuration | Retention (days), Log Read Operations, Archive Before Purge, Archive Directory, Purge Hour (UTC), Purge Batch Size. |
| Session Management | Max Concurrent Sessions (0 = unlimited; beyond the limit, new logins are rejected with 429 session_limit_reached). |
| Multi-Factor Auth | Enabled, Provider — shown as 📅 Planned, not yet enforced in code. |
| Password Policy | Min Length, Require Uppercase, Require Number, Require Symbol — enforced on create/change; existing passwords aren’t retroactively checked. |
Every setting is tagged ✓ Live (enforced the moment you save) or 📅 Planned (a placeholder for a future release, not yet enforced) — MFA is currently the only planned group. A few settings carry extra behavior worth knowing:
- JWT Issuer / Audience — changing either starts a grace period: tokens signed under the old value keep working for
jwt.issuer_grace_daysdays, so you don’t invalidate every live token the instant you save. - Password requirements — apply going forward only; nobody is logged out or forced to reset for not meeting a newly-tightened rule retroactively.
Host Settings
Settings specific to this host’s integrations — bulk import limits, outbound notifications, GitHub publishing, the AI provider that powers config generation, and the shared LoginLink server address. Every field has an inline guide (what it is, why it matters, how to fill it in, where to get the value) — click Open all guides to expand every one at once.
Every setting on this page applies immediately via the same live cache as Security Settings — GitHub, AI and Notify included. Nothing here needs a restart.
Decided during setup
A read-only summary of choices made in the one-time setup wizard: whether setup completed, where CrudAdmin keeps its own data, whether automatic database provisioning is configured, and how many admins exist — each with a link to where it can actually be changed (mostly Storage & data).
Platform address
This deployment’s own real public address — what CrudAdmin’s own login tokens (issuer/audience) and the LoginLink API identifier below are named after. Takes effect immediately, with no grace period, unlike most settings here it needs no restart either way.
Left blank, it falls back to the Public:BaseUrl environment variable if the host sets one — the only way to configure this before this field existed — or, if neither is set, to each request’s own address, which the page flags with a warning since that can vary (behind a reverse proxy, for instance). Normally pre-filled once during the first-run setup wizard, using the address the browser was actually at while setup ran.
Provisioning account
The credential CrudAdmin uses to create real databases on a tenant’s behalf (see Tenants → Managed databases). Three kinds:
| Kind | What it needs |
|---|---|
| A MySQL server you run | One admin login that can CREATE, CREATE USER, GRANT — used to create a database and a scoped-down user per tenant. |
| Plesk server | Panel host/IP, port (default 8443), the target subscription’s webspace id, and either your Plesk admin login or a secret key, plus whether to verify its TLS certificate. One subscription per tenant. |
| cPanel account | Panel host/IP, port (default 2083), account username, and an API token (created in cPanel under Security → Manage API Tokens, shown once). |
A Test action validates the credential live before saving; nothing is stored until the test passes.
Bulk import limits
- Realtime row threshold — uploads at or under this row count process while the user waits; larger ones queue in the background (default 10 — the same setting Bulk endpoints’
Autoexecution mode reads). - Max file size (MB) — largest upload allowed (default 10; can’t exceed any smaller limit your own web server or reverse proxy already imposes).
- Allowed extensions — comma-separated, with the dot (e.g.
.csv,.tsv; default.csv). - Google Drive key file / credentials JSON — stored and ready, but not yet read by bulk import in this build; see Reference → Known limitations.
Notify
- Enabled — turns outbound notifications on/off, checked live on every notification — toggling this takes effect on the very next one, nothing in flight needs to finish first.
- Base URL — the notification service’s address, no trailing path. Read fresh on every notification sent.
GitHub
- Owner / Repo — the GitHub repository Generate Config publishes to (just the names, no slashes).
- Personal access token — a GitHub PAT with push rights to that repo. Stored encrypted; never shown again after saving, only who last changed it and when.
- Config path prefix — the folder prefix inside the repo configs are written under (default
databases). Read live by the generator itself; the separate Config Editor (WebSvc.ConfigVisualizer) still reads its own copy once at startup, so changing this prefix without restarting can leave the two disagreeing about where a config lives until the next restart — the one remaining case on this page that benefits from one.
All three GitHub fields are read fresh on every Generate Config / push — if a push ever fails with Unauthorized — Bad credentials right after saving a new token, the token itself is the thing to check, not a stale process. The diagnostic endpoint GET /api/admin/diagnostics/config-snapshot still exists, mainly to catch the rarer case of an appsettings.json/environment-variable override sitting outside Host Settings entirely.
AI
- Provider — OpenAI, Gemini, or AzureOpenAI. Read live, per Generate Config call.
- URL — leave blank for the provider’s normal address; for Azure OpenAI, your resource endpoint. Read live.
- Model — e.g.
gpt-4o-mini. Read live. - API key — stored encrypted, same as the GitHub token. Read live and passed per call — never cached between requests.
This is the model Generate Config calls to produce table groupings, descriptions, and icons — see Known limitations for the cost and reliability implications of that dependency.
LoginLink server
- Issuer — your LoginLink deployment’s address, set once here so two independent features — new-tenant provisioning (on the Tenants page) and admin sign-in (on the Admin Users page) — can never silently point at two different servers.
- CrudApi’s own API identifier — what an app (like a tenant’s own separate product, calling the data API as its users) asks LoginLink for a token addressed to, instead of needing to be the one specific app each workspace already trusts. Left blank, this defaults to the host’s own configured public address (
Public:BaseUrl) — a fixed value, the same one CrudAdmin’s own tokens use — shown on this page as “In effect right now: …” so it’s never a silent guess. IfPublic:BaseUrlalso isn’t set, the page says so with a warning: without it, the identifier falls back to whatever address served each individual request, which can differ behind a reverse proxy and cause correctly-addressed tokens to be refused intermittently — setPublic:BaseUrlfor a stable value. Editable either way, and changing it re-registers the new value in every workspace already connected to LoginLink. Full detail: Features → LoginLink integration.