v0.4.1

Added

• Workspace deletion. DELETE /admin/tenants/{id} removes a workspace and
  everything scoped to it, in one transaction, plus its ClickHouse click events
  and its Keycloak realm. Only the Owner can delete, the last remaining
  workspace cannot be deleted, and a tenant the caller is not a member of
  answers 404 rather than 403 so the endpoint cannot be used to enumerate
  workspaces. The panel asks for the slug to be typed before it will proceed.
  Deleting frees the slug for reuse. See docs/WORKSPACES.md for what is
  removed immediately and what is removed eventually.
• Workspace creation explains the wait. The request provisions a realm, a
  client, a mapper and a user in Keycloak, so it is legitimately slow. The
  panel now says a sign-in is being prepared, and after a threshold adds that
  it is taking longer than usual and that reloading is safe.
• disabled_reason on the webhook representation in GET /admin/webhooks,
  which lets the panel tell a subscription the user paused from one the system
  disabled.

Fixed

• Webhook destination URLs no longer reach the logs. For Discord, Slack,
  Telegram and the generic connectors the token lives in the URL path, so the
  URL is the credential. It was printed in full at ten log sites. The field is
  now a WebhookUrl whose Display prints host and port only, and reqwest
  does not accept it without an explicit expose(), so reintroducing the leak
  is a compile error rather than a review miss. The subscription's signing
  secret was also readable through the struct's Debug, and no longer is.
• A dead webhook destination is no longer retried forever. 404 and 410
  mean the destination is gone, but every non-2xx response was treated the
  same, so a removed endpoint burned the full attempt budget on every event,
  indefinitely. Those two statuses now get one confirmation attempt, and a
  destination that fails it is disabled with the reason recorded and shown in
  the panel. 429, 5xx, timeouts and transport errors keep the existing
  backoff. 400 and 422 are deliberately not treated as permanent: they
  usually mean our payload is wrong, and disabling a customer's integration
  over our own bug is the worst outcome available.
• Reconnecting Slack over a disabled subscription reactivates it. The OAuth
  merge inherited active: false and the old reason, so the obvious fix
  (reconnect) appeared to work and delivered nothing.
• Log events from main.rs carry fields again. Eight sites built JSON by
  hand inside the log macro, which under QUARK_LOG_FORMAT=json produced a
  serialized object escaped inside message, so the values were text rather
  than queryable fields. Six of them logged errors at info, which no alert
  keyed on severity would ever match.

Imagem do container

docker pull ghcr.io/lucasolopes/quark:0.4.1

Plataformas: linux/amd64, linux/arm64.
Digest: sha256:e2a7fd50ca44730c1293c152804ec0b88ab2339ddbdbe9eecef1aaeb909d42f5

Verifique a proveniencia do build:

gh attestation verify oci://ghcr.io/lucasolopes/quark:0.4.1 --repo lucasolopes/quark