Ignore health status changes in Sentinel push deduplication when the container lifecycle state is unchanged.
Scope stale resource checks to Sentinel servers whose heartbeat is stale, and avoid refreshing resource last_online_at on unchanged statuses.
The endpoint validated private_key_uuid but dropped it from the update,
so the request silently no-op'd. Resolve the UUID to a team-scoped
PrivateKey and include private_key_id in the update payload.
The ownedByCurrentTeamAPI scope was added to StandaloneDocker/SwarmDocker
*after* 4.0.0-beta.470 — running containers on that beta hit a
BadMethodCallException. Rewrites all team scoping to use
whereHas('server', whereTeamId) which works against any v4.x of
Coolify (StandaloneDocker.server_id -> Server.team_id has been there
since the multi-team change).
Also guards attachedTo() with method_exists and falls back to a manual
attached-resource check covering applications + every standalone DB
relation, so delete() doesn't crash on older versions either.
Initial draft called auth()->user()->currentTeam() which returns null in
the API context (Sanctum tokens don't carry the per-user currentTeam
state — that's a session/Livewire concept). Other Api controllers
(ServersController, ScheduledTasksController, etc.) use the canonical
helper getTeamIdFromToken() with a null guard returning 403.
This swap makes all five endpoints work against a real token.
The Destinations section is exposed in the Coolify UI but not via the
REST API. The destination_uuid field is required when creating
applications via create-public-application, but no way to enumerate or
create destinations programmatically existed — this blocks IaC tools
(e.g. an Aspire publisher targeting Coolify).
Adds, scoped to the existing v1 auth:sanctum + ApiAllowed + api.sensitive
group:
GET /api/v1/destinations
GET /api/v1/destinations/{uuid}
DELETE /api/v1/destinations/{uuid}
GET /api/v1/servers/{server_uuid}/destinations
POST /api/v1/servers/{server_uuid}/destinations
The controller uses the existing inline-Validator convention (no Form
Request classes per the API surface's house style), reuses
StandaloneDocker::ownedByCurrentTeamAPI / SwarmDocker::ownedByCurrentTeamAPI
for team scoping (matching ScheduledTasksController etc.), and respects
the `attachedTo()` guard on delete.
No migrations needed — both standalone_dockers and swarm_dockers tables
already carry uuid/name/network/server_id/timestamps.
OpenAPI @OA\ annotations omitted in this commit to keep the diff
minimal; a follow-up can add them in the style of ServersController.
- resolve the GitHub App by a stable identifier during installation
callbacks so installing and re-installing keeps working over the
full lifetime of the App
- verify the installation id received from the callback against the
GitHub API before persisting it
- support re-installing an already configured GitHub App instead of
blocking it
- require an authenticated session and rate limit the setup callback
routes
- extend manifest setup state validity to match GitHub's manifest
code lifetime
Adds feature coverage for the GitHub App setup and installation
callbacks.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Git hosts treat owner/repo names case-insensitively, but the exact
repository match used a case-sensitive comparison, so a payload whose
casing differed from the stored git remote would fail to match and
skip a legitimate deployment.
Lowercase both canonical repository paths before comparing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The manual webhook handlers selected target applications with a
`git_repository LIKE %full_name%` substring query, so a payload
repository name could match unintended applications when repository
names overlap.
Add a `MatchesManualWebhookApplications` trait that validates the
incoming `owner/repo` value and matches `Application.git_repository`
by exact normalized path. Github, Gitlab, Gitea and Bitbucket manual
handlers now use it, reject invalid repository input early, and return
a consistent generic webhook failure payload.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Move Sentinel push handling into a controller and dispatch server update jobs only when container state changes or the force interval elapses. Add opt-in PostgreSQL read/write replica configuration and tune periodic proxy network and storage checks to reduce unnecessary work.
Add feature coverage for replica config, Sentinel push deduplication, deployment log scrolling, and server update job optimizations.
Decode base64 custom_nginx_configuration before model assignment so it is not double-encoded, and allow null values when clearing the setting. Add API coverage for create, update, invalid input, and clearing behavior.
Add model-level hidden fields for secrets, tokens, keys, notification
credentials, deployment logs, and environment values. Allow explicit
read:sensitive API access to reveal gated private keys and deployment logs,
and cover the behavior with feature and unit tests.
Require Docker buildx before Railpack builds, normalize environment
variable keys before validation, and align private deploy key API docs with
the supported dockerfile build pack.
Drop the unstable applications/dockercompose route and controller path now that
service creation is handled by POST /api/v1/services. Add coverage to ensure the
deprecated endpoint stays unregistered while the services endpoint remains
available.
Only eager load nested server settings for service responses when the API token has read:sensitive, preventing read-only tokens from exposing sentinel fields while preserving sensitive access.
Add compose PR fields to the Application hidden attributes and allow database list queries to eager load nested server settings when sensitive access is permitted.
- `/mcp/enable` and `/mcp/disable` now use POST (state-mutating ops)
- `ListServices` queries DB directly instead of loading all projects into memory
- `ListApplications` validates tag arg rejects empty string (not just falsy)
$request->input('force') ?? false reads "false" as a non-empty string,
which PHP coerces to true. Replace with $request->boolean() which uses
filter_var(FILTER_VALIDATE_BOOLEAN), correctly mapping "false" -> false.
Add `exposeNestedServerSecrets()` to Applications, Databases, and
Services controllers so that `read:sensitive`/`root` tokens see
sentinel and logdrain fields on eager-loaded Server + ServerSetting
relations.
ServicesController handles both single models and Eloquent Collections
(listing endpoint passes a Collection per project).
Tests tightened to use JSON-key assertions (`"field":`) to avoid false
positives from field names appearing in values.
Models now declare $hidden for passwords, tokens, db URLs, and compose
fields. API controllers flip from makeHidden-on-deny to makeVisible-on-
allow (can_read_sensitive=true), fixing fields that were never hidden.
Also adds missing fields (mysql/mariadb passwords, logdrain keys, etc.)
to privileged disclosure lists.
Tests added: Feature/Security/ApiSensitiveFieldsTest and
Unit/Models/SensitiveFieldsHiddenTest cover all affected models and
controllers.
Add Model Context Protocol server exposing Coolify infrastructure data
to AI assistants. Includes tools for listing/fetching servers, projects,
applications, databases, and services, scoped to authenticated team tokens.
- Add CoolifyServer with 10 read-only tools (list/get for all resource types)
- Add BuildsResponse and ResolvesTeam traits for shared tool logic
- Add EnsureMcpEnabled middleware guarding /mcp routes
- Add enable/disable MCP API endpoints (root-only)
- Add is_mcp_server_enabled toggle in instance settings and advanced UI
- Add migration for is_mcp_server_enabled column
- Add feature tests for MCP endpoints and toggle API
- Scrub sensitive keys (passwords, tokens, raw IDs) from all responses
Add DetectsSkipDeployCommits trait with two strategies: shouldSkipDeploy
(all commits must contain the marker) for push events, and
shouldSkipDeployAny (any single marker triggers skip) for PR/MR titles
and latest-commit signals.
Apply trait to Bitbucket, Gitea, GitHub, GitLab webhook controllers and
ProcessGithubPullRequestWebhook job. PRs pass pullRequestTitle through
to the job constructor for evaluation.
Introduce a dedicated `audit` log channel (daily rotation, configurable retention via
LOG_AUDIT_DAYS) and a small `auditLog()` / `auditLogWebhookFailure()` helper used to
record state-changing API operations and webhook events.
Instrumented:
- API mutation endpoints (create / update / delete / start / stop / restart) across
applications, services, databases (incl. backups, env vars, storage), servers,
projects + environments, scheduled tasks, private keys, GitHub apps, cloud provider
tokens, Hetzner server provisioning, instance enable/disable.
- Webhook signature verification outcomes for GitHub, GitLab, Bitbucket, Gitea and
Stripe, plus the Sentinel push endpoint.
- Authentication and authorization outcomes via the global exception handler and
the `ApiAbility` middleware (unauthenticated, ability-denied, policy-denied).
The helper is wrapped in try/catch so logging failures never affect the request
path. Successful operations log at `info`; suspicious/denied requests log at
`warning`. Operators wanting a failures-only feed can set `LOG_AUDIT_LEVEL=warning`.
Includes a feature test suite covering the helper, the webhook providers and the
new auth/authorization log paths.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
