swarm.getNodes/getNodeInfo/getNodeApps/getAppInfos accepted a caller-supplied
serverId and ran remote Docker Swarm reads against it with no organization
check, exposing another tenant's swarm topology cross-org (getContainerStats
already had the check; the others did not). getNodeInfo additionally
interpolated nodeId unescaped into 'docker node inspect ${nodeId}'.
Adds an org-ownership assertion to the four unscoped handlers and passes nodeId
through shell-quote in getNodeInfo.
Closes GHSA-jj6h-388v-9rwm
findServerById eagerly loads the sshKey relation (needed for server-side SSH
operations) including the plaintext privateKey. server.one and server.remove
returned that record to the client, exposing the private key to any member
with server:read regardless of canAccessToSSHKeys.
Adds redactServerSshKey() in the server service and applies it to the server.one
and server.remove responses. No client feature consumes the private key (the SSH
key management UI uses the dedicated sshKey router), and server-side callers keep
using findServerById directly, so behaviour is unchanged.
Closes GHSA-w9cp-jqfw-4xj9
The .one / getRepositories / getBranches / testConnection handlers for
github, gitlab, gitea and bitbucket resolved a provider by bare id under
protectedProcedure and returned the full row (OAuth tokens, app private
keys, webhook secrets) with no organization or entitlement check, letting
any authenticated user read another org's git-provider credentials (IDOR).
Adds assertGitProviderAccess() in the git-provider service (rejects cross-org
with NOT_FOUND and non-entitled same-org with FORBIDDEN, reusing the existing
getAccessibleGitProviderIds entitlement logic) and calls it in every read
handler before returning secret-bearing data.
Closes GHSA-66r5-5m5f-57vg
User-controlled git fields (customGitUrl, branch names, repo owner/name,
gitlab namespace, SSH hostname) were interpolated unescaped into git clone /
ssh-keyscan shell commands run via execAsync / execAsyncRemote, allowing
authenticated OS command injection. All such values are now passed through
shell-quote before interpolation (defense at the sink, covering every code
path including the compose branch bypass).
Closes GHSA-qxcw-cx35-2hrw, GHSA-hrfh-82jj-3q46, GHSA-6693-xv3f-69px,
GHSA-qwwm-hc7m-7xp9, GHSA-grrj-6xrh-j6vp, GHSA-x2p2-qq8g-2mqq,
GHSA-cg8g-x23v-5fw8
The randomize/isolated-deployment volume transform split mount strings
on ':' and kept only the first two segments, so an access mode like
:ro, :z or :Z was silently dropped and read-only mounts became
read-write. Keep the full path+mode remainder when rebuilding the
mount string.
Fixes#4818
The @dokploy/server modules get bundled multiple times per process
(esbuild inline copy, compiled node_modules copy, and several Next.js
chunks via transpilePackages), so every module-level side effect ran
once per copy: up to 6 postgres pools, 6 dockerode clients and 6
better-auth instances in the server process, plus the repeated
'Using Docker socket' logs at boot.
- db/index.ts: use the globalThis cache in production too (one pool per process)
- constants/index.ts: cache the dockerode client on globalThis
- lib/auth.ts: wrap betterAuth() in a factory and cache the instance
- Dockerfile: exec node directly instead of leaving pnpm resident (~100MB RSS)
- Added a new service to fetch public whitelabeling configuration for unauthenticated contexts.
- Updated the whitelabeling router to utilize the new service for public requests.
- Enhanced license validation checks to ensure proper access control based on organization licenses.
- Added a check to prevent empty values from being processed in the onValueChange handler for Bitbucket, Gitea, GitHub, and GitLab providers.
- Removed unnecessary defaultValue prop from Select components to streamline the code.
- Updated button styles to remove background color for better consistency across the UI.
- Enhanced volume backup selection to display a message when no volumes are found.
This update enhances user experience by ensuring that empty selections are handled gracefully and improves the overall visual consistency of the UI components.
The Ollama detection matched any URL containing "ollama", which hides
the API Key field for Ollama Cloud (ollama.com) and drops the key from
the createOllama() client, so cloud requests go out unauthenticated.
Narrow the rule to localhost-only Ollama and forward the API key as a
Bearer header when provided.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(backup): redact S3 credentials from logs and error output (#4621)
S3 backup credentials (access key + secret) were logged in plaintext
to Dokploy service stdout via logger.info in getBackupCommand() and
console.error in keepLatestNBackups(). Any operator with access to
service logs could recover S3 credentials.
Added redactRcloneCredentials() pure function that masks
--s3-access-key-id and --s3-secret-access-key values with [REDACTED].
Applied to both the structured logger call and the error handler.
Closes#4621
* fix(backups): redact sensitive information in error logs during web server backup process
Updated error handling in the web server backup function to redact Rclone credentials from error messages before logging and notification. This change enhances security by preventing sensitive data exposure in logs.
---------
Co-authored-by: Mauricio Siu <siumauricio@icloud.com>
OpenAI strict mode for response_format requires every property to be
listed in 'required'. Optional fields must instead be modeled as
nullable so the key stays required while still allowing no value.
Fixes#4267
* fix(domain): validate hostname format to reject invalid characters
Underscores and other invalid characters were accepted in domain
inputs with no validation, causing Let's Encrypt to silently fail
certificate issuance while Dokploy fell back to a self-signed cert.
Fixes#4716
* fix(create-server): update SSH key label for clarity in server creation form
The nightly access-log-cleanup job hardcoded "dokploy-traefik" as the
container name when sending SIGUSR1. In Docker Swarm mode Traefik runs as
a service task named "dokploy-traefik.1.<task-id>", so `docker exec
dokploy-traefik` fails every night with "No such container". The log file
is rotated (inode changes) but Traefik never reopens it, leaving the
on-disk access.log frozen while real logs go to a deleted file handle.
Resolve the running container id dynamically with `docker ps --filter`,
matching the pattern already used elsewhere in the codebase, so it works
for both standalone and swarm deployments. Skip gracefully if no running
container is found.
Closes#4620
Branch names containing '#' (e.g. feat#123) were rejected by
VALID_BRANCH_REGEX when saving a git provider configuration, even
though '#' is a legal git ref character.
Add '#' to the allowed character set. The change propagates to the
backend zod schemas and all provider UI forms, since they share this
constant.
'#' is not a shell injection vector: the regex still rejects every
character needed to terminate a command (; | & $ ( ) ` newline space
quotes), and '#' only starts a shell comment at the beginning of a
word, never mid-argument as in 'git clone --branch feat#123'.
Fixes#4585
cloneGitRepository runs `ssh-keyscan <host> >> known_hosts` as one step
of a `set -e` script. Hosts whose SSH endpoint waits for the client's
identification string first — Hugging Face's hf.co among them — never
complete the keyscan handshake, so it exits 1 and `set -e` aborts the
deploy before `git clone` ever runs.
Make ssh-keyscan non-fatal and let the real ssh client record the host
key on first connect (StrictHostKeyChecking=accept-new), which reaches
hosts ssh-keyscan can't scan. Same TOFU trust model, so no regression;
GitHub/GitLab/Gitea still pre-seed and verify known_hosts as before.
Remove .toLowerCase() transform from registryUsernameSchema.
AWS ECR requires the username to be exactly 'AWS' (uppercase) for
authentication. Docker Hub usernames are case-insensitive for login,
so preserving case is safe for all providers.
Closes#4632
The canAccessToGitProviders legacy override only granted read access, so
members with the Git Providers toggle enabled could not add providers — the
create/delete endpoints require gitProviders.create / gitProviders.delete.
This mirrors how the SSH Keys toggle already grants read/create/delete.
The git-provider remove endpoint now restricts non owner/admin roles to
deleting only their own providers (matching the ownership model used for
visibility and sharing), while owner/admin can still delete any provider in
the organization.
Closes#4695
* feat: update dependencies and enhance UI components in Dokploy
- Updated various package versions in pnpm-lock.yaml to improve compatibility and performance, including React and Tailwind CSS.
- Modified components.json to change the styling to "radix-nova" and added new properties for icon library and menu color.
- Refactored multiple components to use updated class names for better styling consistency and responsiveness.
- Removed unused Radix UI components from package.json to streamline dependencies.
- Adjusted layout and styling in several dashboard components for improved user experience and visual appeal.
- Enhanced tooltip and form item components for better accessibility and usability.
* refactor: enhance UI components in HandleCertificate and SidebarLogo
- Updated the HandleCertificate component to improve dialog content height and textarea styling for better usability.
- Adjusted the SidebarLogo component layout to enhance alignment and spacing, ensuring a more consistent appearance across the sidebar.
- Implemented responsive design adjustments to textarea elements, preventing overflow and improving user experience.
* refactor: improve UI consistency and styling across dashboard components
- Updated Card components in ShowDeployments, ShowSchedules, and ShowVolumeBackups to remove unnecessary border styles for a cleaner look.
- Enhanced TabsList components in ShowProviderForm and ShowProviderFormCompose by adding a variant for improved visual distinction.
- Adjusted spacing in AdvancedEnvironmentSelector for better layout and readability.
- Removed redundant border classes in Service component for a more streamlined design.
* [autofix.ci] apply automated fixes
* chore: update package dependencies and refactor email rendering
- Upgraded React and React DOM to version 19.2.7 across multiple packages for improved performance and compatibility.
- Updated TSX to version 4.22.4 in various package.json files.
- Refactored email rendering from `renderAsync` to `render` in notification utilities and email templates for consistency.
- Adjusted Tailwind configuration usage in email templates to utilize a centralized configuration file.
These changes enhance the overall stability and maintainability of the codebase.
* refactor: update badge variants across dashboard components
- Changed badge variant from "outline-solid" to "outline" in multiple components including columns, show-domains, and various deployment tables for consistency in styling.
- Updated button variants in billing and project templates to align with the new badge styling.
- Enhanced input components to support password generation and error messaging, improving user experience.
These changes streamline the UI and ensure a cohesive design across the application.
* refactor: clean up calendar component imports and structure
- Removed redundant imports of icons from lucide-react and streamlined the import statements for better readability.
- Adjusted the placement of type imports to enhance code organization.
These changes improve the maintainability and clarity of the calendar component.
* [autofix.ci] apply automated fixes
---------
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
The readLogs endpoint only checked deployment.serverId and
deployment.schedule?.serverId to determine where to read log files.
For application and compose deployments on remote servers, serverId
is not stored on the deployment record — the server is resolved from
the parent entity (application.serverId, compose.serverId).
This caused readLogs to fall through to local execAsync, which would
silently fail (2>/dev/null) because the log file lives on the remote
server, returning empty content.
Fix by loading the compose relation in findDeploymentById and adding
fallback resolution through application?.serverId and compose?.serverId.
Fixes#4687
Co-authored-by: Roo <roo@agent.com>
* feat: add builds concurrency management for servers
- Introduced a new `BuildsConcurrency` component to manage the number of concurrent builds for both local and remote servers, gated by license validity.
- Implemented backend logic to resolve effective builds concurrency based on server settings and organization licenses.
- Added unit tests for concurrency resolution logic to ensure correct behavior under various licensing scenarios.
- Updated database schema to include `buildsConcurrency` field for servers and web server settings.
- Refactored deployment queue to support in-memory job processing with configurable concurrency limits.
This feature enhances deployment flexibility and control for enterprise users.
* refactor: enhance deployment cancellation logic and improve Railpack build isolation
- Reintroduced the `initCancelDeployments` function in the server initialization sequence to ensure deployments can be canceled effectively.
- Updated the Railpack build command to use a unique builder name for each build, preventing conflicts during concurrent deployments.
- Enhanced the cancellation logic to reset application and compose statuses to "idle" after canceling running deployments, improving system reliability.
* test: add buildsConcurrency setting to server configuration tests
- Introduced a new `buildsConcurrency` property in the server configuration tests to ensure proper handling of concurrent builds in deployment scenarios.
* feat: implement builds concurrency management and validation
- Added `assertBuildsConcurrencyAllowed` function to validate concurrency settings based on license status.
- Updated `resolveBuildsConcurrency` to reflect new concurrency limits for free and enterprise tiers.
- Enhanced `BuildsConcurrency` component to manage concurrent builds for servers, with UI adjustments for better user experience.
- Introduced a new settings page for managing concurrent builds across servers, ensuring proper handling of deployments.
- Updated database schema to support increased maximum concurrency values for servers and web server settings.
* fix: use canEditDeployGitSource for git provider access on existing deploys
Replaces the simple userId ownership check with a new canEditDeployGitSource
function that correctly handles all role/sharing scenarios. Owner always has
access; admin and member only if they own the provider or it is shared with
the org — being assigned via accessedGitProviders (enterprise) only grants
permission to connect new deploys, not to edit the git source of existing ones.
Adds 26 unit tests covering owner, admin, member (with/without enterprise
license), shared providers, and the key regression case from issue #4469.
* [autofix.ci] apply automated fixes
---------
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
* refactor: improve restore logging for database backups
- Updated restore functions across various database types (Postgres, MySQL, MongoDB, MariaDB, LibSQL, and Compose) to provide clearer logging messages.
- Replaced generic command execution logs with specific messages indicating the database being restored and the source backup file.
- This change enhances the clarity of restore operations and aids in troubleshooting by providing more context in the logs.
* [autofix.ci] apply automated fixes
---------
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
* fix: strip credentials from service-level API responses
Registry passwords and S3 destination credentials were being returned
in service `.one` tRPC endpoints to any user with service-level read
access. Reported by Nihon Kohden Corporation security team.
- Strip registry `password` from `findApplicationById` via Drizzle `columns: { password: false }`
- Strip destination `accessKey`/`secretAccessKey` from all DB service finders (postgres, mysql, mariadb, mongo, libsql, compose, backup, volume-backups)
- Add `findRegistryByIdWithCredentials` for internal use only
- Builders and upload utils now load registry credentials by ID at execution time
- `createRollback` enriches `fullContext` with registry credentials before persisting to DB so rollback execution has what it needs
- Remove `findApplicationByIdWithCredentials` and `ApplicationNestedWithCredentials` — no longer needed
- Backup execution utils load full destination via `findDestinationById` at runtime instead of reading from the joined relation
* [autofix.ci] apply automated fixes
---------
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
Docker Swarm's default start-first update order causes new database
containers to fail with 'DBPathInUse' because two containers compete
for the same data volume simultaneously. Docker then rolls back the
update, silently reverting any env var or config changes.
Using stop-first ensures the old container is stopped before the new
one starts, preventing volume lock conflicts across all database types.
Fixes#4550
- Enhanced the `apiReadTraefikConfig` schema by reintroducing path validation logic to prevent directory traversal attacks and unauthorized access.
- The validation now includes checks for null bytes and ensures paths start with a defined main Traefik path, improving security and robustness.
These changes strengthen the integrity of the configuration handling by ensuring only valid paths are accepted.
- Introduced a new SQL migration file `0171_lucky_echo.sql` to modify the foreign key constraint on the `sso_provider` table, changing the `ON DELETE` behavior from `cascade` to `set null`.
- Updated the journal to include the new migration version and its associated tag.
- Added a snapshot file for version 7 of the database schema, reflecting the current state of the `sso_provider` and other related tables.
These changes enhance the integrity of the database by ensuring that user references are set to null instead of being deleted when the referenced user is removed.
- Updated the alert block in the HandleForwardAuth component to provide clearer requirements for deploying the authentication proxy.
- Added a DnsHelperModal to assist with DNS configuration in the ForwardAuthServers component.
- Refined API input schemas for forward authentication operations to improve type safety and clarity.
- Removed the obsolete forward-auth SSO design document to streamline documentation.
These changes improve the user experience and maintainability of the forward authentication feature across the application.
- Replaced `forwardAuthProviderId` with `forwardAuthEnabled` in the domain schema to simplify the configuration of forward authentication.
- Updated related tests to reflect this change, ensuring consistency across the application.
- Introduced a new SQL migration to create the `forward_auth_settings` table for managing authentication domains and their configurations.
This refactor enhances the clarity and maintainability of the forward authentication logic within the application.
- Added the `VALID_BRANCH_REGEX` import to all Git provider components to ensure consistent branch validation.
- Removed duplicate imports of `VALID_BRANCH_REGEX` to streamline the code and improve readability.
This change enhances maintainability by centralizing branch validation logic across the application.
- Removed the selection of SSO providers from the UI, streamlining the process to enable/disable SSO for domains.
- Updated the API to eliminate the need for a provider ID when enabling forward authentication, relying on the configured settings instead.
- Enhanced user feedback by updating toast messages to reflect the current state of SSO authentication.
- Improved the UI layout for better clarity on SSO status and actions.
This refactor enhances the user experience by simplifying the SSO configuration process and ensuring clearer communication of actions taken.
