Bug fixes:
- Backup refresh no longer re-renders entire page (separate refreshing state)
- SSL cert button no longer flickers when no certs available
- Volume mode selector rewritten to use proper scope system (7 scopes)
- Navigation flicker eliminated when returning from env/volumes pages
- Logout button moved to sidebar footer near theme/locale controls
- Subdomain pattern now shows variable hint tooltip ({project}, {stage}, etc.)
- SSL certificate selector moved to Credentials page with auto-save
- Projects page now has search/filter by name, image, or registry
Consistency improvements:
- New Breadcrumb component replaces 5 inline implementations
- New IconArrowLeft, IconChevronDown components replace inline SVGs
- All inline spinners replaced with IconLoader component
- 10 semantic badge classes with dark mode variants in tokens.css
- Global disabled button cursor-not-allowed rule
- Raw inputs in auth page replaced with FormField components
- Missing aria-labels added to icon-only buttons
- Error panels standardized to use design tokens
- Add proxy provider selector (None/NPM) to settings page with radio cards
- Show warning when switching to None about existing NPM routes
- Conditionally hide SSL certificate picker and NPM credentials when provider is None
- Fix webhook URL regenerate not updating UI (field name mismatch: url vs webhook_url)
- Fix dev-server.sh to persist ENCRYPTION_KEY across restarts via data/.dev-key
- Fix selected radio card background visibility in dark mode
Replace direct npm.Client usage throughout the codebase with the
proxy.Provider interface, enabling pluggable proxy backends. The
deployer, API layer, and proxy manager now use provider-agnostic
route management (ConfigureRoute/DeleteRoute) instead of NPM-specific
API calls. Adds ProxyRouteID (string) to Instance model and
ProxyProvider setting to Settings, with SQLite migrations for
backward compatibility.
- Detect existing projects with same image on quick deploy; show conflict dialog with options
- Move logout button to sidebar header as icon-only
- Replace backup checkbox with ToggleSwitch component
- Allow unsafe-inline in CSP script-src for SvelteKit hydration
- Guard SSE connection behind isAuthenticated() check
- Add notification_url ALTER TABLE migration for existing databases
- Restore RegisterPersistentLogger on event bus
- Expand health endpoint to check DB, Docker, and NPM connectivity (FUNC-M4)
- Add project_id, stage_id, offset query params to deploys endpoint (FUNC-M5, FUNC-M6)
- Add notification_url field to Stage model for per-project overrides (FUNC-M2)
- Add NPM Ping method for health checking
- Sanitize all internal error messages in API handlers (SEC-M4)
- Add audit trail events for admin actions (FUNC-M3)
- Add EventLog event type to event bus
- Add SSE connection status banner showing when real-time updates are lost (UX-H8, UX-M1)
- Add password visibility toggle on login page (UX-H10)
- Add dark mode variants to stat card backgrounds (UX-M11)
- Add overflow-x-auto to tables for mobile responsiveness (UX-H9)
- Add flex-wrap to stage header for mobile overflow (UX-H11)
- Fix theme store system preference listener reactivity (UX-M12)
- Parallelize registry health checks (UX-L4)
Add backup/restore functionality for the SQLite database. Users can
trigger manual backups, configure automatic backups on an interval with
retention policies, list/download/delete backups, and restore from any
backup.
- Backup engine using VACUUM INTO (safe with WAL mode), mutex-protected
- Autobackup via cron with dynamic rescheduling on settings change
- Retention: prune on startup and after each backup, orphan cleanup
- Restore: pre-restore safety backup, streaming copy, graceful shutdown
- Settings UI at /settings/backup with full backup list management
- HIGH: Add sync.Mutex to backup Engine to prevent concurrent
backup/restore operations
- HIGH: Restore uses io.Copy instead of ReadFile to avoid OOM on
large databases
- HIGH: Send HTTP response before closing DB during restore, then
perform destructive operations in a goroutine
- HIGH: Create pre-restore safety backup before overwriting database
- HIGH: Autobackup cron reschedules dynamically when settings change
via callback pattern (same as DNS provider changes)
Add backup/restore functionality for the SQLite database. Users can
trigger manual backups, configure automatic backups on an interval
with retention policies, list/download/delete backups, and restore
from any backup.
- Backup engine using VACUUM INTO (safe with WAL mode)
- Backup metadata tracked in DB, files stored in DATA_DIR/backups/
- Settings: backup_enabled, backup_interval_hours, backup_retention_count
- API: POST/GET/DELETE /api/backups, download, restore endpoints
- Autobackup via cron scheduler with configurable interval
- Retention: prune on startup, after each backup (manual and auto)
- Orphan cleanup: removes backup files without metadata on startup
- Restore: replaces DB and triggers graceful server shutdown
- Settings UI: /settings/backup with toggle, interval, retention config
- Backup list with download, delete, restore actions
- i18n: English and Russian translations
Introduces flexible DNS management. Wildcard DNS is the default (current
behavior). When disabled, users configure a Cloudflare DNS provider and
DNS A records are automatically kept in sync with proxy consumers.
- Provider interface (internal/dns/) with Cloudflare API v4 implementation
- Settings: wildcard_dns toggle, dns_provider, encrypted API token, zone ID
- DNS lifecycle hooks in deployer and proxy manager (best-effort)
- Settings UI: DNS config section with provider picker, zone selector, test
- DNS Records page at /dns with filtering, sync status, reconciliation
- Records visible in both wildcard (informational) and managed modes
- Thread-safe provider access with sync.RWMutex
- Cleanup on provider change: removes old records when switching modes
- CRITICAL: Add binaries and .svelte-kit/ to .gitignore, remove from tracking
- HIGH: Return error from computeExpectedFQDNs to prevent mass DNS
deletion on transient DB errors during sync
- MEDIUM: Log error in rollback DNS cleanup when GetSettings fails
- CRITICAL: Change DNS zones endpoint from GET to POST to avoid
leaking API token in URL query parameters
- HIGH: Add sync.RWMutex to protect dnsProvider field in Server,
Deployer, and proxy Manager against concurrent read/write races
- HIGH: Capture old DNS provider reference synchronously before
launching background cleanup goroutine
- HIGH: Use getDNS()/getDNSProviderLocked() accessors instead of
direct field reads in all DNS operations
Add flexible DNS management to Docker Watcher. By default, wildcard DNS
is assumed (current behavior). When disabled, users can configure a
Cloudflare DNS provider with API token and zone selection. DNS A records
are automatically created/updated/deleted in sync with proxy consumers
(deployed instances and standalone proxies).
- Settings: wildcard_dns toggle, dns_provider, cloudflare credentials
- Cloudflare client: Provider interface with EnsureRecord/DeleteRecord/ListRecords
- DNS lifecycle hooks in deployer and proxy manager (best-effort)
- Settings UI: DNS config section with provider picker, zone selector, test button
- DNS Records page at /dns with filtering, sync status, reconciliation
- Records visible in both wildcard and managed modes
- Cleanup on provider change: removes old records when switching modes
- Add 'absolute' volume scope for direct host paths (NFS, external mounts)
- Allowlist in settings: allowed_volume_paths (JSON array of prefixes)
- Validation: absolute source must be under an allowed prefix
- Empty allowlist = absolute scope disabled entirely
- Settings API exposes/validates allowed_volume_paths
- Frontend type updated with absolute scope
Critical fixes:
- IDOR: verify volume belongs to project before resolving path
- Upload: override global 1MB body limit for upload endpoint (100MB)
High-priority fixes:
- Symlink escape: use filepath.EvalSymlinks in safePath validation
- Remove host filesystem path from browse API response
- Sanitize Content-Disposition filenames, force application/octet-stream
- Strip directory components from upload filenames
- Browse and Download buttons on each volume row in the editor table
- Download entire volume as ZIP directly from the editor (no browser needed)
- File type icons for common extensions in browser
- Ephemeral volumes excluded from browse/download actions
- CRITICAL: validate volume Name against path traversal (safe regex)
- HIGH: log data migration errors instead of silently ignoring
- HIGH: reject empty source when switching from ephemeral scope
Replace confusing shared/isolated volume modes with explicit scopes:
- instance: per-deploy isolated directory
- stage: shared within a stage across deploys
- project: shared across all stages
- project_named: named group within a project
- named: global named volume across projects
- ephemeral: tmpfs in-memory mount
Includes schema migration (shared→project, isolated→instance),
backward-compatible deployer resolution, scope metadata API endpoint,
and redesigned volume editor UI with scope guide cards and hints.
- Classify Docker errors into categories (socket_not_found, connection_refused,
permission_denied, timeout, tls_error) with platform-specific hints
- Enrich GET /api/health with structured diagnostics (category, hints, platform)
- Expandable hints panel in sidebar when Docker is disconnected
- "Retry now" button for immediate re-check
- Collapsible raw error details for advanced users
Use $effect instead of onMount to start SSE and health polling,
so they activate on client-side navigation after login without
requiring a full page reload.
- GET /api/health endpoint returning Docker connectivity status
- Sidebar shows Docker connection dot (green=connected, red=disconnected)
- Stale scanner returns store-only results when Docker is unavailable
- Polls health every 30s
Stale container detection, standalone proxy management with validation
and health monitoring, unified proxy viewer, persistent event log,
container stats, webhook notifications, and dashboard enhancements.
- Fix chi duplicate Route() panic by consolidating read/write routes
into single Route blocks with nested admin Group
- Add logout button to sidebar with token cleanup
- Add missing settingsAuth.password i18n key
Critical fixes:
- Fix StaleContainer frontend type to match nested backend response shape
- Guard ContainerID[:12] slice against empty/short IDs in ListAllProxies
High-priority fixes:
- Support comma-separated severity/source in event log filtering (IN clause)
- Eliminate N+1 queries in ListAllProxies and FindStaleInstances (pre-load maps)
- Stop leaking internal error messages to API clients (use slog + generic msgs)
Add all frontend pages for observability & proxy management:
- Proxy Viewer: /proxies with grouped view, filtering, health indicators
- Proxy Creation: form with live validation, diagnostic hints, edit/delete
- Stale Containers: /containers/stale with dashboard widget, cleanup actions
- Event Log: /events with filters, pagination, real-time SSE streaming
- Navigation: proxies and events links in sidebar
- i18n: full EN/RU translations for all new features
- Settings: stale threshold configuration
Add standalone proxy management:
- Multi-step validation pipeline (DNS, TCP, HTTP) with diagnostic hints
- Proxy lifecycle: create/update/delete via NPM API with SSL auto-assign
- Periodic health monitoring (5min) with event log on status transitions
- Unified /api/proxies/all endpoint merging standalone + managed proxies
- Frontend types and API functions for downstream UI phases
Add database foundation for observability features:
- event_log table with severity/source filtering and pagination
- standalone_proxies table for user-created reverse proxies
- stale_threshold_days setting (default 7 days)
- Auto-persist warn/error events from event bus to database
- SSE broadcast of persistent events for real-time UI updates
- Frontend types and API functions for downstream UI phases
When the SSL certificate is changed in settings, automatically
updates all existing NPM proxy hosts managed by Docker Watcher
in the background. Clears SSL if cert is removed.
Add enable_proxy boolean to stages (default true). When disabled,
the deployer skips NPM proxy host creation — useful for internal
services, workers, or externally-routed containers. UI shows
toggle in Add Stage form and "No Proxy" badge on stage header.
alexei.dolgolyov