- Fixed handleMouseOut to properly detect when leaving tooltip triggers
- Added document mouseleave handler to hide tooltips when mouse leaves viewport
- Added scroll event handler to prevent tooltips getting stuck during scrolling
- Improved event handling to prevent premature tooltip hiding from child element bubbling
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Change config loader to set name to null instead of host address when no custom name is provided
- This ensures discovered node names are used by default
- Custom names are only applied when PROXMOX_NODE_NAME is explicitly set
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add support for custom display names via PROXMOX_NODE_NAME configuration
- Handle multi-node clusters properly by prefixing node names with endpoint name
- Fix dashboard and nodes views to show configured display names
- Fix non-sequential endpoint numbering issue in settings (node_2, node_8, node_14)
- Improve endpoint configuration loading to handle any index number
Fixes#100🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Removed diagnostics icon button from header
- Removed JavaScript code that managed icon visibility
- Removed event listener for diagnostics icon click
- Diagnostics functionality remains available in settings menu
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Backup .env file before removing old installation
- Restore .env file after extracting new version
- Fix both update and migration processes
- Ensure proper ownership and permissions on restored .env file
- Store release data when checking for updates
- Use server API for update checks to get Docker status
- Implement proper applyUpdate function with progress tracking
- Handle Docker deployments with appropriate warning message
- Add WebSocket listeners for update progress events
- Strip http:// or https:// from PBS_HOST values in settings form
- Update placeholders to show IP/hostname format without protocol
- Fix discovery cycle to trigger for PBS-only configurations
- Resolves issue where PBS data wouldn't load due to malformed host URL
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Previously, the config loader required sequential numbering (2, 3, 4...) for additional Proxmox/PBS endpoints. If a user had PROXMOX_HOST_2 and PROXMOX_HOST_4 (skipping 3), only endpoint 2 would be loaded.
This fix scans all environment variables to find any PROXMOX_HOST_N or PBS_HOST_N patterns, regardless of numbering sequence. Now users can have endpoints numbered 2, 5, 10, etc. and all will be properly loaded.
Fixes#96🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added SHA-based version tracking to avoid GitHub CDN caching issues
- Uses GitHub API to check latest commit SHA for the installer
- Falls back to simple diff check if jq is not available
- Updates embedded SHA when downloading new version
- Much more reliable than simple file comparison
This restores the update method from the original installer that
properly handles GitHub's aggressive CDN caching.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Check for Express version mismatch after extracting tarball
- Reinstall dependencies if Express version is wrong
- Fixes service startup failure due to path-to-regexp error
- Works around issue where GitHub releases have wrong Express version
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added back the self-update check for the installer script
- Installer now checks for updates to itself before running
- Skips update check when running from pipe (curl | bash)
- User is prompted before updating the installer
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Removed npm dependency installation from installer (tarballs include deps)
- Cleaned up output with better icons and formatting
- Simplified installation flow since tarballs are pre-built
- Reduced installer from 590 to 522 lines
- Fixed messaging to be cleaner and more professional
The installer now expects properly built tarballs from create-release.sh
that include all dependencies and pre-built CSS.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Install all dependencies (including dev) initially
- Build CSS with tailwindcss available
- Remove dev dependencies after CSS is built
- Fixes installation failure due to missing tailwindcss
- Check if CSS is already built before attempting to rebuild
- Skip CSS build if tailwindcss is not available (dev dependency)
- Continue installation even if CSS build fails (tarball should have pre-built CSS)
- Fix issue where installation would fail due to missing tailwindcss
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Set should_update_deps=true when Express version mismatch is found
- Previously, the installer would detect the mismatch but still skip npm install
- This was causing the path-to-regexp error to persist even after detection
The installer will now properly update dependencies when it detects
that Express version is not 4.19.2.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Remove 'local' keyword usage outside of functions
- Variables should_update_deps, current_express_ver, and express_version are now global
- Fixes "./install-pulse.sh: line 2204: local: can only be used in a function" error
This was preventing the dependency update logic from running properly.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Changed installer to always run npm install on updates, not just fresh installs
- Added Express version check even for tarball installations
- Forces dependency update if Express version doesn't match expected 4.19.2
- Removed manual fix script as it's no longer needed
The issue was that tarball installations include pre-built node_modules,
which prevented dependency updates when package.json changed.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add fix-express-version.sh script to manually resolve path-to-regexp errors
- Script performs complete node_modules cleanup and reinstall
- Verifies correct Express 4.19.2 version is installed
- Provides immediate solution while installer is being debugged
Run this script on affected servers:
sudo /opt/pulse/scripts/fix-express-version.sh
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Stop the service before updating dependencies
- Remove node_modules directory to force clean install
- Add Express version verification after install
- This ensures old Express 4.21.x dependencies are completely removed
The service recovery was masking the real issue - old node_modules
were not being updated properly, keeping the problematic Express version.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Switch from npm install to npm ci for clean installs based on package-lock.json
- Add fallback to npm install if npm ci fails
- Remove --unsafe-perm flag and use --omit=dev for production installs
- Show dependency installation output (filtering npm WARN messages)
This ensures that package-lock.json changes are properly applied during updates,
fixing issues where old dependencies remain after updates.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Downgrade Express from ^4.21.2 to 4.19.2 to avoid path-to-regexp compatibility issues
- Express 4.21.x includes a newer version of path-to-regexp with breaking changes
- Routes with multiple parameters (e.g., /api/thresholds/:endpointId/:nodeId/:vmid) were causing startup failures
- This resolves the "pulse-monitor.service: Failed with result 'exit-code'" error
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Replace overly broad npm notice detection with specific checks
- Add detection for actual npm start misconfiguration in ExecStart
- Separate npm errors/warnings from service configuration issues
- Provide clearer troubleshooting guidance for each error type
This prevents false positives when npm notices appear in logs
during normal operation with direct node execution.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Replace npm execution with direct node server/index.js in systemd service to eliminate npm-related startup failures and update notices. Add improved diagnostics to detect and explain npm execution issues.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Add clearer messaging when service recovery succeeds to provide better user feedback during installation process.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Runtime alert acknowledgment data should not be tracked in version control.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add validation to prevent API calls with missing nodeId/vmid parameters
- Implement proper guest selector initialization for edit modal
- Add server-side error handlers to return JSON instead of HTML for API routes
- Include debug logging for threshold configuration troubleshooting
- Ensure dropdown selection properly populates all required hidden fields
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
• Merge separate "Alerts" and "Custom Thresholds" tabs into single "Alert Configuration" tab
• Improve UX by organizing all alert settings in one logical location
• Add comprehensive custom per-VM/LXC threshold management system with migration support
• Include blue "T" badges for VMs with custom thresholds in dashboard
• Implement full CRUD operations for threshold configurations via REST API
• Support migration-aware threshold keys (endpointId:vmid) for cluster environments
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Previously, configuration created via the web UI was lost when containers were
recreated or updated, requiring users to reconfigure each time. This was due to
the .env file being stored inside the container filesystem.
Changes:
- Add persistent volume mount for configuration directory in docker-compose.yml
- Update ConfigApi to auto-detect Docker vs development environment
- Configure dotenv loading to use persistent config path when available
- Update file watcher to monitor correct .env file location
- Remove conflicting host .env file mount
- Update README with new volume configuration instructions
Configuration now persists across container updates, eliminating the need for
users to reconfigure after updates.
Fixes#94🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Previously, adding additional PVE credentials would remove all existing PBS
credentials from the .env file, and vice versa. This was caused by the credential
removal logic not checking if the new configuration actually contained endpoints
of the same type before removing existing ones.
The fix ensures that:
- PBS credentials are preserved when adding additional PVE endpoints
- PVE credentials are preserved when adding additional PBS endpoints
- Existing endpoints are only removed when explicitly managing that endpoint type
Fixes#91🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Replace loose equality/inequality operators with strict comparisons and proper type conversion for guest vmid matching:
- Use parseInt() with strict equality for all guest ID comparisons
- Prevents type coercion bugs between string and number vmids
- Ensures consistent behavior across different data sources
- Improves reliability in multi-cluster environments
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Fixed guest lookup logic to properly handle guests with the same vmid on different nodes by implementing unique guest identification throughout the backup data pipeline:
- Updated guest filtering to use vmid-node combinations instead of simple vmid matching
- Enhanced PBS vs PVE backup system awareness for node-specific vs centralized filtering
- Fixed calendar heatmap data structures to prevent guest data mixing
- Improved calendar date filtering to use unique guest identifiers
- Added node-aware filtering for backup tasks and snapshot detection
This resolves issues where the filtered summary and calendar would incorrectly display or mix data from guests sharing the same vmid across different nodes.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Apply the same sticky header pattern used in dashboard to fix column header
positioning in storage table. Individual header cells are now sticky instead
of relying on row-level sticky positioning.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Changes EnvironmentFile to EnvironmentFile=- in systemd service template, allowing service to start even when .env file doesn't exist during fresh installations.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Prevents "fatal: not a git repository" error when install script falls back to git update on non-git directories.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
courtmanr@gmail.com