Phase 1 of Pulse AI differentiation:
- Create internal/ai/context package with types, trends, builder, formatter
- Implement linear regression for trend computation (growing/declining/stable/volatile)
- Add storage capacity predictions (predicts days until 90% and 100%)
- Wire MetricsHistory from monitor to patrol service
- Update patrol to use buildEnrichedContext instead of basic summary
- Update patrol prompt to reference trend indicators and predictions
This gives the AI awareness of historical patterns, enabling it to:
- Identify resources with concerning growth rates
- Predict capacity exhaustion before it happens
- Distinguish between stable high usage vs growing problems
- Provide more actionable, time-aware insights
All tests passing. Falls back to basic summary if metrics history unavailable.
- Changed patrol schedule from preset dropdown to freeform number input
- Users can now set any interval (min 10 minutes, max 7 days, or 0 to disable)
- Added patrol_interval_minutes to API request/response (preset is now deprecated)
- Backend validates: min 10 minutes when enabled, max 10080 (7 days)
- Frontend shows human-readable duration next to input (e.g., '6h', '2h 30m')
Also improved Auto-Fix Mode safety:
- Removed '(recommended)' from preset options (was subjective)
- Added 'I understand the risks' acknowledgement checkbox
- Toggle is disabled until user explicitly acknowledges the risks
- Shows prominent warning when Auto-Fix is enabled
- Acknowledgement is session-based (must re-acknowledge on page reload)
- Add 'content' type to StreamDisplayEvent for tracking text chunks
- Track content events in streamEvents array for chronological display
- Update render to use Switch/Match for cleaner conditional rendering
- Interleave thinking, tool calls, and content as they stream in
- Add fallback for old messages without streamEvents for backwards compat
Previously, tool/command outputs stayed at top while AI text responses
accumulated at the bottom. Now all events appear in order like a
normal chatbot.
- Add clear_anthropic_key, clear_openai_key, clear_deepseek_key, clear_ollama_url flags to API
- Backend handles clearing with confirmation prompt
- Each provider accordion shows Test and Clear buttons when configured
- Clear button requires confirmation before removing credentials
- Frontend automatically refreshes settings after clearing
- Add /api/ai/test/{provider} endpoint for testing individual providers
- Add 'Test' button to each provider accordion (visible when configured)
- Shows test result inline (success/error message)
- Update help links with direct URLs to API key pages:
- Anthropic: console.anthropic.com/settings/keys
- OpenAI: platform.openai.com/api-keys
- DeepSeek: platform.deepseek.com/api_keys
- Ollama: ollama.ai
Backend:
- Add per-provider API key fields to AIConfig (AnthropicAPIKey, OpenAIAPIKey, DeepSeekAPIKey, OllamaBaseURL, OpenAIBaseURL)
- Add NewForProvider() and NewForModel() factory functions for multi-provider instantiation
- Update ListModels() to aggregate models from all configured providers with provider:model format
- Update Execute/ExecuteStream to dynamically create provider based on selected model
- Update TestConnection to use multi-provider aware provider creation
- Add helper functions: HasProvider(), GetConfiguredProviders(), GetAPIKeyForProvider(), GetBaseURLForProvider(), ParseModelString(), FormatModelString()
Frontend:
- Remove legacy single-provider UI (provider grid, single API key input, single base URL)
- Add accordion-style UI for configuring all providers independently
- Add model grouping by provider in selectors using optgroup
- Update AIChat model dropdown with grouped provider sections
- Add helper functions for parsing provider from model ID and grouping models
API:
- Add multi-provider fields to AISettingsResponse and AISettingsUpdateRequest
- Add /api/ai/models endpoint for dynamic model listing
- Update settings handlers for per-provider credential management
Users can now:
1. View all suppression rules (both from dismissed findings and manually created)
2. Create manual rules like 'ignore performance issues on debian-go'
3. Delete rules when they want alerts to come back
Backend:
- Added SuppressionRule type for user-defined rules
- Added suppressionRules storage to FindingsStore
- Added AddSuppressionRule/GetSuppressionRules/DeleteSuppressionRule methods
- Added isSuppressedInternal check for manual rules
- Added API handlers and routes for /api/ai/patrol/suppressions
Frontend:
- Added SuppressionRule interface
- Added getSuppressionRules/addSuppressionRule/deleteSuppressionRule API functions
- Added getDismissedFindings for viewing dismissed findings
Example usage:
POST /api/ai/patrol/suppressions
{
'resource_id': 'debian-go',
'category': 'performance',
'description': 'Dev container runs hot - expected'
}
The main issue was that finding IDs included the title, which the LLM
generates differently each time. 'High CPU on minipc' vs 'Node minipc
experiencing high CPU load' got different IDs, making dismissals useless.
Changes:
1. LLM findings now get IDs based on resource+category only, not title
2. Add() now checks if finding is suppressed before adding as new
3. Add() now checks dismissed findings and only reactivates on severity escalation
4. IsSuppressed() now matches by resource+category only, not title
5. Added isSuppressedInternal() for use when lock is already held
Now when you dismiss 'performance issues on minipc', any future patrol finding
about performance on minipc will be recognized as the same issue and stay dismissed.
The ForcePatrol() function was using the HTTP request context, which gets
cancelled immediately when the API response is sent. This caused LLM analysis
to fail with 'context canceled' before it could complete.
Now uses context.Background() so the goroutine runs independently of the
HTTP request lifecycle.
Also fixed dropdown hover gap issue in the dismiss menu.
Knowledge store notes are now included in the patrol LLM prompt. When users
save notes about resources (e.g., 'This VM intentionally runs hot'), the patrol
AI will see these notes and avoid flagging documented behavior as issues.
Changes:
- Added knowledge store reference to PatrolService
- Added SetKnowledgeStore() method to configure the store
- Enhanced buildPatrolPrompt() to include knowledge context
- Connected knowledge store to patrol in service.go SetStateProvider()
This complements the dismissed findings context to give the LLM a complete
picture of what the user considers normal/expected behavior.
Implements a comprehensive feedback system that allows the LLM to 'remember'
user decisions about findings, preventing repetitive/annoying alerts.
Backend changes:
- Extended Finding struct with dismissed_reason, user_note, times_raised, suppressed
- Added Dismiss(), Suppress(), SetUserNote(), IsSuppressed() methods to FindingsStore
- Added GetDismissedForContext() to format dismissed findings for LLM context
- Enhanced buildPatrolPrompt() to inject user feedback context
- Added POST /api/ai/patrol/dismiss and /api/ai/patrol/suppress endpoints
- Updated IsActive() to exclude suppressed findings
Frontend changes:
- Added Dismiss dropdown with options: Not an Issue, Expected Behavior, Will Fix Later
- Added Never Alert Again option for permanent suppression
- Expected Behavior prompts for optional note to help LLM understand context
- Added visual badges: recurrence count (×N), dismissed status, suppressed indicator
- Display user notes in expanded finding view
Also fixes:
- Fixed 403 error on Run Patrol (compilation errors from partial refactoring)
- Removed non-LLM patrol checks - patrol now uses LLM analysis only
- Fixed function signature mismatches in alert_triggered.go
The LLM now receives context about previously dismissed findings and is
instructed not to re-raise them unless severity has significantly worsened.
- Add alert-triggered AI analysis for real-time incident response
- Implement patrol history persistence across restarts
- Add patrol schedule configuration UI in AI Settings
- Enhance AIChat with patrol status and manual trigger controls
- Add resource store improvements for AI context building
- Expand Alerts page with AI-powered analysis integration
- Add Vite proxy config for AI API endpoints
- Support both Anthropic and OpenAI providers with streaming
Keep only the simple AI-powered approach:
- set_resource_url tool lets AI save discovered URLs
- Users ask AI directly: 'Find URLs for my containers'
- AI uses its intelligence to discover and set URLs
Removed:
- URLDiscoveryService (rigid port scanning)
- Bulk discovery API endpoints
- Frontend discovery button
The AI itself is smart enough to iterate through resources
and discover URLs when asked.
- Add URLDiscoveryService for scanning all resources at once
- Scans common web ports (80, 443, 8080, 8096, 3000, etc.)
- Automatically saves discovered URLs to resource metadata
- Add API endpoints for start/status/cancel discovery
- Progress tracking with results reporting
Endpoints:
- POST /api/ai/discover-urls/start - Start bulk discovery
- GET /api/ai/discover-urls/status - Check progress
- POST /api/ai/discover-urls/cancel - Cancel discovery
- Add MetadataProvider interface for AI to update resource URLs
- Add set_resource_url tool to AI service
- Wire up metadata stores to AI service via router
- Add URL discovery guidance to AI system prompt
- AI can now inspect guests/containers/hosts for web services
and automatically save discovered URLs to Pulse metadata
Usage: Ask the AI 'Find the web URL for this container' and it will:
1. Check for listening ports and web servers
2. Get the IP address
3. Verify the URL works
4. Save it to Pulse for quick dashboard access
- Add host metadata API for custom URL editing on hosts page
- Enhance AI routing with unified resource provider lookup
- Add encryption key watcher script for debugging key issues
- Improve AI service with better command timeout handling
- Update dev environment workflow with key monitoring docs
- Fix resource store deduplication logic
- Add Claude OAuth authentication support with hybrid API key/OAuth flow
- Implement Docker container historical metrics in backend and charts API
- Add CEPH cluster data collection and new Ceph page
- Enhance RAID status display with detailed tooltips and visual indicators
- Fix host deduplication logic with Docker bridge IP filtering
- Fix NVMe temperature collection in host agent
- Add comprehensive test coverage for new features
- Improve frontend sparklines and metrics history handling
- Fix navigation issues and frontend reload loops
The AI service now uses only buildUnifiedResourceContext() for
infrastructure context, since the resourceProvider is always set
during router initialization.
Removed:
- buildInfrastructureContext() function (~288 lines of dead code)
- Legacy fallback path in buildSystemPrompt()
The unified resource context provides a cleaner, deduplicated view
of infrastructure that includes:
- All resources grouped by platform and type
- Top CPU/Memory/Disk consumers
- Active alerts on resources
- Infrastructure summary statistics
This completes the AI service migration to unified resources.
This cleanup addresses transition debt from the unified resources migration:
Frontend cleanup:
- Move all Resource→Legacy type conversions to useResourcesAsLegacy() hook
- Add asNodes() and asDockerHosts() adapter functions to the hook
- Simplify DockerRoute, HostsRoute, DashboardView to use the centralized hook
- Remove ~300 lines of duplicated adapter code from App.tsx
- Remove debug console.log statements from Dashboard.tsx
- Fix CPU value conversion (divide by 100) for Dashboard compatibility
Backend fixes (from previous session):
- Fix parentID format in converters (VM, Container, Storage) to match Node.ID
- Format changed from 'instance/node/nodename' to 'instance-nodename'
- Update tests to match new parentID format
This consolidates all legacy type conversion logic in one place,
making future cleanup easier when components are migrated to use
unified resources directly.
Backend:
- Call SetMonitor after router creation to inject resource store
- Add debug logging for resource population and broadcast
Frontend:
- Add resources array to WebSocket store initial state
- Handle resources in WebSocket message processing
- Use reconcile for efficient state updates
The unified resources are now properly:
1. Populated from StateSnapshot on each broadcast cycle
2. Converted to frontend format (ResourceFrontend)
3. Included in WebSocket state messages
4. Received and stored in frontend state
5. Consumed by migrated route components
Console now shows '[DashboardView] Using unified resources: VMs: X'
confirming the migration is working end-to-end.
- Added PopulateFromSnapshot method to resources.Store
- Extended ResourceStoreInterface to include PopulateFromSnapshot
- Monitor now calls updateResourceStore before broadcasts
- This ensures resources are fresh on every WebSocket broadcast
Without this, the store would only be populated when /api/resources or
/api/state endpoints are hit, leaving WebSocket broadcasts empty.
- Extended StateFrontend with Resources field containing unified resource data
- Added ResourceFrontend and related types for frontend-compatible resource data
- Extended ResourceStoreInterface to include GetAll() method
- Monitor now injects resources into WebSocket broadcasts
- Added helper method getResourcesForBroadcast() to convert resources to frontend format
- All existing tests pass
This enables the frontend to access unified resources via WebSocket state.
- Removed /resources page and associated frontend components
- Removed ResourcesOverview.tsx, UnifiedResourceRow.tsx, columns.ts
- Removed frontend types/resource.ts
- Updated unified-resource-architecture.md to mark Phase 4 as ABANDONED
- Removed unified-view-migration-plan.md
- Backend unified resource model remains for AI context
This is a checkpoint before attempting full frontend migration to unified model.
The Resources page was showing 0 resources because the store was only
populated when /api/state was called (from the dashboard). Now the
resources are populated on-demand when /api/resources is accessed.
Changes:
- Added StateProvider interface to ResourceHandlers
- SetStateProvider() method for injecting the monitor
- HandleGetResources now calls PopulateFromSnapshot before querying
- Router injects monitor as state provider during SetMonitor()
This ensures the /resources page works even when accessed directly
without visiting the main dashboard first.
This commit implements the Unified Resource Architecture for AI-first
infrastructure management. Key features:
Phase 1 - Backend Unification:
- New unified Resource type with 9 resource types, 7 platforms, 7 statuses
- Resource store with identity-based deduplication (hostname, machineID, IP)
- 8 converter functions (FromNode, FromVM, FromContainer, etc.)
- REST API endpoints: /api/resources, /api/resources/stats, /api/resources/{id}
- 28 comprehensive unit tests
Phase 2 - AI Context Enhancement:
- Unified context builder for AI system prompts
- Cross-platform query methods: GetTopByCPU, GetTopByMemory, GetTopByDisk
- Resource correlation: GetRelated (parent, children, siblings, cluster)
- Infrastructure summary: GetResourceSummary with health status counts
- AI context now includes top consumers and infrastructure overview
Phase 3 - Agent Preference & Hybrid Mode:
- Polling optimization methods in resource store
- ResourceStoreInterface added to Monitor
- SetResourceStore() and shouldSkipNodeMetrics() helper methods
- Store automatically wired into Monitor via Router.SetMonitor()
- Foundation ready for reduced API polling when agents are active
Files added:
- internal/resources/resource.go - Core Resource type
- internal/resources/store.go - Store with deduplication
- internal/resources/converters.go - Type converters
- internal/resources/platform_data.go - Platform-specific data
- internal/resources/store_test.go - 28 tests
- internal/resources/converters_test.go - Converter tests
- internal/api/resource_handlers.go - REST API handlers
- internal/ai/resource_context.go - AI context builder
- .gemini/docs/unified-resource-architecture.md - Architecture docs
All tests pass.
- Extended AI context selection to host rows in HostsOverview
- Added resourceId prop to StackedMemoryBar for sparkline support
- Relocated guest URL editing from GuestRow name click
- Added GuestNotes component with URL field in AI sidebar
- Refined host routing in AI service backend
- Minor animation and styling improvements
- Implement 'Show Problems Only' toggle combining degraded status, high CPU/memory alerts, and needs backup filters
- Add 'Investigate with AI' button to filter bar for problematic guests
- Fix dashboard column sizing inconsistencies between bars and sparklines view modes
- Fix PBS backups display and polling
- Refine AI prompt for general-purpose usage
- Fix frontend flickering and reload loops during initial load
- Integrate persistent SQLite metrics store with Monitor
- Fortify AI command routing with improved validation and logging
- Fix CSRF token handling for note deletion
- Debug and fix AI command execution issues
- Various AI reliability improvements and command safety enhancements
- Extract ostype from LXC container config (debian, ubuntu, alpine, etc.)
- Map ostype values to human-readable names (e.g., "debian" -> "Debian")
- Add OSName field to Container model and ContainerFrontend
- Add icons for NixOS, openSUSE, and Gentoo in frontend
- LXC containers now show OS icons alongside VMs in the dashboard
Supported LXC OS types: alpine, archlinux, centos, debian, devuan,
fedora, gentoo, nixos, opensuse, ubuntu, unmanaged
- Add AI service with Anthropic, OpenAI, and Ollama providers
- Add AI chat UI component with streaming responses
- Add AI settings page for configuration
- Add agent exec framework for command execution
- Add API endpoints for AI chat and configuration
Host agent was including Docker overlay2 mounts from TrueNAS SCALE's
.ix-apps directory in disk totals. These mounts inherit the ZFS pool's
AVAIL space, causing massively inflated storage numbers (e.g., 173 TB
per container overlay instead of actual usage).
Changes:
- Add /mnt/.ix-apps/docker/ to container overlay path exclusions
- Use ShouldSkipFilesystem() in host agent disk collection (was only
using ShouldIgnoreReadOnlyFilesystem() which missed container paths)
- Add test cases for TrueNAS overlay paths
Related to #718
- Rename checkFlapping to checkFlappingLocked to clarify lock contract
- Replace goto statements with structured control flow
- Wire up unused recordAlertFired/recordAlertResolved metric hooks
- Add trackingMapCleanup goroutine to prevent memory leaks from stale entries
- Tighten alert ID validation to alphanumeric + safe punctuation
- Fix history save error handling to properly manage backup lifecycle
- Add auto-migration for deprecated GroupingWindow field
- Refactor 300+ line UpdateConfig into focused helper functions
- Unify duplicate evaluateVMCondition/evaluateContainerCondition
- Add constants for magic numbers (thresholds, timing, flapping)
- Update tests to match new backup behavior
- Add unit tests for internal/buffer package
- Fix misleading "ring buffer" comment (it's a bounded FIFO queue)
- Remove unused BufferCapacity config field from both agents
- Rewrite flaky integration test to use polling instead of fixed sleeps
When new nodes are added to a Proxmox cluster after Pulse was
initially configured, they weren't showing up in Settings. The
existing "Refresh" button only triggered network discovery, not
cluster membership re-detection.
Changes:
- Add POST /api/config/nodes/{id}/refresh-cluster endpoint
- Add "Refresh" button in cluster node panel in Settings
- Re-detect cluster membership and update stored endpoints
Related to #799
The config package has a more robust IsPasswordHashed function that
handles truncated hashes. The auth package had a simpler duplicate
that was only used in tests. Removed the duplicate and its test
(already covered by config/config_utils_test.go).
Reduces deadcode findings from 7 to 6.
The isLoopback function in internal/hostagent/agent.go was unused in
production code - it was a duplicate of the same function in
internal/hostmetrics/collector.go which is actively used.
Removed the dead code along with its associated tests to reduce
maintenance burden and improve code clarity.
Refactor the test to avoid unreachable code after panic by
checking a flag set before the panic instead of after. This
resolves the go vet warning while maintaining test coverage.
