Skip to content
2026-10-09

2026-10-09

Added pagination support for timeline media and marker APIs, including cursor metadata and configurable page sizes for smoother browsing of large result sets

Hub API

  • Added pagination support for timeline media and marker APIs, including cursor metadata and configurable page sizes for smoother browsing of large result sets
  • Improved live video and HLS playback responsiveness by reducing repeated device permission checks during active streaming sessions
  • Enhanced workflow authoring with configurable file-based contracts for custom stages, Start node controls, validation rules, and editor metadata updates without service restarts
  • Added stronger workflow validation and compatibility handling for conditions, unavailable stages, secret fields, and forwarder delivery settings to improve reliability of custom workflows
  • Improved workflow forwarding configuration with support for explicit RabbitMQ destination settings, retry controls, dead-letter queue options, and safer credential handling
  • Added support for enforcing camera group access permissions in workflow definitions to prevent unauthorized group usage
  • Optimized marker category and notification statistics queries for significantly faster performance at scale when recommended indexes are applied
  • Improved muted-device handling during storage uploads and verification requests for more consistent upload behavior

Hub Workflows

  • Added camera group-aware workflow triggering and filtering, enabling automatic workflows and conditions to respond accurately to device group memberships
  • Improved workflow dispatch reliability by ensuring node settings are resolved consistently at dispatch time and retried safely if workflow configuration lookup fails
  • Enhanced forwarder workflow handling with stricter destination validation and clearer separation between deployment-managed and custom destinations
  • Prevented workflows from inheriting outdated or invalid forwarder destination settings after configuration changes or removals
  • Improved storage credential lookup performance by reducing unnecessary database data transfer during site resolution

Hub ANPR

  • Added per-workflow ANPR node settings so individual runs can override plate classes, confidence thresholds, frame budgets, movement filtering, and tracking sensitivity without affecting other workflows
  • Improved isolation between workflow runs to ensure node-specific ANPR settings are applied only to the current message being processed
  • Enhanced validation and fallback handling for invalid node settings, with unsupported values safely reverting to deployment defaults
  • Improved privacy and security of workflow responses by preventing run parameters, upstream inputs/results, and storage credentials from being returned in routed results
  • Preserved signed media URLs across downstream workflow stages to reduce unnecessary storage fetches and improve pipeline efficiency

Hub Vlm

  • Added support for per-workflow VLM settings, allowing individual workflow nodes to customize prompts, frame sampling, output limits, reasoning effort, and timeouts without changing deployment-wide defaults
  • Expanded Azure OpenAI reasoning effort options with new xhigh support
  • Improved request isolation so workflow-specific model settings and timeouts do not affect other runs or shared deployment configuration
  • Added stronger validation and fallback handling for invalid workflow node settings, preventing bad inputs from disrupting processing
  • Improved timeout handling across fetching, model requests, and full workflow processing for more predictable execution behavior
  • Enhanced privacy protections by ensuring credential-bearing workflow parameters are never echoed back in results or logs

Hub Frontend

  • Added a configurable workflow Start experience with customizable fields, trigger modes, scheduling, and connection controls driven by deployment-provided contracts
  • Introduced smarter workflow condition building with guided field selection, typed inputs, upstream stage context, and support for matching related values within the same detected object
  • Redesigned workflow connection panels to organize settings, schedules, and conditions into structured sections with improved handling of retained and read-only values
  • Improved workflow editor reliability by validating stage and Start contracts defensively, preserving unavailable nodes safely, and preventing invalid workflow graph configurations
  • Expanded workflow stage settings with richer input types, validation, defaults, multiselect support, and improved handling of secret and optional values
  • Updated workflow editing interactions so inspectors open only on deliberate selection, reducing accidental panel changes while dragging or placing nodes
  • Improved marker labeling across analytics views and detail panels to show classification-aware names consistently
  • Optimized device status polling so navigation badges update efficiently without continuously loading full device lists on unrelated pages
  • Improved device navigation status indicators to use lightweight online/total device statistics for faster and more accurate badge updates
  • Refined audit page pagination styling and responsiveness for a more consistent experience across screen sizes

Hub Cleanup

  • Added optional deployment-wide retention controls for workflow_runs, allowing old workflow execution records to expire independently of media retention policies
  • Improved workflow run cleanup accuracy for legacy and current timestamp formats, preventing runs from being retained or deleted incorrectly
  • Increased cleanup batch size defaults to improve performance and throughput on large datasets
  • Reduced database overhead during cleanup by optimizing delete operations and workflow run indexing behavior
  • Improved index migration handling so outdated indexes are safely replaced without interrupting cleanup operations

Pipeline Monitor

  • Reduced unnecessary device latestMedia updates by throttling refreshes to once every 5 minutes by default, improving stability and database efficiency during high recording activity
  • Added support for configuring the latest-media refresh interval with LATEST_MEDIA_UPDATE_INTERVAL_SECONDS, including an option to update on every recording
  • Improved reliability of latest-media tracking so delayed recordings no longer overwrite newer device activity
  • Automatically recovers from invalid future-dated latest-media timestamps to prevent devices getting stuck with stale activity
  • Optimized stored device latest-media snapshots to keep only essential media details while preserving the latest recording context for readers
  • Added explicit reporting when a latest-media update is skipped due to throttling instead of treating it as a failed write

Pipeline Notification

  • Improved notification processing performance by reducing the amount of data loaded during duplicate detection checks
  • Reduced alert lookup payload sizes by excluding large site and group assets, improving efficiency for notification delivery
  • Prevented sensitive site and storage credential details from being included in notification and webhook-related data lookups
  • Fixed an issue where private alert configurations could unintentionally affect alert execution during rollout scenarios

Pipeline Analysis

  • Improved pipeline performance and reduced database bandwidth usage when resolving site vault credentials
  • Reduced unnecessary data retrieval during site vault lookups, improving efficiency for recordings linked to sites with large metadata payloads
  • Improved reliability of site vault resolution logic with additional safeguards around credential field handling

Pipeline Thumbnail

  • Improved service reliability by allowing thumbnail processing to continue even when tracing or telemetry connections are unavailable
  • Reduced startup failures caused by tracing configuration issues, improving resilience in environments without telemetry endpoints

Pipeline Sprite

  • Improved service reliability by allowing the application to continue running even when tracing services are unavailable
  • Added graceful fallback behavior for tracing, preventing startup failures when telemetry endpoints cannot be reached

Vault

  • Added per-storage-provider URL signing controls, allowing each provider to use either Vault-signed URLs or direct provider presigned URLs independently of the global setting
  • Vault-signed URLs can now fully hide underlying storage endpoints from clients on a per-provider basis
  • Added URL signing configuration options to the storage provider management UI and API
  • Improved validation for storage provider signing settings with clearer error handling for unsupported modes

Factory

  • Improved region-of-interest editing reliability by preserving saved polygon coordinates accurately across camera resolution changes and agent restarts
  • Added support for storing the original camera resolution used when creating motion detection regions, improving region alignment on resized or different-resolution snapshots
  • Fixed cases where outdated agent configuration data could overwrite newer saved region settings in Kubernetes deployments
  • Configuration updates in Kubernetes now return clear errors when changes cannot be fully applied to running agents
  • Improved deployment creation reliability with better duplicate deployment detection and clearer error handling
  • Enhanced Kubernetes-backed configuration handling to consistently load the latest persisted per-agent settings from ConfigMaps and Secrets

Learn

  • Added an optional media proxy feature that hides direct storage and Vault URLs behind secure Hub API links while preserving streaming and seek support
  • Introduced a new “With analysis” media filter to quickly find recordings that contain detections, markers, descriptions, or other analysis results
  • Expanded and published the Organisations and Projects documentation with clearer explanations of workspaces, access boundaries, project isolation, and multi-project workflows
  • Improved guidance for organisation and project switching, including how workspace context affects recordings, workflows, and background processing
  • Clarified default project behavior and legacy compatibility handling to support smoother upgrades and more predictable project visibility across existing deployments

Website

  • Added dedicated product pages for Agent, Vault, Factory, and Hub with improved navigation and discoverability
  • Introduced a new public CV page for Cédric Verstraeten, including a downloadable PDF version and personalized branding
  • Improved search engine visibility with expanded sitemap coverage, route-specific metadata, and structured profile data
  • Enhanced the website navigation experience with new dropdown menus and responsive mobile behavior
  • Added direct “Explore product” links from architecture and composition sections to related product pages

Helm Charts

  • Added support for deployment-wide workflow identities with stable IDs to preserve workflow references and history across updates and renames
  • Introduced configurable media proxy support so media can be served through the Hub API instead of exposing direct storage or Vault URLs
  • Added new workflow editor contract packaging and validation capabilities, enabling controlled custom workflow stages, conditions, and reusable configuration catalogs
  • Expanded workflow management with bundled Start, ANPR, and Forwarder workflow contract support for richer workflow authoring and validation
  • Added new media filtering option for analysis results in the frontend
  • Added configurable retention policies for workflow run history cleanup
  • Improved MongoDB observability by allowing per-service application names in MongoDB connections
  • Added the ability to temporarily disable agent heartbeat polling to reduce database activity during maintenance or scaling events
  • Updated classification catalog management to use ConfigMap-based configuration, improving consistency and customization support across deployments