Release 2026/05/14
v3.9.3
Observ May Updates
This release brings stronger event logic, richer result review, and broader detection coverage—making Observ more flexible for event configuration, faster for result validation, and more practical across real-world scenarios.
New Features:
✅ API V2 Upgrade
We’ve introduced a new version of the Observ API for richer integration and more scalable workflows, including event, task, camera, and location APIs, along with real-time event subscriptions.
Highlights include:
Pagination-ready query patterns for list APIs
More complete task, event, camera, and location GET, POST, PATCH, or batch requests
More information on the event type detecting logic details
🤖 Enhanced VLM Logic: Multi-Condition Trigger Support
VLM-based event analysis is now more flexible for more advanced logic combinations.
You can now define event decisions using multiple logical conditions, enabling:
More precise trigger control
Better handling for complex scenarios
More adaptable event evaluation logic across different use cases
🔗 Custom Events: Multi-Object Logic Selection
Custom object detection events now support multi-object logic selection, making it easier to define event rules with multiple target objects and logic combinations.
🗺️ New Map Page
A new map page is now available. Users can click into the map to view related cameras and detailed information.
🖼️ Single Image Inference Result View
You can now have inference results at the single-image level with a more complete detail panel.
Based on the UI you shared, users can:
Upload images directly
Detection objects and VLM events
Review event analysis results in detail
Users can now upload images and inspect detailed inference information in one place, including detected objects, VLM, and event analysis details.
🚗 Person / Vehicle Check-In & Check-Out Events
New check-in and check-out events are now available for cars and people, enabling tracking of entry, exit, dwell status, and dwell time within a defined area.
🧩 General Model Update
We’ve updated the general model to improve accuracy, especially for weapon and person detection, and added new classes:
road pothole
road crack
road patch
These new classes help detect road cracks and uneven or patched road surfaces.
Coming Soon:
We’re preparing the next wave of updates to strengthen VLM reasoning, expand temporal understanding, and improve platform governance:
Add bbox and zone context into object detection + VLM flow, giving VLM more explicit grounding for targeted reasoning, such as describing a specific person’s status or clothing.
Support sending multi-frame sequential images to VLM for temporal reasoning.
Add platform security activity logs for key operations, such as task or resource creation, deletion, and editing.
These updates will help Observ deliver more context-aware analysis, better decision quality, and stronger operational traceability in real-world deployments.
Support
For technical support, please contact our support team during business hours.
Last updated