Flussonic 26.05: Faster troubleshooting, richer stats, smoother NVR and mosaic workflows

Here is what we did for media server

Find the root cause faster using stats and errors.
Make ONVIF and configuration behavior more robust.
Plan GPU transcoding capacity and input diagnostics more easily.
Reduce delays and surprises in live/DVR scenarios.

Faster troubleshooting without digging through logs. Stream statistics now include a description of the last transcoder error, which helps you understand what went wrong faster. We also added more input counters (including UDP MPEG-TS unicast/multicast) and expanded DeckLink telemetry, making signal quality and stability easier to assess.

More robust integrations and configuration. ONVIF event handling now tolerates gaps in XML structures: motion detection works more reliably and supports a wider range of cameras. We fixed rare cases where on-disk settings could be ignored when config_external had issues. We also fixed URL parsing in configuration and edge cases with stream names that contain /.

Easier GPU capacity planning. We added NVENC benchmarking tools to help estimate throughput and capacity. NVENC raw format support was improved (including r210/ARGB), reducing cases where pipelines fail due to unsupported input formats. The API now provides more detailed GPU hardware information as well.

Fewer production issues in large deployments. API responses for large result sets are faster. The JSON logger now respects different message size limits per log level and truncates oversize messages to avoid log bloat and performance degradation. We also removed Quicksync support from the configuration parser and fully removed libsrt, keeping the Erlang SRT implementation — reducing build and dependency complexity.

More predictable live and DVR behavior. Markers inserted via API now appear in the rewind playlist without noticeable delay during live playback. VOD now correctly applies segment_duration changes when configuration is updated.

Here is what we did for central

More resilient stream distribution under errors and high load.
New delivery capabilities via CDN zones and centrally managed playback authorization.
Fewer unexpected assignments inside the cluster.

More resilient distribution. Central Layouter’s distribution algorithm was improved: mass stream reassignments are no longer blocked by a single stream processing error, and disabled streams are reassigned more reliably under high load.

More control over delivery and playback. We added the CDN zones API and support for it via the central playback balancer, making geo-distribution and delivery policies easier to manage. We also added Central-managed playback session authorization so access control can be enforced centrally.

Fewer incorrect assignments. We fixed a bug where an agent without a stream could be assigned to a streamer in a different namespace. Cluster configuration for mutual DVR cross-replication was improved as well.

Here is what we did for watcher

Find cameras faster and return to work without losing context.
Make mosaics and episodes easier to use: more layouts, less lag, faster navigation.
Reduce manual NVR work and make archive behavior more predictable.
Expand API v3 capabilities and protect sensitive data in public endpoints.

Faster navigation with fewer clicks. The camera list now preserves scroll position when you return from a camera page, so you don’t have to find your place again. Filters and UI state are retained more reliably across navigation, including switching between old and new UI versions. Camera search and onboarding are smoother: there is manual camera search, found-device lists behave more predictably, and unadded cameras are shown first. We also added a “search all local agents” screen and made WNVR redirects and flows clearer.

More usable mosaics and episodes for daily work. We added new grids and the ability to select a specific layout (including 6x8), and we improved mosaic performance and stability. Episode feeds are easier to navigate: fewer “unloaded” areas, a scroll-to-top button, and filtering episodes by the selected camera. We also added episode statistics (for example, view count) and the ability to leave comments.

Less manual NVR upkeep and a more resilient archive. NVR autosync is now enabled by default, and redundant manual sync buttons were removed — keeping camera lists up to date takes less effort. You can now configure how many servers store an archive copy, which helps balance retention resilience and capacity. We also fixed archive viewing issues in unstable states (for example, when a camera is temporarily unavailable). The default preset was updated: archive is enabled by default for new Watcher installations.

API v3 and security. We added a password-recovery endpoint in API v3. Watcher now has API v3 endpoints for settings such as branding, external auth, and maps, along with an updated UI for managing them. We also improved security by removing sensitive input URLs from a public API endpoint.