Release Notes: DL Streamer Pipeline Server 2025#
Version: 2025.2.0#
Release Date: December 2025
Added:
souphttpsrcelement in docker file.
Fixed:
Wrong formatting in compose file.
Added missing
eis_mqtt_publish_doc.mddocumentation.Fixed WebRTC GPU pipeline functionality for Xeon+dGPU hardware combinations.
Updated Helm chart version in documentation and dockerhub to be in sync with current one.
Resolved issue where videoconvert was dropping tensor data provided by gvametaconvert.
Updated Helm chart on Dockerhub.
Incorrect communication between containers has been fixed by configuring proper env variables.
RSTP connection error recovery mechanism.
Updated:
DL Streamer updated to 2025.2.0.
v3.1.0#
Release Date: August 2025
Added:
Support for Ubuntu22 and Ubuntu24 based docker images.
Separate optimized and extended runtime docker images.
Publisher for InfluxDB to store metadata.
OPC UA is now configurable in REST request.
Improved logging by consuming log levels from
.envinstead of fromconfig.json.WebRTC bitrate is now configurable.
Logs can be queried and monitored in real time with Open Telemetry.
ROS2 publisher for sending metadata (with or without encoded frames).
Enabled VA-API based pipelines for RTSP and WebRTC streaming.
Fixed:
Cleanup: Remove confidential info such as email and gitlab links. Removed unused model downloader tool, gRPC interface.
Bug in appsink synchronization behavior not being consistent with gstreamer/DL Streamer.
Bug in appsink destination and publisher configurations.
WebRTC with GPU inferencing falls back to CPU if vah264enc is missing.
Updated:
DL Streamer updated to 2025.1.2.
Interface to Model registry updated with environment variables instead of
config.json.Documentation updates: Cross stream batching, latency tracing, tutorial on launching and managing pipelines.
v3.0.0#
Release Date: April 2025
Updated:
Rebranded Edge Video Analytics Microservice (EVAM) to Deep Learning Streamer Pipeline Server (DL Streamer Pipeline Server).
v2.4.0#
Release Date: March 2025
Added:
Enabled frame publishing over WebRTC protocol to a MediaMTX server.
New REST API to get pipeline instance status:
GET/pipelines/{instance_id}/status.
Fixed:
Fixes in model update flow.
Fixed an issue where overlay was improper for published frames when source has I420 image format.
Updated:
DLS upgraded to 2025.0.1.2.
Geti™ SDK upgraded to version 2.7.1, sample model files updated.
RTSP and WebRTC allows watermark overlay on frames using REST.
Updated documentation.
v2.3.0#
Release Date: March 2025
Added:
Image blob write support for S3 API compliant storage.
Metadata and image blob (optional) publish support over OPC UA protocol.
OpenTelemetry support to publish gathered metrics to Open Telemetry collector.
MRaaS model update support for non-Intel® Geti™ models (YOLO and OMZ) loaded through gva DL Streamer elements that perform inference.
Optimized docker image size - removed unused libraries.
Fixed:
Warnings from OpenVINO™ telemetry.
Updated:
DL Streamer updated to 2025.0.1.
Updated third party programs list for components with copyleft licenses.
Updated documentation.
v2.2.0#
Release Date: February 2025
Added:
Support for synchronous REST API, timeout and base64 image in Image Ingestor.
MRaaS support in Helm chart.
Support for NV12 and I420 image formats in EVAM Publisher.
Option to send frames optionally in image ingestion REST API requests.
Payload support in config when auto-start is enabled.
Insourced pipeline server.
Fixed:
Bug in Geti™ UDF loader’s color space handling.
Bug in not being able to run multiple instances simultaneously of the same pipeline on dGPU and iGPU.
Bug in allowing to re-run pipeline with a failed model-instance-id.
Updated:
DL Streamer updated to 2025.0.0.
Geti™ SDK updated to version 2.5.0.
Updated documentation, license, copyright.