We have released Gabriel version 2.0. The highlights for this release are:
- Clients can now send frames from multiple sources. This can be used to run early discard filters on the client, which choose the frames that get sent to the server. Two early discard filters running on a client would be two separate sources.
- Support for multiple cognitive engines, running on the same server or elsewhere.
Gabriel servers allow multiple clients to concurrently use the same cognitive engine. Communication between clients and servers uses the WebSocket protocol. Communication between the server and standalone cognitive engines uses ZeroMQ. All data is serialized using protocol buffers.
We have written client libraries for both Python and Android. Both libraries include tools to measure frame rate and average message round trip time (RTT).
All source code for Gabriel is available in this repository. We have also published Gabriel components on PyPi and Maven Central.
OpenRTiST has been updated on the Google PlayStore with Gabriel v2.0.1 and the backend Docker container has also been updated on Docker Hub. We have introduced a few new tags for the OpenRTiST Docker container to prevent developmental builds from impacting users. There is now a stable tag which maps to the most recent release of OpenRTiST. The stable and v2.0.1 tags currently reference the same container. The latest tag is used for active development and has CI/CD between GitHub and Docker Hub.