Releases: livekit/ingress
Releases · livekit/ingress
Version 1.4.2
Version 1.4.1
Version 1.4.0
CHANGELOG
This version decreases significantly the glass to glass latency of the
bypass_transcoding WHIP pipeline.
ADDED
- Set participant metadata (#222)
- Report the region in CreateIngress and UpdateIngressState RPC calls (#227)
- Use pacer when transcoding is enabled (#230)
- Try restarting the output part of the pipeline up to 3 times if the room disconnected (#240)
- Remove the jitter buffer in the non transcoded WHIP path (#245)
FIXED
- Call RegisterListActiveIngressTopic on start (#216)
- Switch server-sdk-go to v2 (#219)
- Pick up NULL CPU monitor from protocol (#220)
- Remove some verbose logs. Gather PLI count in the stats object. (#221
- Fix typo and handle backward compatibility with livekit-server properly (#229)
- Fix panic if the whip handler receives data before the stats gatherer is set (#233)
- Use a different idle threshold in availability handler and when deciding to accept a request (#234)
- Avoid nil pointer dereference if NewLKSDKOutput fails to connect to the room (#235)
- Make sure to forward Sender Reports to the synchronizer in the WHIP handler (#243)
- Update to Go SDK 2.1.0 to fix SCTP compatibility issue (#246)
- Ensure we match WHIP simulcast layers properly (#247)
Version 1.3.0
Changelog
Added
- Add EndedAt field to Ingress State (#40)
- Log room name and participant ID (#191)
- Gather input stats for url pull sessions (#192)
- Add support for SRT in URL Pull ingress (#206)
- Compute input media jitter (#205)
- Gather pipeline output statistics (#207)
- Add WHIP Simulcast Support (#209)
Fixed
Version 1.2.2 (#184)
Version 1.2.1
Changelog
Fixed
- Flush between each frame when copying data on the WHIP relay (#157)
- Fix CanAccept logic (#159)
- Avoid panic when querying the stats monitor before service start (#161)
- Allow reconnections when the output fails (#164)
- Do not an error if ingress update failed because the ingress is not found (#163)
- Make sure that we resize layers to even dimensions (#172)
Ingress version 1.2.0
Changelog
Added
- Implement high motion presets (#153)
Fixed
- Add queues between pipeline steps (#146)
- Small quality/resource usage/latency tuning for x264 and vp8 encoding parameters (#147)
- Adopt COMPLETE state for URL ingress. Use udp port multiplexing by default for WHIP (#148)
- Do not attempt to unpublish a track from a closed room (#150)
- Use souphttpsrc as it provides better error messages (#151)
- Improve Ingress A/V sync (#154)
Ingress version 1.1.0
This is the v1.1.0 release of the ingress service. This release adds support for pulling media from URLs and fixes source resolution handling
Changelog
Added
- Log redacted ingress info and set stereo and disableDTX publication options (#130)
- Add Debug handler with pprof support (#131)
- Report average bitrate in the ingress state (#133)
- Support for GST pipeline debug dot files on the debug handler (#136)
- Support for pulling URLs (#138)
Fixed
- Do not report instance as available while draining (#124)
- Handle better the case where 2 publishers with the same ingress ID connect to the same ingress instance (#125)
- Check stream key in WHIP DELETE call (#126)
- Make sure handler processes use the same node id as the service (#129)
- Update the WebRTC settings engine to match the sessions parameters (#139)
- Adopt jitter.PopSamples in the WHIP handler (#141)
- Only accept specific file types in url pull (#142)
- Fix layer dimension calculation logic (#143)
Ingress version 1.0.0
This is the v1.0.0 release of the ingress service. This release stabilizes the API and adds support for the WHIP protocol
Changelog
Added
- WHIP support (#97)
- Expose heath and availability on the WHIP HTTP server (#109)
- Add support for optimally bypassing transcoding when ingesting WHIP (#111)
- Set the source track information in the ingress state (#121)
Fixed
LiveKit Ingress v0.9.0
Changelog
This release breaks the existing ingress protocol API.
Fixed
- Update Ingress state to
failed
when we fail to start the handler (#68) - Configure logrus to provide an output that is similar to the zap output for go-rtmp logging (#70)
- Drain active ingress when receiving SIGTERM (#71)
- Use codec constants instead of mime types (#74)
- Prevent go-rtmp panics on bad user input (#80)
- Mark the track as stereo when we have more than channel (#76)
- Adopt psrpc Errors (#83)
- Adopt core.Fuse (#82)
- Fail the health check while draining (#84)