LiveKit Community

just to understand: are you trying to have two different rooms connected at the same time from within the same app?
cc <@U02GAC58A79>

it will create two diffrent room for different calls and participants will connect to their different rooms

i also found that when second call is being made by 2 other device ,some times for  first call remote participants is getting disconnect for while and getting connect after some time.

Just so I understand: you have four participants, which you would like to have in two different calls?

Are you connecting to the same room name for each call, or does each call have a different room name? Personally I find it easier to have a different room for each call in apps that are used for calling and conferencing.

for each different call we are using different room name.

It should be related to the life cycle of localTrack, maybe you need to clone localAudioTrack/localVideoTrack, and then put it into a different room to prevent one of the tracks from terminating

In the native/flutter sdk, you cannot use getUseMedia as many times as chrome to get localStream, so you need to manage the life cycle of localStream yourself.

<@U02GAC58A79> please can you give any reference how can i do this?

which version of Egress are you using? this seems like a bug that's been fixed: <https://github.com/livekit/egress/pull/197>

<@U01KZ61GV5Z> - my egress version is 1.5.2 which version should I upgrade to to get the fix?

The fix was in <https://github.com/livekit/egress/releases/tag/v1.5.3|1.5.3>

<@U01KZ61GV5Z> : could you update the helm chart for egress seems it’s not up to 1.5.3?

image.png

I see you've opened a PR for this, we'll take a look.

Ingress has been tested fairly extensively with ffmpeg. Could you ensure you provide a an "app" field in the RTMP url? ffmpeg's rtmp implementation requires it

Specifically, how are you using it, and is there any context on the lack of handshake support on their side? This seems like a violation of the RTMP protocol

here is the srs: <https://github.com/ossrs/srs>

Screenshot 2023-01-31 at 7.32.48 PM.png

I have autosubscribe Off and I subscribe selectively for tracks and then enable disable tracks as and when required.

I waited for very long , but at the end . Had to refresh and publish fresh streams.

yeah it should resume correctly. Can you share the session here? which SDK are you using?

This is logging the RTMP Url associated to the ingress. The server is not doing anything with this URL directly

<@U04LRBN31L4> can you clarify what you mean by `do a reconnect event` ?

Do you mean you interrupted your internet connection and then reconnected it?

<@U01KZ61GV5Z> i tried the reconnect event.  Simulate a reconnect

And then I can't get the remote streams working

I see, I'm assuming using the option within the options menu?

<@U01KZ61GV5Z> are you able to reproduce it?

I have not, just trying to get more info so <@U02GAC58A79> could take a look later.

I managed to fixed it by unsubscribing and then subscribing back on each track when `RoomReconnectedEvent` appear

right, you shouldn’t have to hack around it

<@U04LRBN31L4> When you simulate a reconnect, can the remote audio track be received? I tested that when there are two remote tracks, the simulated re-connect will hear the sound, but the video stream will come later,

I'll check. I only tested for video so far

<@U02GAC58A79> I just tested it. neither microphone or video work

ah, okay, what version of livekit-server are you using?

hi, <@U04LRBN31L4> After several rounds of testing, in the local livekit server and cloud, it seems to be working fine, except that there is a random delay of 0~4 sides in the video rendering. If possible,  can you provide some reproducible test environment or browser console logs? , and I can try to track down the error.

<@U04LRBN31L4> looks like you are reproducing this in Safari? are you able to repro with any other browsers?

I was able to reproduce it on Chrome, Firefox and Safari

I will try again today and see how it goes

logs_live_kit.txt

logs_live_kit_2_browsers.txt

<@U02GAC58A79> <@U01KZ61GV5Z>
Sorry for my late response. I tested again. I use the flutter library v1.1.12.

I tried the following scenarios — all reconnect events were created from the debug Chrome from osX:
• Test 1 — debug Chrome osX with WIN Firefox — I did reconnect and everything worked fine
• Test 2 — debug osX Chrome with WIN Firefox and osX Safari — I did reconnect and it didn’t recovered anymore
• Test 3 — debug osX Chrome with WIN Firefox and debug osX Safari — I did reconnect several times from Safari this time — It didn’t recover and I had to force close it due to a high increase in GPU `com.apple.WebKit.GPU`  went crazy to 100+ % in activity monitor — This issue with safari using a lot of GPU it’s almost always replicable if I do several reconnects events there or multiple streams stops and starts again
• Test 4 — debug osX Chrome with WIN Firefox and debug osX Safari — I did reconnect several times from chrome and it started at some point, but the streams were not displaying correct data. All streams were showing osX Safari.
Log file `logs_live_kit.txt` attached is for the scenario with 3 devices ( 2 osx chrome, 1 win safari), log coming from debug flutter web within osx chrome with the output that the streams didn’t come back alive after reconnect.

Log file `logs_live_kit_2_browsers.txt`  attached is for the scenario with 2 devices ( 1 osx chrome, 1 win safari), log coming from debug flutter web within osx chrome with the output that the streams didn’t come back alive after reconnect.

Thanks for the details, I'll check this bug:pray:

UPDATE for test where safari overloads GPU
(this slowdown only happens on osX Safari)
—
I added the <https://pub.dev/packages/camera|camera plugin> on a page and render all the the cameras I have connected (2 real and 1 virtual). Did lots of screen resizes, page refresh etc and the GPU didn’t load more than 20%.
—
So maybe some webrtc interactions or the way example is built makes Safari to go crazy with GPU

thank you for the detailed bug report! I wish every report we get is like this one.

<@U04LRBN31L4> I fixed some issues after testing yesterday, you can test it is this branch

<https://github.com/livekit/client-sdk-flutter/pull/238>

<@U02GAC58A79> I just tested it and it doesn’t work.

I deployed it here using my server:
<https://livekit-sample.web.app/#/>

even with only 2 users, it doesn’t reconncet anymore

I tested again on all browser. I tested it with
firefox osx
chrome osx
safari osx
firefox win
firefox android 13

on firefox stream leaves the last frame on screen. on the others the stream goes black.

if I rejoin the room from one of the other device, all the streams starts again on the browser that got stuck after reconnect

Thank you for deploying the webapp :pray:, I can reproduce now, let me try to find the bug

It's deployed as release build, but I can deploy a debug one too

I can only open this page in safari on iOS, and an error will be reported on osx and the UI cannot be displayed

i can open it from android, ios browsers windows browsers and also other browsers on mac

yes, weird, not working in private windows.

<@U02GAC58A79> could it be because it’s built with canvakit?
```flutter build web --release --web-renderer canvaskit```

I re-deployed it with normal not specific rendered:
`flutter build web --release`

with `flutter build web --release` the one that’s now deployed
I can’t get safari working almost at all. it gets stuck pretty quick, and then it depends, it either kicks me out of the room, or never gets any video stream, or after like 20-30 sec gets the stream and then looses it again.

Losing of stream happens on both safari and firefox on mac. They are also not able to send at all

so I gues that the web renderer plays a big role here

hi <@U04LRBN31L4> , I reproduced the issue you described, and committed a key fix in this PR, please update to the latest commit for testing, looking forward to your feedback.
<https://github.com/livekit/client-sdk-flutter/pull/238>

The problem is related to the signaling timing. I tested it on mac-chrome &lt;---&gt; win-firefox, and the simulated re-connect works very well.

<@U02GAC58A79> Thank you man!

It looks like the problem is fixed. I testes it on osx Firefox Chrome Safari, win Safari, iPad Safari, Android Firefox and it works

There is definitely something wrong with web safari. Now it blocks and when I hit reconnect every camera stream goes black and no interaction works anymore. After like 30-40 seconds interactions work again, but still no stream.
If I try to hit reconnect again, after 10-15s it gets kicked out of the room. It’s osX Ventura 13.2, M1 Max, Safari 16.3 (18614.4.6.1.5) build with `flutter build web --release --web-renderer canvaskit` .
I tested it on another mac, like this one, same os same browser and I got the same behaviour.
I also deployed it without --web-renderer and is doing the same

while safari is blocked, it sends the video stream and also audio, but everything else is blocked and black, and then after a while it just disconnects

Yes, safari should be a separate issue, I compiled flutter on m1, the web app won’t even load properly, we might need to try `--web-renderer html`,
I will continue to debug for safari.

I definitely think it’s a separate issue

I might also be related to safari + flutter engine + camera integration.
I didn’t try it on intel mac. I don’t have any

I got a friend testing current build on a mac with intel using safari.
Build was done using: `flutter build web --release`

He had no issues at all, everything worked as expected.

It looks like the m1 version of the flutter sdk can't compile the web app correctly

hmm. I compiled it using m1 mac and deployed it. My friend using mac with intel had it running perfectly

Does your m1 mac have Lock Down mode enabled by default? After shutting down on m1 it seems to work, the Lock Down setting seems to disable wasm loading?

livekit-safari-logs.txt

<@U02GAC58A79> coming back to this thread :))

Currently I have deployed here <https://livekit-sample.web.app/#/> the sample app with latest flutter livekit sdk, version 1.2.0.
It’s buit using html web render from my m1 mac. On Safari, and Safari only :))
I have this bug where the user gets disconnected after several seconds.

Steps to reproduce:
1. Open example or  <https://livekit-sample.web.app> in Safari
2. join a room
3. Allow devices permissions
4. — Problem 1 — It needs around 15 sec to display the local camera feed, but the camera led starts right away
5. Local Camera is displayed
6. After some seconds it gets disconnected.
I did this in a Safari private window, enabling all room properties [simulcast, adaptive_stream, fast_connect, dynacast]. It’s only happening at first connection. If I connect again using the same browser session it seems to work.

I attached here the logs.

Do you have any idea why camera is loading this slow in safari?

hey <@U04LRBN31L4>, thank you for your feedback, as you reported, I also reproduced the issue on safari, there may be two problems here, 1, is MediaStream rendering (a safari-specific issue, `videoElement.srcObject = stream` cannot be call multiple times ), 2, the client connection may be stuck or failed due to the issue of the `IceServers` field in `RTCConfigration`, but these two issues should be fixed soon

<@U04LRBN31L4> Can you try this branch?
<https://github.com/livekit/client-sdk-flutter/pull/241>
It should fixed rendering and publish issue for mac and iOS safari, but it will be a bit flickering when the local video track changes.

it works great, but it seems that a new bug got introduced :))

web session from mac doesn’t receive video streams from mobile web session

I have mac m1 with safari
Android 13 with firefox

---
Android 13 displays the stream from mac
Mac doesn’t display stream from android phone

also tested with a windows firefox and in the setup:
Mac m1 safari
Win 11 Firefox
android 13 Firefox

==
Mac display local + Win
Win display local + Mac
Android display local + Mac + Win

but the safari is way more faster and smooth now

logs.txt

here are the logs from safari.
It’s also deployed on <https://livekit-sample.web.app/>

Looks like firefox on Android 13 doesn't publish track correctly

just tested with a huawei android 10, firefox, same

thanks, very useful feedback, let me see what happened on mobile FF

<@U04LRBN31L4> Does this issue only reproduce on Android 13 + FF? I’m using Android 12 + FF and it seems to work perfectly.

Can you try to test with Android FF again after clearing the cache? I updated the system to Android 13 and tested it with the <https://livekit-sample.web.app> you provided, and it works fine on FF.

<@U02GAC58A79> Just checked it. it only happens with `simulcast` turned on. I tried in both firefox private and also `focus`

logs-firefox-android13-simulcast-off.txt

logs-firefox-android13-simulcast-on.txt

<@U02GAC58A79> I added here the logs I have. They are both from android 13 with firefox. One is with simulcast off, and the stream is published; the other one is with simulcast on, and the stream is not published.

<@U04LRBN31L4>  Can you try [v1.2.1](<https://github.com/livekit/client-sdk-flutter/releases/tag/v1.2.1>), my test result is FF on android 13 with simulcast works fine

<@U02GAC58A79> just tested it.
It’s deployed here <https://livekit-sample.web.app/>

I testes with a Huawei with android 10 and firefox with simulcast on and does the same.

just video is not displayed. audio works fine

so we have livekit_flutter 1.2.1

Device A: Android 10 - Firefox 110.0.1-20230227191040
Device B: Mac Os - Safari

Device A — Simulcast ON -&gt; connects to the room
Device B — whatever state for simulcast -&gt; connects to the room

Device A — Display both camera feed from A and B, receive sound from B
Device B — Display feed from B (local feed only), receive sound from A

can you reproduce it here&gt; <https://livekit-sample.web.app/> ?

okay, this is a bug that I reproduced before I fixed it, can you temporarily switch to feat/e2ee, and confirm that e2ee is turned off during the test

sure, can you tell me where is this setting?

OK, don't worry, e2ee is disabled by default，you don’t need set anything 

视频片段 (2023-03-15 20:34:51).MOV

This may be a bug related to the device model, I use Redmi k40 android 13 + FF, can run with simulcast correctly

If we can’t fix the issue on these specific devices, we may only use the model list to force close simulcast to ensure usability. Even lose some features.

just so I'm following.. does 1.2.1 have the bugfix? or is it only in e2ee branch?

I tested two branches, 1.2.1 and e2ee, both can work normally on FF on Android 13, so it may be a bug related to the device model.

Got it.. that would be strange since it's FF on Android.. it should be using its own webrtc implementation. Can you see if it makes a difference if the recipient is running Safari?

Just tested it, Android FF works fine with mac Safari and iOS Safari (all 3 with simulcast enabled).

I have a couple of androids here.. let me try too

cool, you can test it with <https://livekit-sample.web.app/>

If I use Chrome or Firefox with FF on Android, it works fine. However, if I use Safari on the Mac, FF on Android fails to decode or publish.

But the issue here seems to be the device missing H.264 codec (multi-codec simulcast should resolve that issue)

i tested on a google pixel 1, Android 10, Firefox.
Same behaviour

hey <@U03SN0JBTJT>, just to confirm, is this the library you’re using?
<https://github.com/livekit/components-js>

hey <@U01KZ0SCDKL>, thanks for your response,
No, I use `livekit-client`  and `@livekit/react-core'`

<@U03SN0JBTJT> while not officially deprecated yet, if you can, i’d recommend migrating to the new LK Components stuff. However, going to cc <@U02EU2ZPTQU> in case he might have an idea of what’s going on.

it sounds like it may be particularly related to how you’re managing state in your app?

ok i will see that !
Thanks <@U01KZ0SCDKL>

hard to tell what’s going on, metadata should generally not impact the video playback at all. As <@U01KZ0SCDKL> said, this sounds state handling related.
If you keep hitting this issue it would be great if you could provide a minimal reproduction repo for us to take a look

I installed the egress service by using docker run.

the snippet you pasted is showing manually running pulse audio from the terminal.

Since it fails in the docker start, so I ran the shell commands by copy and paste the commands in the docker, it reproduces the issue

which version of the image tag are you running against?

I am using image labels latest should be 1.5.3.

It works in my arm based server, now I am using another two x86 servers, both of them have such issue.

Ok, give 1.5.5 a try, it's just released earlier today.

There was a pulse-audio related issue that had been fixed in 1.5.4+, but I'm not sure if that's what you are seeing

we have not seen this then. please report an issue in the Egress repo with as much repro steps as you can provide. clear repro steps would help us to get an idea for reproduction.

When I go to the docker, using curl <http://www.bing.com|www.bing.com>, error reported: "curl: (6) getaddrinfo() thread failed to start"

I added option docker "--privileged" in the docker run command, it works. A documentation issue? <@U01KZ61GV5Z>

Hmm, not sure, the underlying native view doesn't support this directly

Does setting border radius on a parent component work?

It looks like `fq` is part of the stats on that host. I'm unfamiliar with these types of stats, though.

```tc -s qdisc show dev enp1s0
qdisc fq 0: root refcnt 2 limit 10000p flow_limit 100p buckets 1024 orphan_mask 1023 quantum 3028b initial_quantum 15140b low_rate_threshold 550Kbit refill_delay 40.0ms 
 Sent 13465397 bytes 43697 pkt (dropped 0, overlimits 0 requeues 0) 
 backlog 0b 0p requeues 0
  flows 837 (inactive 836 throttled 0)
  gc 0 highprio 6 throttled 864 latency 7.796us```

Running on docker on a Ubuntu 20.04.5 LTS host, in case it matters.

looks like `fq` is supported by this `go-tc` package we’re using: <https://github.com/florianl/go-tc/blob/main/qdisc.go#L122>

checking to see if we’re just using an old version

we are using a very old version, but it does still support `fq`
<https://github.com/florianl/go-tc/blob/v0.4.2/qdisc.go#L122>

<@U02C10ZEP8R> we run it with prometheus stats and have never seen this error. any chance you can try upgrading the `go-tc` package to see if your issue gets fixed?

I could try, but looking through the go-tc code, it looks like it isn't supported in that section of the code in the latest release.

Did I read it wrong? In that case, maybe it’s fine to ignore errors when it can’t read network stats

<@U02BBTRGF98> you were reading that part correctly, but the error seems to be coming from here <https://github.com/florianl/go-tc/blob/main/attributeTcMsg.go#L368>, which even in main doesn't have `fq` as a supported kind for some reason.

I'll try to figure out the repro steps soon. I'm not sure if it's some combo of packages, the fact that I'm on vultr, or the specific version of ubuntu that the server is on. My guess is one of those is making for an unexpected result from the network stats.

Nice find. I think it’d make sense not to return an error here: <https://github.com/livekit/livekit/blob/master/pkg/telemetry/prometheus/node.go#L122>

and the system packets gauge will just show 0

Yeah, that sounds better than erroring out completely.

made a branch <https://github.com/livekit/livekit/pull/1386>

I wanted to follow up that this is still on my radar. I had an issue with my VPS provider that I was trying to get resolved so I could test it properly. I at least know what is going on with that now and have a workaround, so hopefully I can get this test in the next day or so.

thanks. I’ll get this changed merged in today anyway.

스크린샷 2023-02-06 오후 4.59.28.png

스크린샷 2023-02-06 오후 4.59.18.png

Hi,  can you provide a couple of details?
1. Which client are you using? 
2. By "lot of time", how much time does it take? Is it 1 second or 5 seconds or more?

화면 기록 2023-02-06 오후 5.26.54.mov

1. It is react client
2. Sometimes it's 1 second, Sometimes it's 5 seconds.


Thank you.

I think this is the result of using the load test. The canned streams do not generate key frames on request. So, there is variable amount of delay depending on when a key frame becomes available in the canned stream.

Is this your own deployment of the server? Or using LiveKit cloud?

Is it possible for you to test with real users?

I am using my own AWS server.
When testing with real users, only about 10 people can test frequently,
It is not an environment that can be tested with a large number of other people.

Got it. With your layout  :point_up::skin-tone-4: , 10 people should sufficient to do 3 pages right? With adaptive streaming, there should not be performance differences with number of off screen participants.

As Raja mentioned, the latency that you see is due to the lack of keyframes in the sample video that our load tester embeds.

the typical activation delay for adaptiveStream and dynacast is about 2 RTTs; typically &lt;500ms

<@U02CM28PB4Y> <@U01KZ61GV5Z> Thank you so much
(With my layout, there is no problem even with 15 people.)

do you have the `react-native-webrtc` package installed, and called `registerGlobals()` in your index.js file?

I installed `react-native-webrtc`, but didn't call `registerGlobals()`.

It worked now, thanks !!

Hi <@U04NH50U1QC>
Some answers:
1. this _should_ serve as an example integration: <https://github.com/livekit/client-sdk-react-native/tree/main/example>
2. cc <@U01NZKTNE1H>
3. we don’t have an admin panel/frontend/dashboard for self-hosted deployments yet, but it’s something we’re planning to release later this year

Thanks <@U01KZ0SCDKL>

But the react native example doesn't include the call service code for iOS (It is written `TODO`)

oh :face_with_open_eyes_and_hand_over_mouth: <@U01NZKTNE1H> wdyt?

you’d be looking at react-native-callkeep, but they have some code that interferes with the audio session settings. I haven’t had time to fork and fix that for our uses.

you can see a previous implementation of it here: <https://github.com/livekit/client-sdk-react-native/blob/966278a1eff9eaf3dd23eff74bf7bbc6481925a9/example/src/callservice/CallService.ios.ts>

what do you get when you run the <https://livekit.io/connection-test|connection tester>?

What OS are you running the Go SDK based application on?

It seems that the TURN is not properly deployed right? We are using a coTURN

right, it doesn't look like it's configured correctly.

one thing to check is to ensure your UDP ports are open. Go SDK does not support ICE/TCP. It could either do TURN or ICE/UDP

Hi <@U01KZ61GV5Z>, we are having 2 problems:
1. (TURN disabled) Connecting with ICE/UDP in the same private network as the LiveKit. It works with ICE/TCP though.
2. (TURN enabled) When we enable the internal TURN, the TURN check doesn't work in any connection test.

TURN config:
`turn:`
	`enabled: true`
	`domain: <http://livekit.domain.com|livekit.domain.com>`
	`udp_port: 3478`
	`tls_port: 5349`
	`cert_file: /path/to/cert.pem`
	`key_file: /path/to/key.key`
	`relay_range_start: 51000`
	`relay_range_end: 51200`

Thanks!

Unfortunately, livekit sdk does not support p2p calls yet, all data needs to be relay through livekit-server,

You can apply such filters to camera frames yourself before passing them to the local camera track, but there’s no built-in affordance for this. What platform are you on?

What library will you use for the "face filters"? OpenCV?  If you have already implemented this the next step is not so hard.

Wait are you developing a native Windows app or webapp?

Flutter mobile app for Android and iOS <@U02MC5V3C2H>

<@U03GEMJGJJ3> We are actually struggling on that next step...

By passing token, I am getting this error on the web

```Unhandled Runtime Error
Error: token endpoint needs to be defined```

Hi <@U04A1328AJY>
the idea for the `useToken` hook is that your app has a defined endpoint that returns a token. you can have a look here:
<https://github.com/livekit/components-js/blob/main/examples/nextjs/pages/api/livekit/token.ts>

be aware that this implementation isn’t actually doing any authentication, as it’s only an example. for a secure implementation you would usually check against your db whether that user is actually allowed to join the room

the `NEXT_PUBLIC_LK_TOKEN_ENDPOINT` then specifies where that endpoint is reachable. In the example that would be `/api/livekit/token`

Thanks for the swift reply, <@U02EU2ZPTQU>. I am a backend engineer and we are currently using the Go SDK to build out the skeleton (token generation for any room’s participant is being created at the backend)

I have some experience with Vue.js so I am trying to test the UI part using the reach repo

ah, in that case the `useToken` function is not really ideal. it’s main use case is currently to ease the setup with nextjs directly.
But you can have a look at the useToken source (it’s quite simple) and adjust it for your needs to request the token directly from the server that’s hosting your go script

Sounds great. That is helpful; I will revert should I have more questions.

Quick question <@U02EU2ZPTQU>, the `NEXT_PUBLIC_LK_TOKEN_ENDPOINT` can be made to point to local endpoint, right (for test sake)? Exp: `GET localhost:8000/token` ? I could see that you pass both `roomName` and `userIdentity`  as params when making a call to the URL, which the backend can use to perform some authorization?

yeah, that’s the idea! but probably this won’t cover all use cases. If you have feedback on how to generalise it / make it more applicable for other use cases, that would be great!

This is neat. It works for our use case “to an extent”. I will be sure to share edge cases when encountered. Thank you, <@U02EU2ZPTQU>

Ok I think this is doing it:

`const stream = await navigator.mediaDevices.getUserMedia({ audio: true });`
`const audioTrack = stream.getAudioTracks()[0];`
`await room.localParticipant.publishTrack(audioTrack);`

you could just do `localParticipant.setMicrophoneEnabled(true)`

more examples <https://docs.livekit.io/client/publish/|here>

`localParticipant.setMicrophoneEnabled(true)` didn't work. No errors, but I can't hear anything.

^ maybe for now just the publishin, not the receiving; that is almost the same like the video/audio example.

Web Audio API methods like `createGain()` and `gainNode.gain()` is used to adjust the volume before sending it to livekit.

<@U01KZ0SCDKL> thx ^^ is Neil the expert in Spatial audio thingsy?

he is — he’s working on a spatial sample which he’ll be sharing with folks here soon :slightly_smiling_face:

btw is it also possible to stream strings beside the audio over webrtc(livekit)?

Is it a all-in-one solution for sharing audio+text messages+position of the users?

<@U01UEMQ8C3F> we lost you in the metaverse? O_O

<@U03GEMJGJJ3> while we’re waiting for Neil to unplug from the Matrix, have you checked out our <https://docs.livekit.io|docs>?

you can use this for position + text messages or any sort of arbitrary data/strings:
<https://docs.livekit.io/client/data-messages/>

also, here’s a spatial thing I did a year ago which may help (but doesn’t use Web Audio API methods):
<https://github.com/livekit/hq>

Oh my this got lost in my notifications. The GainNode can be used if all you care about is adjusting volume spatially.

I have a sample app rolling out soon that uses the `PannerNode` which gives you stereo sound

<https://docs.livekit.io/oss/deployment/>

Also any suggestions on how can I achieve full reconnection , when ausubscribe is off and adaptivestream is off.

which version of the SDK are you using? are you able to reproduce this with the JS sample app?

I think the new engine with new Peerconnections dont have access to old senders

This happens when this log appears
`livekit-client.esm.mjs:19726 leave-reconnect disconnected`