LiveKit Community

has renamed the channel from "webrtc-chat" to "lounge"

Imagepipe_17.jpg

Hey <@U067EKS4BD2>, thanks for this note. Let us know if you have any questions as you evaluate or prototype with LiveKit. 

We’d of course love to help as many folks as possible make a smooth transition from Twilio over the next year. 

What do you think would be most helpful (in terms of information and/or resources) for us to provide towards that process?

In our case, we're actually rebuilding our video call functionality from scratch for v2 of our platform, and assuming that takes less than a year, then we wouldn't need any migration from Twilio guides. We'd just need LiveKit to work well and be awesome.

However, in the same email I received there's a link to a migration guide to Zoom <https://developers.zoom.us/docs/video-sdk/twilio/> with which you may find something useful.

I think you should boost awareness of your platform by emphasising your SDKs and guides for frameworks beyond JS, native Android and iOS. Because these are the only ones offered by Twilio Video. Zoom offer broader integration options, but I get the feeling there's a lot of people out there who prefer to just not use Zoom for whatever reasons they may have.

image.png

:wave: I didn't design this specific diagram, but we generally use a combination of <https://www.figma.com/figjam/|Figjam> and <https://www.figma.com/|Figma>. Figjam makes it easy to quickly diagram things with boxes/arrows. To get our own unique style we'll bring the diagram back into Figma to do custom styling and tweaks which provides more granular styling tools. Hope that helps!

Are there any plans for an Elixir client/sdk? :smile:

it's not currently on our roadmap but it does come up in discussion from time to time. some developers have just interacted directly with our <https://docs.livekit.io/reference/server/server-apis/|server API>, but you'll still need a realtime client.

can you share more about your application and how you're hoping to integrate LiveKit?  that sort of information always helps us to prioritize!

<@U0760MHCHQD> 
I’m also interested in an elixir client. 

My use case is an agent. I’m building a mock interview (<http://InterviewGenius.io|InterviewGenius.io>) where we’ll let anyone practice mock interviews. The goal is to make it feel as realistic as possible. 

I already have an existing implementation that’s 100% elixir (using membrane). It’s a Santa app (<http://callsanta.ai|callsanta.ai>). Though I’m looking to simplify the stack and move the webrtc pieces off of our plate. 

I looked through the docs, and it seems like I’d need a client. The server side APIs only manage the room creation, not the generation of responses. 


That’s right - you’d need a client for this. We’ll keep you posted if there’s any progress on elixir!

I’d be open to writing my own client if the APIs are documented. Is that a possibility?

I don’t think it’s well documented, but all of our more recent clients (python, nodejs, and unity) are based off the common core of our rust SDK which should make things simpler. I’m sure we’d do the same with elixir when we take it on. You might look at those and see what it would take. 

Let's gooo <@U01KZ0SCDKL> :smile:

<https://www.youtube.com/watch?v=RNebPdvPB7k>

Hey,  so i have been playing with the JWT, and it appears a small TTL of less than a minute isnt respected tokens are valid upto a minute

second of all the application I am trying to develop allows users to kick people out of the stream but with the refresh tokens having a exp of 10 minutes... a user could grab this JWT and rejoin

any thoughts?

There are different approaches to this, if TTL isn't a solution. One way would be: knowing that each token has a unique `identity` , you could keep track of the kicked out identity within the roomData, or in an external DB, to validate if the token is allowed to rejoin. This would be handled within your app logic.

hi, been facing this issue with invites, tried inviting multiple teammates and all get the same page after login

Google Meet has recently added their “Adaptive Audio” feature which detects if multiple people are in the same room and then automatically chooses which mic to use and which one to mute.
I believe this feature is implemented in their SFU so I was wondering if you guys have considered anything similar for LiveKit?
See more: <https://support.google.com/meet/answer/14263133>

No David. We are not looking at it currently. It is a rare use case (although I am not sure with more people returning to office). It is also quite involved and uses ultrasound on client side to actually detect other devices and uses smart switching to have only the active microphone enabled - <https://blog.google/products/meet/adaptive-audio-google-meet/#:~:text=The%20team%20adopted%20a%20new,without%20turning%20any%20speakers%20off>.

Yeah I thought it's pretty complicated.
Thanks for the quick answer, it's also not a big use case for us, I was just curious.

Actually, this company <https://www.around.co/> did it long before anybody else. Have no idea why they did it and if they really had a customer/a lot of demand for it or did it predicting it would be a big thing.

Hmm, I think there is possibly a "simple" ways to do this. Just brain storming some ideas here. But I wonder if the machines could play a tone that is just above the range of human hearing. Then any machines that receive the tone can mute the track or the participant that generated the tone. There would need to be some logic for how to identify local tone producers but I think it is not too hard.

Doing full on echo canceling I think would be too much processor to work

Just need to figure out the optimal sound. I expect it is range of tones and range of volumes:

From <https://blog.google/products/meet/adaptive-audio-google-meet/|article>:
```Then they tried using ultrasound. By emitting high-frequency sounds undetectable to the human ear, the laptops can identify the presence of other laptops in close proximity and begin acting together as a group. This eliminated the need for users to manually configure their devices or select the room they were in. "But it was really tricky because the ultrasound needed to work reliably on any device, and be precise — if audio leaks from the room next door, it shouldn't think you're in the same room," Henrik says. The team adopted a new type of ultrasound to increase accuracy, and tuned the frequency and volume to optimize reach without being audible.```

same_room_detection.MP4

Cobbled together tone generation and detection into setting menu. Now need to sort out how to do station identification so client can unsubscribe from near participants track.

For transmitting identification I'm not sure if it should be over the data message, or some sort of FSK, or some sort of time deviation slotting, or something else all together. I sort of like FSK with the tone because then station could be identified (maybe) before they even join the room.

To play you need to set .env.local to:

```NEXT_PUBLIC_SHOW_SETTINGS_MENU=true
NEXT_PUBLIC_ULTRASONIC_DETECTION_ENABLED="true"
NEXT_PUBLIC_ULTRASONIC_FREQUENCY="20154"  # 21kHz as default```
hacked up code <https://github.com/livekit-examples/meet/compare/main...yepher:meet_same_room_beacon:main|here>


The video is a little hard to see but basically I have the settings->effects window open. Then in client B I click the detect local button then in client A I click the generate tone button. As soon as I click the generate tone button the listening client pops up an alert box that it was detected. Second time around I show the FFT to see that it is a narrow tone.

I played with embedding identification in the beacon with this toy teletype I put together but it is misses most data currently: (here is some version of it)
<https://yepher.com/ultrasound-comms.html>

I need to read <https://eudl.eu/pdf/10.1007/978-3-030-38819-5_12|UltraComm: High-Speed and Inaudible Acoustic Communication> and see if that has any good tips. They claim 16k bps

Of course data channel would be a much easier and reliable solutions :slightly_smiling_face:

Fun little project :slightly_smiling_face:

Hey all! Just dropping the message here because in the channel <#C088ZNU7QQ5|> the `Run-LLM` is not responding at all.

We want RunLLM , Livekit Assistant is Dumb , Dumber and Dumbest.

We have been trying to respond to the partnership email to Michael and it’s not delivering. 
Is anyone else facing this? 

Email: <mailto:michael@livekit.io|michael@livekit.io>

`<http://mx.google.com|mx.google.com> rejected your message to the following email addresses:
Michael Engle (<mailto:michael@livekit.io|michael@livekit.io>)
There's a problem with the recipient's mailbox. Please try resending your message. If the problem continues, please contact your email admin.

 
<http://mx.google.com|mx.google.com> gave this error:
The email account that you tried to reach is inactive. For more information, go to <https://support.google.com/mail/?p=DisabledUser> 8926c6da1cb9f-4f26371b2f1si8888765173.63 - gsmtp`

Hey <@U07TMEGRG21>! You can reach out to me instead at <mailto:varun@livekit.io|varun@livekit.io>

Hey, is the RunLLM, in the <#C088ZNU7QQ5|> channel down? It is not able to answer the question/ queries. It keeps giving the same answer to any question asked. Just dropping the message here to let you know.

Hey, any road map on when agent 1.0  will be available for Node.js?

If you are working on a production grade agent we recommend you use the Python Agent library.

Makes sense, thank you.
Node agent feels about a decade behind Python, and I somewhat hoped it would see improvements this year, but it seems that isn’t the case.

The agent-js team is in the process of giving it a lot of love but it will take time to bring it up to speed

Appreciate the update. If the agent-js team needs real-world feedback or testers, happy to help. Just trying to avoid having to learn Python out of desperation :smiley:

This is interesting. You can chat with openSource GitHub repos by putting `<https://deepwiki.com/>` in place of the `<https://github.com/>`. For instance to chat with LiveKit Repo:

*Agent:* <https://deepwiki.com/livekit/agents>
Python SDK: <https://github.com/livekit/python-sdks>
LiveKit: <https://deepwiki.com/livekit/livekit>