bitter-gold-53424
07/16/2024, 3:43 AMfast-ice-43108
07/16/2024, 8:40 AMcalm-helicopter-50047
07/16/2024, 9:00 AMfull-dog-23669
07/16/2024, 1:51 PMcalm-helicopter-50047
07/17/2024, 4:34 AMadorable-house-73941
07/17/2024, 9:53 AMTRACK_PUBLISH_REQUESTED
to go to TRACK_PUBLISHED
?
We are seeing some clients that publish tracks, seemingly successfull (track published
in logs), but in the cloud dashboard only the TRACK_PUBLISH_REQUESTED
shows up, not the TRACK_PUBLISHED
. We have some mechanisms to detect stale tracks, which triggers in these cases, causing a re-try (track un-publish and new publish), resulting in the same error.orange-tiger-19162
07/17/2024, 12:02 PMorange-tiger-19162
07/17/2024, 12:15 PMgorgeous-elephant-19666
07/17/2024, 6:58 PMflaky-postman-7418
07/18/2024, 1:48 AMlittle-eve-27396
07/18/2024, 3:54 AMauto_disconnect
setting in JobRequest.accept
? I'm noticing the ROOM_EMPTY
and PUBLISHER_LEFT
options & wanted to better understand how they work, but i'm weirdly not able to find any uses of the field. seems like it's just set on AcceptData and that's it? 🤔 I'm probably missing something thoughlittle-eve-27396
07/18/2024, 4:19 AMparticipant_disconnected
when the client goes into RECONNECTING and participant_connected
when the client transitions into RECONNECTED (this would be the inverse of how the client observes the remote participants, so that would make sense). just want to confirm there isn't any additional state we get on the server to infer that a client is more likely to come back sooncolossal-electrician-36427
07/18/2024, 5:24 AMsome-petabyte-69250
07/18/2024, 8:30 AMsome-petabyte-69250
07/18/2024, 8:30 AMblue-glass-37058
07/18/2024, 10:46 AM{
"message": "HTTP Request: POST <https://api.openai.com/v1/chat/completions> \"HTTP/1.1 200 OK\"",
"job_id": "AJ_5jfPkbMHiSkN",
"pid": 27844,
"timestamp": "2024-07-17T09:51:44.652418+00:00"
}
I need to extract job_id
, `message`and pid
little-eve-27396
07/18/2024, 6:27 PMjolly-air-55852
07/18/2024, 6:31 PMParticipant.audioLevel
property. Does this property indicate the audioLevel
of the incoming audio, or the audio level of the mic for outgoing track.
• Track seems to have a muted
property, and an updateMuted
function to mute a track. So in order to 'deafen', is it right to go ahead and mute all subscribed tracks.
• finally, if a client 'deafens', what's a good way for others clients in the conference to know about it.breezy-alarm-70231
07/18/2024, 10:01 PMcolossal-electrician-36427
07/19/2024, 12:16 AMpolite-agent-58519
07/19/2024, 8:03 AMpython
assistant.say
await assistant.say("Welcome to a NetShow AI Agent Demo. Meet and talk with Bella. Say Hello! Bella. And enjoy your conversation.", allow_interruptions=True)
You can test it with the NetShow Phone Agent - AI Bella at (949) 990-3030.
The issue is that the phone agent's response time is 4 to 8 seconds to start the intro message. However, when using rules and extensions/rooms, the call is answered immediately and the intro message plays as soon as it answers. I'm trying to replicate this for agents without rules.
For example, with AI Agent Rooms at 949-990-5100, Room pin: 1234, it takes the standard four to six seconds to say the welcome message, and then the agent finally responds. This delay isn't acceptable.
I'm concerned about this 4 to 6-second delay for it to say "Hello." I don't think it's an issue with my server. With the number 650-680-4883, your setup plays the agent right when the phone answers within a second and a half, without using the default "hello" in your file. Could you share the Python file for this particular agent connection or something similar?
I want to start selling our advanced web agents and Twilio/LiveKit phone agents, which work together on workflows, function calls, and tasks.
Here is my server hardware for the LiveKit setup and testing workflow:
- 8 GB Memory
- 2 Intel vCPUs
- 160 GB Disk
- SFO3 - CentOS 8 Stream x64
I also tried the simple TTS example from here:
https://github.com/livekit/agents/blob/main/examples/text-to-speech/openai_tts.py
The simple TTS takes 4 to 5 seconds for the first word.
Here are some timing comparisons:
- LiveKit - AI Bella (949 990 3030): First audio in 4 seconds
- LiveKit - AI Larry (949 990 4866): First audio in 4 seconds, goodbye in 6 seconds
- LiveKit - AI Interviewer (949 990 5100): First audio in 0.2 seconds, says "enter room pin and press hash to confirm," then says hello in 4 seconds, total of 18 seconds
- Twilio - Test AI Assistant Agent (+1 949 828 2564): First audio in 0.2 seconds, says "Hello, Bella is here, how can I help you today," goodbye in 6 seconds
- LiveKit - Demo Agent (650 680 4883): First audio in 1.5 seconds, live agent response in 2 seconds after question or hello
Please let me know my best option other than switching to Telnyx and not using Twilio unless that's the only fix; thanks again.
-----
Please allow two other questions, if you can:
2. The agent audio is coming out at a lower volume level. Can we control the audio volume level in the code?
3. How can I obtain the complete audio and transcript of a room session from the backend?
Does LiveKit provide the full audio of the session? If not, what alternative solutions can we use to capture both the audio and the transcript?
Thank you!breezy-alarm-70231
07/19/2024, 9:17 PMhelpful-lifeguard-86445
07/20/2024, 6:23 AMstrong-toddler-72393
07/20/2024, 6:06 PMimportant-hair-14673
07/21/2024, 5:16 AMimportant-hair-38324
07/21/2024, 11:22 PMrich-sandwich-34942
07/22/2024, 10:25 AMroom disconnect
delegate on iOS, but I have noticed that it does not operate in real-time. To detect when a user kills the task, I have integrated a separate socket server during calls. However, the socket disconnect
event is frequently triggered when the app is in the background. When the microphone mode is on during a LiveKit call, this issue does not occur because the OS prioritizes processing in Background Mode.
However, it appears that when a user mutes their microphone, LiveKit internally disables iOS’s AudioEngine
, causing the app to exit microphone mode. Consequently, when the app transitions to the background in this state, the session disconnect
event is often triggered.
My objective is to keep the local microphone active (without sending audio to the server) when a user mutes their microphone. Is there a way to customize the settings to achieve this behavior on iOS?
Thank you for your assistance.
Best regardsfull-iron-41629
07/22/2024, 10:13 PMthankful-branch-77619
07/23/2024, 12:32 AMhelpful-lifeguard-86445
07/23/2024, 7:00 AM