livekit-recorder

Disclaimer

This project has been stable for our use at Cloud Ground Control. However, we will be moving to livekit-egress when it comes out. The good news is I will be contributing to that project, so for any suggestions, you can file it here or talk to us on Slack!

LiveKit Recorder

Service to record each participant in LiveKit room without transcoding. Prioritises minimum resource consumption for scalability. For composite recording (multiple participants in one layout), check out https://github.com/livekit/livekit-recorder.

Features

• Single track recording
• Track composite recording (video + audio)
• Docker support
• Upload to S3
• Structured logging
• Custom file name
• Job queue

Motivation

The original https://github.com/livekit/livekit-recorder was intended mainly for compositing participants, but consumes significant resources due to Chrome. We only want to record each participant, and that recorder was too expensive to scale. Our solution is to record each RTP stream via LiveKit's server-sdk-go.

How it works

The project has a service which is responsible for managing recordbots. The bots use selective subscription so we can have multiple bots in the room without duplicated recording, making it scalable through a Load Balancer. To stop the recording, either send a POST request to stop, or disconnect the participant from the room. We then use ffmpeg to containerise the output files.

Prerequisite

Make sure you have ffmpeg installed (https://ffmpeg.org/download.html)

Quickstart

First, start the server

LIVEKIT_URL=ws://... \
LIVEKIT_API_KEY=... \
LIVEKIT_API_SECRET=... \
APP_PORT=8000 \
LOG_LEVEL=debug \
go run main.go

Then, join the LiveKit room my-room as my-participant.

Next, create a POST request to /recordings/start with the body:

{
    "room": "my-room",
    "participant": "my-participant",
}

After recording for some time, to stop, either disconnect from the room, or create a POST request to /recordings/stop with the body:

{
    "room": "my-room",
    "participant: "my-participant"
}

You should have a file in the recordings/ folder.

Triggers

There are 2 ways to perform recording.

Manual

Use POST endpoints /recordings/start and /recordings/stop with the payload specified in quickstart.

Webhooks

The endpoint /recordings/webhooks will readily receive LiveKit's webhooks. Recording will start when receiving the event participant_joined and stop automatically on the event participant_left.

Environment Variables

Required

S3 upload

To enable S3 upload, make sure to set the following variables. Both region and bucket variables are required to create an S3 uploader.

For our use case, we have one bucket for different environments. If we specify S3_DIRECTORY=livekit and a file named my-file.mp4, the resulting file will be saved as livekit/my-file.mp4 on S3.

Deployment

We have shipped a Dockerfile which can be built locally. Unfortunately we don't have plans to have a DockerHub account, so you'll need to clone the repository, build the image, and push to container registry of your choice (ECR, etc.).

Issues

For any bug reports or the service not working as expected, either file an issue here or contact me in LiveKit's Slack. You can also reach out to me via email at dennis.wirya@advancednavigation.com .