rtsp-stream

rtsp-stream is an easy to use out of box solution that can be integrated into existing systems resolving the problem of not being able to play rtsp stream natively in browsers.

Table of contents

• How does it work
• Authentication
  • No Authentication
  • JWT
• Easy API
• Configuration
  • Transcoding
  • HTTP
  • CORS
• Run with Docker
• UI
• Proven players
• Coming soon

How does it work

It converts RTSP streams into HLS based on traffic. The idea behind this is that the application should not transcode anything until someone is actually watching the stream. This can help with network bottlenecks in systems where there are a lot of cameras installed.

There's a running go routine in the background that checks if a stream is being active or not. If it's not the transcoding stops until the next request for that stream.

Authentication

The application offers different ways for authentication. There are situations when you can get away with no authentication, just trusting requests because they are from reliable sources or just because they know how to use the API. In other cases, production cases, you definitely want to protect the service. This application was not written to handle users and logins, so authentication is as lightweight as possible.

No Authentication

By default there is no authentication what so ever. This can be useful if you have private subnets where there is no real way to reach the service from the internet. (So every request is kind of trusted.) Also works great if you just wanna try it out, maybe for home use.

JWT Authentication

You can use shared key JWT authentication for the service.

The service itself does not create any tokens, but your authentication service can create. After it's created it can be validated in the transcoder using the same secret / keys. It is the easiest way to integrate into existing systems.

The following environment variables are available for this setup:

You won't need the private key for it because no signing happens in this application.

Easy API

There are 2 main endpoints to call:

POST /start

Starts the transcoding of the given stream. You have to pass URI format with rtsp procotol. The respond should be considered the subpath for the video player to call. So if your applicaiton is myapp.com then you should call myapp.com/stream/host/index.m3u8 in your video player. The reason for this is to remain flexible regarding useability.

Requires payload:

{ "uri": "rtsp://username:password@host" }

Response:

{ "uri": "/stream/host/index.m3u8" }

GET /stream/host/*file

Simple static file serving which is used when fetching chunks of HLS. This will be called by the client (browser) to fetch the chunks of the stream based on the given index.m3u8

GET /list

This (kind of a debug) endpoint is used to list the streams in the system. Since the application does not handle users, it does not handle permissions obviously. You might not want everyone to be able to list the streams available in the system. But if you do, you can use this. You just have to enable it via env variable.

Response:

[
    {
        "running": true,
        "uri": "/stream/185.180.88.98-streaming-channels-101/index.m3u8"
    }
]

Configuration

You can configure the following settings in the application with environment variables:

Transcoding related configuration:

The project uses Lumberjack for the log rotation of the ffmpeg transcoding processes.

HTTP related configuration:

CORS related configuration

By default all origin is allowed to make requests to the server, but you might want to configure it for security reasons.

Run with Docker

The application has an offical docker repository at dockerhub, therefore you can easily run it with simple commands:

docker run -p 80:8080 roverr/rtsp-stream:1

or you can build it yourself using the source code.

UI

You can use the included UI for handling the streams. The UI is not a compact solution right now, but it gets the job done.

Running it with docker:

docker run -p 80:80 -p 8080:8080 roverr/rtsp-stream:1-management

If you decide to use the management image, you should know that port 80 is flexible, you can set it to whatever you prefer, but 8080 is currently burnt into the UI as the ultimate port of the backend.

You should expect something like this:

Logs

With the UI solution the following files are created in /var/log:

• rtsp-stream-ui.err.log
• rtsp-stream-ui.out.log
• rtsp-stream.err.log
• rtsp-stream.out.log

Proven players

The following list of players has been already tried out in production environment using this backend:

• Angular - videogular
• React - ReactHLS

Coming soon

Codebase will be refactored as soon as I'll find some time to do it. 🙏 That will mean a major version bump. The goal is still the same. Keep it relatively simple and easy to integrate. Solve the issue of not being able to play RTSP natively in browsers.

Plans for the future:

• Throw out URI based directory creation completely
• Add better logging and debug options
• Separate HTTP from Stream processing completely
• Add option to remove streams from the client (Could be tricky, gotta figure out if this should be an option even if non-authenticated mode is used)
• Add better documentation about how to debug streams
• Add documentation about how to create issues
• Add guide for PRs