To let many users look at different recordings in a browser, give each of them a web Viewer link to their recording.
The link names the data in its url parameter, so you only need to host the data, not one Viewer per user.
A gRPC proxy shows the same data to everyone
rerun --serve-web, rerun --serve-grpc, and rr.serve_grpc() start a gRPC proxy.
The proxy sends every message it receives to every connected Viewer.
It has no per-user sessions: all users see the same stream.
The proxy also keeps a buffer of past messages, so that a Viewer that connects late gets the history.
The buffer is limited by --server-memory-limit (or server_memory_limit in serve_grpc), 1 GiB by default, and the oldest data is dropped when it is full.
This is a good fit for a live demo or a shared screen, but not for giving each user their own data.
Give each user a link to a file
Save each recording as an .rrd file (for example with rr.save()), and host the files on any HTTP server or object store.
Then give each user a link with the file's URL in the url parameter:
https://rerun.io/viewer?url=https%3A%2F%2Fdata.example.com%2Fruns%2F42.rrdSee Open a file by URL for the details: URL-encoding, presigned URLs, and the CORS headers that the file host must send.
Experiment trackers such as MLflow or Weights & Biases need nothing special: store the Viewer link with the run, for example as a link or an HTML artifact.
Give each user a link to a catalog server
If you have many recordings, or want users to browse and query them, put them on a catalog server and link to each recording (segment) there.
The open-source rerun server serves a directory of .rrd files; Rerun Hub is the managed version.
Generate segment URLs in dataframes shows how to create a link per segment.
Put the segment URL, URL-encoded, in the url parameter of the web Viewer page.
The server only accepts browser requests from pages on localhost, 127.0.0.1, and rerun.io (and its subdomains) by default.
If you serve the Viewer from your own domain, allow it with rerun server --cors-allow-origin "https://viewer.example.com".
Where to host the Viewer
- Use the hosted Viewer at
https://rerun.io/viewer. - Embed it in your own page with an
<iframe>or the JavaScript package. See Embed Rerun in Web pages. - Serve it yourself with
rerun --serve-web, which hosts the web Viewer on port9090(--web-viewer-port).
Use a Viewer version that can read your files. See RRD file backwards compatibility.