[Feature request] read-only spectator/view link for live device screen

[hassineabd]

Problem

Today, viewing a device's live screen in the hub UI is tied to actually controlling it (tap/swipe/type). There's no way to let a second person just watch what's happening on a device — e.g. to show a bug to a teammate, pair on a manual test, or let a reviewer
follow along — without giving them control or requiring them to reserve/release the device themselves.

Proposed feature

A "spectator" / read-only mode for a live device session:

• The person driving the device (who holds the reservation) can generate a shareable link (or the hub exposes a "view" URL per device) that opens the same live screen stream in read-only mode — no tap/swipe/type controls, no reservation required.
• Multiple spectators can open the link concurrently.
• Spectator view should reuse the existing MJPEG/WebRTC stream rather than opening a second stream from the provider, to avoid extra load on the device/provider.
• Optional: expire the link when the underlying reservation ends, or make it explicitly revocable.

Implementation questions to discuss

• New endpoint/websocket that fans out the existing stream, or a UI-only permission flag on the existing session route (same URL, minus the control panel)?
• Auth: does a spectator need a GADS account (view-only role), or is a token-based link acceptable for a self-hosted internal tool?
• Where does this fit relative to the workspace/permission model — role per workspace vs. a one-off share token?

I'd like to implement this — happy to open a PR once we agree on the approach above
(endpoint shape + auth model). Let me know which direction you'd prefer.

[ilendemli]

I just wanted to request something similar. In our case, we would love to have a monitoring page that displays all device screens.

[shamanec]

Considering the different supported video stream types creating a page that displays all device screens will be tough, especially for WebRTC because the browser does decoding and will probably go OOM or just have very bad performance I think. But I can probably add a button for observing from the device list, something along the lines of:

1. If device is not in use - no observing, makes no sense
2. If device is being used for automation - anyone can observe the device screen
3. If device is being used by a person - ask for observing and the person should approve before it loads

This is just on the top of my head, but seems like the most reasonable approach, what do you guys think?
Also for the questions of the original issue, I dont like the idea of having links that anyone can use to observe a device, I think you should have a GADS account to be able to observe a device and this also solves the workspace/permissions model, why would you be observing devices that you do not have access to.

[ilendemli]

Maybe only selected devices from a list are shown instead of all. (Filter options?, remembering the selections?)

1. If device is not in use - we would like to see in which state the device is, even if it is not running
2. If device is being used for automation - anyone can observe the device screen
3. If device is being used by a person - imo no need to ask for permission, but sure if that's the way you want it to be

[shamanec]

1. I could add a button to observe the screen also for devices not in use but I dont see the actual value of this when you can just control it and check the state. Also this has another problem - device might be locked or nothing happens on the device so no frames are coming and in this case you will just see a black screen.
2. That seems reasonable
3. I think some kind of permission acceptance makes sense, even if this is an in-house company tool watching someone as they are working without their knowledge looks a bit like spying :D