Swarmbot shows a device online but no screen: find the failing layer with Diagnostics
Right-click a device in the Swarmbot wall and open Debug and Logs → Diagnostics. It checks the device channel, accSys, the 19012 route, protocol communication and a real image request; shared devices also check the platform, WebRTC and the owner-side device.
When a device appears online in Swarmbot but remains on “waiting for first frame,” will not open a stream, or shows video without control, run Diagnostics first. It works with USB, Wi-Fi ADB, ACC, Runner 19013 and shared devices. In the device wall, right-click the target, open Debug and Logs, and choose Diagnostics.
The check is read-only. It does not restart the phone, restart accSys or change device settings. Instead, it follows the same route Swarmbot currently uses and tests each layer. That gives you a more useful answer than a single online indicator.
Why “online” does not always mean “screen ready”
A device appearing in the wall usually means that Swarmbot has discovered it or still has its device record. A working screen needs more: the computer must reach the device, accSys must be available, the 19012 route must work, the service protocol must respond, and the phone must actually produce an image.
USB status is easy to overread. Android’s documentation says that the device state in adb devices means the device is connected to the ADB server, but it does not guarantee that Android is fully booted and operational. offline means the device is not connected or is not responding. See Android’s ADB device-state explanation
For that reason, Swarmbot does not stop after confirming ADB. Diagnostics continues through the actual screen path.
How to read the diagnostic result
The window shows an overall result, total duration and the status of each check. The visible view stays concise; focus on these layers.
1. Device channel
This answers: “Can Swarmbot reach the device?”
- USB: checks whether USB ADB is online and authorized.
- Wi-Fi ADB: runs a lightweight command to confirm that the network ADB connection can really execute work, rather than merely being discovered.
- Runner 19013: checks the Runner service and its command interface on a custom system.
- Direct ACC: if the device already exposes accSys on 19012, no separate ADB channel is required.
If this layer fails, start with the cable, USB debugging authorization, Wi-Fi route or Runner service. Later accSys and image checks may be skipped because their upstream channel is missing.
2. Runtime and accSys process
For devices where Swarmbot deploys accSys, the check reviews runtime compatibility and the accSys process. A direct ACC device has no ADB or Runner process-query channel, so “skipped” is expected there; the protocol check later becomes the source of truth.
Some customized Android builds hide process names or return unusual process output. Swarmbot treats a successful 19012 handshake plus a real image as stronger evidence. If both work, a missing conventional process name alone will not mark an otherwise healthy device as failed.
3. The 19012 access route
This answers: “How is the computer reaching accSys?” Each connection type has its own path.
- USB: verifies the ADB forwarding tunnel from the computer to port 19012 on the phone. If an existing forward is unreachable, Swarmbot rebuilds it once and checks again. Android documents
adb forwardas host-to-device port forwarding. See the ADB forwarding reference - Wi-Fi ADB: prefers a direct connection to port 19012 on the phone’s LAN address; the protocol layer can also identify the Wi-Fi ADB device channel.
- Runner 19013: Runner is used to discover, deploy and start accSys, while the screen service is still verified on 19012.
- ACC: checks the existing 19012 service directly.
When the device channel passes but this route fails, the problem is already narrowed to a port, tunnel or accSys listener rather than the entire phone being offline.
4. accSys connection
An open port is not enough. Swarmbot establishes the same service connection it uses for the device and sends a Ping, then displays the protocol round-trip time.
A passing result means that the control service can communicate. It still does not prove that the phone can generate a screen image, so Diagnostics proceeds to one final layer.
5. Real image request
Finally, Swarmbot requests a 256-pixel-wide JPEG diagnostic frame and shows a preview when successful. This is the most useful check for “online but no picture.”
- Success: the wall image path works. The report says whether it received a new frame or the most recent valid cached frame.
- Warning — screen asleep: accSys is still connected, but the device cannot currently produce an image. The wall uses the existing control channel to wake the display and immediately retry; restarting accSys should not be the first response.
- Warning — capture busy: the service and port are alive. The wall keeps the last valid frame and retries at a short interval, so this does not by itself mean the USB tunnel is broken.
- Failure: when the protocol works but valid images continue to fail, the detailed report helps isolate the device-side capture path.
For a problem limited to one phone, try Run again and observe whether waking the display changes the real-image result. If many devices fail at the same layer, check the shared computer network, ADB or sharing path instead.
Why shared devices have extra checks
A shared phone is attached to the owner’s computer, not the recipient’s. Swarmbot first checks the public-platform connection and owner presence, then the WebRTC peer, RPC, live-control, wall-snapshot and video data channels. Once RPC is open, it asks the owner’s Swarmbot client to run the same low-level checks on the real device and returns those results.
WebRTC’s official documentation explains that peers exchange connection information through signaling and use ICE with STUN or TURN to establish a usable path. Read the WebRTC peer-connection guide A connected public platform therefore confirms the signaling entry point, not that video and control data channels are already open.
Use these clues:
- Owner not online: keep the Swarmbot client that owns the phone signed in and connected.
- Peer connection failed: inspect signaling, ICE, TURN and DataChannel states in the copied report.
- Video works but control does not: check the granted control permission and live-control data channel.
- RTC connected but owner-side diagnosis failed: the issue has moved to the owner-side device path; read the checks prefixed with the owner side in the report.
For the full workflow, see How to share devices with Swarmbot.
When to copy the detailed report
The dialog intentionally shows short conclusions. Copy detailed report includes per-step technical details, connection type, timing, recent accSys lifecycle events and device runtime logs. Preview image data is omitted, and common token, password and API-key fields are redacted.
When asking for help, send:
- the complete copied diagnostic report;
- the approximate time of the problem;
- the Swarmbot device number and connection type;
- whether the failure was no wall image, stream startup, control, or one specific operation;
- whether waking the phone, running the check again or reconnecting the cable changes the result.
A screenshot saying “retry in 15 seconds” is rarely enough. Never include passwords, active share codes or API keys. The detailed report lets support distinguish a device-channel issue from accSys, image capture or WebRTC without asking you to repeat every step.
A faster troubleshooting order
- Right-click the target device and run Debug and Logs → Diagnostics.
- Find the first failed or warning check instead of reading only the overall badge.
- Fix connectivity when the device channel fails; inspect the port or tunnel when the 19012 route fails; wake the display and retry when protocol passes but image capture warns.
- For a shared device, read the path in order: public platform → owner presence → WebRTC data channels → owner-side device.
- If it still does not recover, copy the detailed report before contacting support.
For first-time setup, start with the Swarmbot connection guide. If connectivity is healthy and a later workflow step fails, switch to Swarmbot automation troubleshooting. Separating connection faults from workflow faults is usually more productive than restarting every component.