Skip to content

Troubleshooting

Common issues and step-by-step solutions to get your device working.

Common Issues

Device not powering on

Start with the requirements for your exact hardware.

Check power

Use the adapter, cable and power profile specified for the device. Check that connections are secure and the outlet works; a shared connector does not establish electrical compatibility.

Record the symptoms

Note the device model, revision, lights, sounds and recent changes. LED meanings and reset procedures vary by device.

Avoid destructive recovery

Do not factory-reset or reflash a device before confirming the supported procedure and its effect on your data. If hardware appears damaged, stop using it and contact support.

Network and MeshStack connections

Distinguish local connectivity from a peer connection.

Local connectivity

Verify the address and port supplied by the setup flow. Check the device network settings and router client list; do not assume a fixed hostname or IP.

Peer state

A signed-in or online MeshStack control plane does not mean the peer tunnel is established. Running describes the adapter; Connecting means the peer connection is not yet established.

Permissions and conflicts

Use the supported OS elevation prompt if required. Keep firewall and UAC protections enabled. Record other active VPNs for diagnostics rather than removing them automatically.

Useful diagnostics

Include the app version, OS, connection states and the time of the failure. Remove keys, tokens and private addresses when they are not needed for support.

Slow or unresponsive inference

Locate the delay before changing settings.

Identify the route

Record whether the request used a local runtime, mesh peer or hosted gateway. Check the actual model and the application error or receipt.

Check resources

For local inference, inspect the runtime memory use and hardware status using its supported tools. Reduce context length or concurrent requests before trying a smaller compatible model.

Restart deliberately

If a service is stuck, preserve useful diagnostics and use the documented restart procedure for that component. A restart is not proof that an underlying fault is resolved.

Heat and sustained workloads

Follow your product operating limits.

Ventilation

Keep the vents clear and use the placement instructions for your model. Do not rely on a single clearance or temperature threshold across all devices.

Reduce load

Pause a sustained workload if the device reports a thermal warning. Inspect reported temperatures using the supported tools and contact support if the warning persists.

Maintenance

Follow the device cleaning and service instructions. Do not open or modify hardware as a generic troubleshooting step.

Model loading and storage

Check the runtime, artifact and available space.

Storage requirements

Check the exact model artifact size and the runtime requirements, including download or conversion space. Disk size and inference memory are different constraints.

Failed downloads

Use the download method supported by the model source and installed runtime. Check available space, network errors and any provided checksum before retrying.

Before deleting files

Confirm which files are model artifacts and which are user data or configuration. Back up important data and avoid generic formatting or deletion commands.

API errors

Use the HTTP status and service-specific message.

401 or 403

Check whether the request needs a website session or an inference key. Confirm the key is active and has the required entitlement and scope; do not paste it into a support ticket.

Model unavailable

Refresh the endpoint model list and use an exact served identifier. Catalog presence alone does not establish that your key can successfully run a model.

429

A usage or rate limit was reached. Follow the response guidance and retry later; repeated immediate retries do not reset the limit.

Timeout or service unavailable

Check the selected service status and record the timestamp and request identifier if provided. Repeated errors should be escalated with redacted diagnostics.

Still having issues?

Our support team is here to help with any problems not covered here.

Submit a Support Ticket