Troubleshooting

Start with the message shown in the app. 3sparks Chat uses specific states such as unreachable server, connected with no models, no selected model, and selected model unavailable.

Connection problems

• Confirm the provider API key or local server address.
• Use Fetch Models after creating or editing a connection.
• For local servers, use a LAN IP or Tailscale IP instead of localhost.
• Check firewall prompts on the computer running a server.

Model problems

• If no model is selected, choose one on the connection or in chat settings.
• If the selected model is unavailable, use Switch Model or fetch models again.
• If context is too large, lower token limits or remove large attachments.
• If vision fails, choose a model that supports image input.

Replies that stop when you leave the app

• On iOS 26 and later, replies keep streaming in the background. See Background Generation.
• On iOS 25 and earlier, and on visionOS, iOS suspends the app after roughly 30 seconds and the reply stops.
• A very long generation can still be ended early by the system background budget.
• Force quitting the app from the app switcher ends a running generation immediately.

Sync problems

• Confirm iCloud is available on each device.
• Restart the app if sync activation says restart is required.
• Use Sync Now or reopen the app when data appears stale.
• Review duplicates or conflicts carefully.

Keyboard problems

• Confirm the keyboard is added in iOS Settings.
• Enable Allow Full Access.
• Choose a reachable keyboard connection and model.
• Remember on-device Local Models are main-app only.

Local Model problems

• Keep enough free storage for downloads.
• Start with smaller models.
• Reload the model if generation fails.
• Delete and redownload if a model or projector file appears corrupt.