Troubleshooting

• Open Settings → AI Providers. Make sure the active tab matches the provider you want to use.
• For OpenRouter: paste the key, save, then wait for the model list to load before picking a model.
• For Ollama or LM Studio: confirm the server is running and reachable. Use the test button rather than guessing.
• For Built-In on Free: check the usage meter in the sidebar footer. You might be out of monthly tokens.

• Switch to a model with a larger context window.
• Trim long persona system prompts or scenario base prompts.
• If you're running locally, check the server has enough RAM/VRAM for the model.

• Use Chrome, Edge, or Opera. Safari and Firefox don't support Web Bluetooth.
• The page has to be served over HTTPS (or be on localhost). A non-secure proxy will block the picker.
• Restart the device in pairing mode before clicking Scan for Devices.

• Open the extension's detail page from My Toys and check the Personas section — is the persona you're chatting with toggled on?
• Confirm the device shows as connected in the top bar (green status, not gray).
• Try the Test action on the extension settings page. If that doesn't move the toy, the AI's tool calls won't either.

• Check the battery indicator on the Devices tab (V3 only). Low battery causes drops.
• Stay within a couple of metres of the computer, line-of-sight if possible.
• Other Bluetooth peripherals on the same machine can interfere — headphones and watches are common offenders.

• Open the extension on its category page (My Toys, My Utilities, or My Integrations) and check the Enabled here filter.
• From the detail page, scroll to Personas and confirm the current persona has the extension toggled on.
• If global settings ask for a key, endpoint, or credential, fill them in. Extensions don't run with empty config.

Open the extension card and click Update. Nothing auto-applies, so you get to read the changelog before installing.

• Turn on Developer Mode and Show developer features under Settings → Advanced.
• Pick a local extension folder. Each extension is its own subfolder with a manifest.json and a JS module.
• Click Scan workspace.

• Check Settings → Image Generation → Photo Behavior. Turn photos off is the obvious culprit; Send pics every few turns needs enough turns to have elapsed.
• For ComfyUI or Automatic1111, confirm the server is running at the configured URL (defaults: http://localhost:8188 and http://localhost:7860).
• Automatic1111 must be started with --api.
• For Cloud images, adult content needs NSFW enabled under Content.

• Edit the persona's Visuals and Advanced Settings tab. Look at the base prompt, default clothing presets, and any attached LoRA files.
• If a scenario is active, its Environment and Clothing fields also feed the prompt.

• If you're in Local mode and cleared browser data, the chats really are gone unless you have an export.
• If you signed in from a different browser, profile, or device, sign in there to check.
• If you're in Cloud mode and you're sure the account is right, contact support before doing anything destructive.

The pencil icon next to a History row only appears in Cloud mode. In Local mode the session title is whatever the first message generated — the row groups under Today, Yesterday, or Earlier and you can delete but not rename.

• The import preview shows persona, session, message, memory, and summary counts. Zeroes mean an empty account.
• The archive's export date is shown too — useful for spotting a stale backup.

• Some scenarios stack base prompts, environment, character definitions, and triggers. All of that is sent every turn.
• Memories above ~50 entries grow the context noticeably. Trim or lower the importance of older ones.
• Image generation tools also consume tokens for the prompts they build.

The field shows a countdown to when you can change it again. There's no override; the cooldown keeps handles stable for other users who may reference yours.