Troubleshooting

Fix LM Mini connection errors, Home sync, context-full chats, voice, tools, and common setup issues.

Can't connect to LM Studio

1. Settings → Provider is actually LM Studio (on-device/cloud will ignore that URL)
2. In LM Studio, the server is running and Serve on Local Network is on
3. Phone and PC share Wi‑Fi (guest / AP isolation blocks LAN)
4. URL looks like http://192.168.x.x:1234 — never localhost on the phone
5. If LM Studio requires auth, paste the API token
6. VPN/corporate filters — try without VPN, or use remote access

No model at all? Use Get Started to download an on-device pick or connect a computer — Getting started.

Connect

On Mac: LM Mini Home → Share with phone. Connect for Mac is deprecated through the end of 2026.

1. Re-scan the QR (old pairings go stale after reinstall)
2. Confirm the Mac can chat locally first (a model in Home, or LM Studio/Ollama on that Mac)
3. Keep Share with phone running — menu bar is enough; don’t quit Home
4. Windows/Linux: LM Mini Connect 1.5+ (newer QR for multi-backend)

Phone and Mac sync

Toggle lives at Settings → Remote Access → Sync chats with Home (Home pairs only).

1. Share with phone is on
2. Toggle sync off and on to retry
3. You should see “chats and folders are now in sync”, or a prompt to open Share with phone

Sync copies chats and folders, plus the personas you pick under Sync personas — not settings or models. Guide: Phone and Mac sync.

On-device model won't start

1. Download finished (not paused) and selected in the on-device picker
2. Enough free storage and RAM for that quant
3. MLX-only models need supported Apple Silicon; otherwise pick GGUF
4. After download on Android, open a new chat if an old empty-model state lingers

Context is full / prompt too large

1. Check the context bar in chat
2. Start a new chat, or pick a model with a larger window
3. With Pro: Settings → Models → When context is full → Roll or Cut middle, or tap Compact in the composer. Free chats stop when the window is full

Details: Chat — when context is full.

Tools / Pro Search do nothing

1. Enable tool use is on
2. Model advertises tool calling (try a recent instruct model)
3. Pro Search needs Pro where required
4. SearXNG URL reachable from the phone
5. Watch the transcript for tool rows — if absent, the model never called a tool

Voice / mic problems

1. OS microphone + speech permission granted
2. Correct STT language
3. Kokoro on Mac / PC (Pro) needs Remote Access pairing with Share with phone (Home) or Connect on Windows/Linux, plus Kokoro actually running
4. On Mac Home, allow Speech Recognition in System Settings if Mac system STT is silent
5. Keep LM Mini open during voice mode — it runs in the foreground

Image generation errors

1. Backend URL reachable (A1111 API / ComfyUI)
2. Workflow is API format for ComfyUI
3. Checkpoint / LoRA names match files on the server
4. On-device SD only on builds that ship Core ML assets

App Lock / forgot PIN

1. Settings → Data → App Lock is a Pro setup. If the lock is already on, it still asks for the PIN after Pro lapses.
2. Wrong PIN: wait if the app briefly blocks more tries, then try again.
3. Forgot the PIN? If you signed in with Apple, Google, or email while the app was unlocked, tap Forgot PIN? on the lock screen: sign in again, open the verification email, then choose a new PIN.
4. Without a linked account, reinstalling LM Mini is the only way to clear the lock. The PIN is not in iCloud, Home sync, or cloud backup.
5. Face ID / Touch ID / fingerprint bouncing back to the lock screen was fixed in 1.9.8 — update the app.

Guide: Settings — App Lock.

Still stuck?

• Getting started and Providers
• Blog for the “why”
• Update the app; changelog
• In-app feature requests / support if it looks like a regression