PocketLLM Server

Support

Get help

Open an issue on GitHub: github.com/renjithsmenon/pocketllm-examples/issues

You’ll need a free GitHub account. Choose the template that fits: bug report, model compatibility, or question.

Issues are public. Never paste your API key, your prompts, or anything personal. An API key in a public issue lets anyone on your network use your server. If you shared one by accident, regenerate it in PocketLLM: Settings → API Key → Regenerate API Key.

Helpful details to include:

• iPhone model and iOS version
• PocketLLM version (Settings → About)
• The model you were using (name and size)
• What you did, what you expected, and what happened
• For connection problems: the client (Python, curl, an agent framework…) and the HTTP status code you received

Troubleshooting

Connection refused or timing out

• Check that the server is running and PocketLLM is open in the foreground. iOS pauses the server when the app is backgrounded or the phone locks. Settings → Keep Screen On helps.
• Make sure your computer and iPhone are on the same Wi-Fi network. Guest and many corporate networks isolate devices from each other.
• Use the exact endpoint shown on the Server tab; the iPhone’s address can change when it reconnects to Wi-Fi.

401 Unauthorized The API key is missing or wrong. Send it as Authorization: Bearer <key>. If you regenerated the key in the app, update your clients.

429 Too Many Requests (model_busy) PocketLLM generates one response at a time. Send requests one after another.

503 Service Unavailable (model_not_loaded) No model is loaded. Load one in the Models tab.

400 tools_unsupported The loaded model’s chat template can’t express tool calls. Load a model whose template supports tools; each model’s details show this.

A model won’t load It may need more memory than iOS allows the app. Try a smaller model or a smaller quantization (Q4 rather than Q5/Q8), lower Context Size in Settings, or close other apps. Expand a model in the Models tab to see its estimated memory fit.

A download stops or fails Downloads continue while you use other apps, but restart if PocketLLM is force-quit mid-download. Tap Resume or Retry. Check you have enough free storage.

Frequently asked questions

Which models work? See compatible models. Single-file GGUF models with a mainstream architecture generally work; only models listed as tested have been verified.

How does pricing work? Try everything free for 7 days with no payment details. Then a one-time Lifetime purchase unlocks unlimited serving: no subscription. Without it, PocketLLM keeps working and the server runs in 10-minute sessions.

I bought Lifetime on another device or reinstalled. Open Settings → Plan → Restore Purchases while signed in with the same Apple ID.

I have a code. Open Settings → Plan → Redeem Code. Codes are redeemed through Apple’s own sheet.

Is my data sent anywhere? Inference runs on your iPhone; prompts and responses are never sent to us or any cloud service. See the privacy policy.

This site is open source. Improve this page.