Menu

Support

Sotto runs language models on your own device. There is no account and no server, so there is nothing to sign in to and nothing to be locked out of. If something is not working, it is happening on your device, and the answers below cover the cases that come up most.

Contact

contact@eonix.lk

Expect a reply within three working days.

Security reports

Email, not a public issue

See the security model.

Privacy

What leaves your device

The complete list is short.

When you write in, it helps to include your device and OS version, the Sotto version from Settings › Advanced › About, and which model you were using.

Getting started

Is Sotto free? Are there any purchases?
Sotto is free. There are no in-app purchases, no subscription, no paid tier and no ads. The only thing that can ever cost you money is a Google Cloud API key, if you choose to set up the optional search tool — that is billed by Google under your own account, not by Sotto, and Google’s free allowance is 100 searches a day.
Does Sotto work offline?
Yes. Inference runs on your device. Once a model is on the device, airplane mode changes nothing. The only features that need a network are downloading a model and any tool you explicitly set up to call the internet.
Do I need to download a model?
Not if your device supports Apple Intelligence and you have turned it on — the on-device model is ready with nothing to download. Otherwise pick an open-source model from the catalog, or import a .gguf file you already have. You do not need both.

Troubleshooting

“Turn on Apple Intelligence in System Settings”
Apple’s on-device model needs Apple Intelligence enabled and a device that supports it. Open the system Settings app, find Apple Intelligence & Siri, and turn it on. If your device does not support it, download an open-source model from Library instead — everything else in Sotto works the same way.
“Apple Intelligence is still downloading its model”
The system model downloads in the background after you enable Apple Intelligence. It can take a while on a slow connection. Try again later, or use an imported model in the meantime.
A model will not load: “not enough memory”
The estimate is the weights plus the KV cache. Either pick a smaller model or a smaller quantization, or lower the context length in Settings › Models. As a rough guide: a 1B model is comfortable on any supported device, 3B wants about 3 GB free, and 7B wants about 6 GB and is really a Mac model.
A download is stuck at “waiting to start” on iPhone
Settings › Privacy › Downloads on Wi-Fi only is on and you are on cellular. Either join Wi-Fi or turn that setting off. Model files are large — often several gigabytes — so the default is deliberate.
A download failed part way
Downloads resume. Reopen Library and press Retry; it continues from where it stopped rather than starting again. Check you have enough free space for the whole file.
Answers are slow
Speed depends on the model size, the quantization and your device. Smaller models are much faster. On a device under thermal pressure, everything slows down. The tokens-per-second figure in the inspector tells you what you are actually getting.
The model said something wrong, or something I did not want
Sotto does not review, filter or fact-check model output — no one does. Models invent facts confidently. Treat an answer as a draft and verify anything that matters. If a particular model behaves badly, try a different one; the catalog covers a range of publishers, and each model’s licence and publisher are listed next to it.

Your data

Where are my conversations stored? Can I get them out?
On your device only, in Sotto’s own container. Settings › Privacy › Export all conversations writes them to a JSON file wherever you choose. Erase all data in the same place deletes everything — conversations, personas, imported models, preferences and any tool API keys in the keychain — with no copy anywhere else.
I turned off “Store conversations” and my old chats are still there
That setting takes effect from the next launch and governs new conversations. Existing chats stay on disk until you erase them.
Can I lock the app?
Settings › Privacy › Require Face ID to open locks Sotto whenever it goes to the background.
What leaves my device?
Only four things, all started by you: a model download, the optional weekly catalog check (off by default), the Google search tool (off by default), and any HTTPS tool you create. Settings › Privacy shows a running count of the bytes Sotto has actually sent, so you can check this yourself. Full detail is in the privacy policy.

Tools

Sotto ships 25 tools that run entirely on your device. 4 are switched on after installing — date and time, calculator, unit converter, and a search of your own past chats — and the rest are one switch away in Tools.

You can also create tools that reach the internet. Those are off until you configure them, and default to Ask every time, which shows you the exact request before anything is sent. Setup instructions for Google search are on the tools page.

A tool stopped working

Open Tools and press Run once on it. That runs the tool directly, with no model involved, and shows you the real error — an expired key, a changed address, or a server that is down. Every tool is capped at 20 seconds, and a single reply may make at most 4 tool calls.

Turning tools off

One switch at the top of Tools disables the whole feature. A persona can also be restricted to local-only, in which case it never sees a tool that uses the network.

Reporting a bug

Email contact@eonix.lk with what you did, what you expected and what happened. If Sotto crashed and you have turned on Settings › Privacy › Crash reports, the report is waiting in Settings › Advanced for you to read and attach — it stays on your device until you choose to send it.