Privacy for the iOS app.

Who this covers.

This is the privacy policy for Sveda for iOS. Use this URL in App Store Connect. It is not the marketing-site notice and not Sveda Cloud.

The app is a native client for a Sveda server you choose. We do not operate a public chat backend for this app. Messages you type go to the server URL you enter.

Publisher questions: hello@sveda.dev. Questions about chats stored on a server belong to that server’s operator.

What the app stores.

On the device the app may store the server URL and an unsent composer draft in App Storage, and the access key, embed token, and a random visitor identifier in the iOS Keychain. Keychain items are marked this-device-only. The visitor identifier (a UUID, at most 64 characters) is sent with chat requests so history can stay attached to this install.

The app does not create an email or password account with us. That visitor identifier is a device identity for history, not a social login.

What leaves the device.

When you tap Open Chat, and later when you send a message or manage history, the app transmits to the origin you entered: the visitor identifier, the access key or embed token as authentication, the chat text you send, model and thinking choices, and history list, load, rename, and delete requests. The iOS composer does not attach files; if a connected server returns file parts in history, those names are shown in the thread.

We use that data only to run the client: connect, stream replies, and keep history on the server you chose. We do not sell it, use it for advertising, or track you across apps or websites. The app has no analytics, advertising, or social SDKs.

Third-party AI and other operators.

The operator of the server you connect to is the controller of data stored there. That operator may send prompts and related metadata to AI model providers they configured — for example OpenAI, Anthropic, DeepSeek, or another compatible endpoint. Those providers process the data under their own terms.

We do not receive those payloads. We do not share iOS app chat with analytics vendors, advertising networks, or other third-party SDKs, because this app includes none. Equal protection under this policy therefore applies to processors we would control; we do not appoint any for this app. The server URL and model providers are chosen by you (or by the operator of a server you chose). They are not our subprocessors, and we cannot bind them to this notice. If you run the server, their policy is yours. If you connect to someone else’s Sveda server, that operator’s policy applies in addition to this one.

Consent.

Entering a server URL and tapping Open Chat is your permission to send the data above to that server, including onward transfer to AI models that operator configured. You can withdraw by switching runtime or using Settings → Reset this device. After you leave, the app does not send new messages to that origin.

Retention and deletion.

On the device, credentials and the visitor identifier remain until you reset them or delete the app. An unsent draft remains until you send it, clear it, or reset. On the server, chat history remains until you delete individual threads, reset this device while the server is reachable, or the operator’s own retention applies.

Settings → Reset this device deletes chats for this visitor on the current server when it is reachable, then clears Keychain credentials, the visitor identifier, and local storage. If the server is unreachable, local data is still cleared; remote copies may remain until you or the operator delete them. Deleting the app from the device removes local data; it does not by itself wipe the server.

Children.

This app is not directed at children under 13. Assistant output is unrestricted and can include material that is inaccurate or unsuitable for minors. We do not knowingly collect personal information from children. If you believe we have, write to hello@sveda.dev.

Reporting assistant output.

Long-press an assistant reply, or open Settings → Report this reply. Submit stores the reason and the reply text on the server you are connected to. The operator of that runtime can read the stored reports.

Assistant output.

Model replies can be wrong or harmful. They are not medical, legal, or financial advice. Links inside replies are produced by the server you connected to.

Contact and changes.

Privacy questions about the iOS app: hello@sveda.dev or Telegram. The same address is the App Store support contact. We may update this page when the app changes. The date at the top is the current version.