CLEAR BY DESIGN
Privacy
Updated September 15, 2026
ToFlux is operated by the individual developer 吴磊 (Wu Lei). This notice explains the main ways this website and the ToFlux apps handle data. For support or privacy questions, contact: wleo19x6@gmail.com
Visiting this website
This website introduces ToFlux and provides support information. It has no registration or file-upload form and does not include advertising or analytics SDKs. Sites hosts the website. The hosting service processes access records needed to serve pages, protect the service, and investigate problems.
Audio and local transcription
After you grant permission and connect your devices, opening the Watch app can start listening automatically. Live audio travels over an authenticated, encrypted local-network connection to your own Mac. The Mac transcribes it locally and returns text to the Watch. Audio reaches the Mac before you tap Send; the send controls determine when text is inserted or submitted to Codex. You can pause listening or revoke microphone permission in system settings.
Audio is buffered in memory during the session. The current app does not archive long-term recording files, upload live audio through CloudKit, or send live audio to a developer server or Codex for recognition. Drafts, text awaiting confirmation, and receipts may be stored locally on the Watch to display and restore state.
Device connections and first-time pairing
Connections use randomly generated device identifiers, device information, local-network addresses, and trust credentials. App settings are stored locally, and associated keys are stored in the system Keychain. These identifiers distinguish paired devices; they are not advertising identifiers.
First-time pairing uses the CloudKit private database of the same Apple Account to exchange device discovery information, pairing requests, identifiers, timestamps, and status. You compare six-digit numbers on both devices and approve the request on the Mac. Pairing configuration and credentials then use encrypted fields. The developer has no administrative interface for reading your CloudKit private database. Apple provides and processes the cloud service; private-database storage and field encryption do not mean Apple processes no service metadata.
The iCloud text channel
When the iCloud text channel is configured and available, confirmed text, the send mode, message identifiers and timestamps, receiver configuration, and processing receipts are stored in the same Apple Account’s CloudKit private database. They support delivery, duplicate prevention, and status checks. The app uses encrypted fields for relevant text and credentials. iCloud does not carry or recognize the current live audio stream. Apple’s rules also apply to the cloud service.
Mac input and Codex
The Mac app uses Accessibility permission to locate the target app and input field and enter text. It temporarily reads and writes the clipboard, then attempts to restore the previous contents if another action has not changed the clipboard. Send & Run also requests submission. Once text reaches Codex, its handling and retention depend on your Codex settings and the applicable service terms. ToFlux cannot retract or delete submitted content.
ToFlux currently has no advertising, cross-app tracking, or third-party analytics SDKs, and no feature that automatically uploads recordings or dictated text to a developer server. Support material you choose to send and diagnostics handled by the operating system are separate from automatic upload of dictated text by the app.
Storage and cleanup
Ending or canceling a session clears the related runtime audio state. This is not a guarantee of immediate, complete physical erasure from storage media. While running, the Mac app keeps up to 30 in-memory activity records containing text. These are replaced over time and released when the process exits. Persistent duplicate-prevention records contain identifiers, digests, status, and timestamps, rather than ordinary dictated text. Those records and connection information are not deleted when you pause listening.
While online, the Mac attempts to clean up historical cloud text commands created more than 24 hours earlier. Records may remain longer if the Mac is offline, cleanup fails, or the service is unavailable. This is not a guarantee that all cloud data is deleted after 24 hours, and it does not clear pairing records.
You can clear the Watch draft when no message is awaiting confirmation. Clearing it does not delete CloudKit commands or Codex content that was already submitted. Uninstalling an app does not guarantee removal of Keychain items, backups, or cloud data. There is currently no single delete-everything control. Contact the developer for cleanup guidance for your version. The developer cannot directly read or verify the contents of your private database on your behalf.
Permissions and choices
The app requests Watch microphone access, Mac Accessibility access, and applicable local-network or Bluetooth permissions as needed. Refusing a permission affects the corresponding feature. You can pause dictation, leave the dictation screen, revoke permission in system settings, or quit the Mac receiver. Stopping these features does not automatically delete previously saved drafts, credentials, or cloud records.
iCloud pairing and the text channel require an available Apple Account and iCloud service. ToFlux does not ask you to give the developer your password, verification codes, or pairing keys.
Contact and updates
If you contact the developer by email, 吴磊 (Wu Lei) receives your email address and the information you choose to provide, to reply and investigate the issue. Share only what is needed and hide private information in screenshots. You may use the support email above to request deletion of support material you provided. Such a request does not automatically delete data on your devices, in CloudKit, or in Codex.
This notice will be updated with a revised date when app features or service data handling change. New uses or additional permissions will be explained in the relevant information or workflow.