Local computation
Kwip ML is a visual machine-learning builder and learning application from KWIP LLC. Its supported workflows compute locally on your Mac. KWIP does not operate a service that automatically receives your datasets, models, training results or lesson submissions from the app.
Kwip ML does not require a KWIP account and does not include an advertising or first-party analytics upload feature. This policy covers the application; visiting this website is covered separately by the KWIP website privacy notice.
Information the app processes
Depending on your choices, the app processes source files, column names and types, selected data, workflow labels and connections, configuration, generated code, prepared data, training metrics, predictions, tensor values (arrays of numbers), model weights and checkpoints. It also stores settings, lesson progress and local permission records. The app includes third-party runtime libraries; their license notices are included with the application. See third-party notices and source access.
Projects and run evidence are different. A project can refer to an external source while a run retains a captured copy of that data, prepared partitions and fitted preparation settings. Failed or cancelled runs may retain partial diagnostics. File hashes help verify consistency; they do not anonymize content.
File access and notifications
You choose external source files and output destinations through macOS file selection. Saved local access grants allow supported future access to those selections. You can review and revoke saved source access in Settings. Revocation does not delete the original source, earlier snapshots, caches or exports.
Completion notifications require an explicit enabling action and macOS permission. Their content is generic and does not include project names, file paths, metrics or source data. macOS notification settings and Focus can affect delivery.
Capture, retention and deletion
Activation-value and prediction-row capture can be configured for future runs; factory defaults enable both. Timing/memory measurements are optional and off by default. Project choices can override application defaults. Turning capture off does not erase previous observations, mandatory prepared data, fitted preparation settings, checkpoints or other replay evidence. Explicit previews may still show values.
Cache cleanup removes the records covered by that operation, not original source files or user exports. Resetting learning progress changes completion tracking and preserves editable lesson files and historical evidence. Source permission revocation is separate from deletion.
To remove saved work, manage its project, run and export files in their saved locations. Removing one copy does not remove backups, synchronized copies or exports in other destinations. Your selected folder or backup provider controls any separate copy.
Sharing and exported artifacts
Saving or exporting data, generated code, weights or model packages is a user-directed action. These artifacts may include labels, architecture, learned information and provenance. Default trained-model packages omit raw/prepared tables, source bindings and fitted preprocessing state, but their weights and code can still be sensitive.
Exported Python and inference programs are executable artifacts. They are not automatically installed as new app functionality. Review a destination and its access before sharing; the receiving person, application or service controls its copy.
Information you send to support
If you contact [email protected], KWIP receives your email address, message and any attachments you choose to send. We use that correspondence to respond and investigate your request. Email is handled through our email provider; it is separate from local app computation.
The app’s Support summary is a limited local file. It excludes names, identifiers, paths, timestamps, raw errors/logs, source rows, tensor values, graph/code and weights. It can include status, artifact counts, failure codes and selected execution/capture policy. Nothing is automatically emailed or uploaded. Other logs and archives do not inherit those exclusions.
Please avoid sending confidential data or credentials. Contact us to ask about or request removal of information you sent to support; we will explain any records we need to retain to handle your request.
Apple services
Apple handles App Store downloads, purchases and related account or transaction information under its own policies. Those services do not give KWIP access to your local training data. Optional macOS diagnostic sharing is controlled by Apple and your system settings.
Policy updates and contact
This policy describes the application’s local processing and support practices. If the application’s data practices change, we will update this page to explain them.
For privacy questions, email [email protected]. For help using the application, visit Kwip ML support.