Personal AI Assistant
Turning everyday words, voice messages, and images into connected, editable personal records.
- Status
- In development
- Platforms
- macOS & iOS
- Built with
- Swift, SwiftUI
A thought is easy to capture.
Making it useful takes more work.
A conversation, a plan, a meal, an idea: everyday information rarely arrives neatly organised. This project starts with a simple question—can an assistant turn that input into useful records while keeping the original context?
The app captures text, dictated speech, and images, then connects the resulting notes in a user-controlled Obsidian vault. The native interface provides ways to explore people, events, goals, and the moments linking them.
Capture, connect, and stay in control
- CaptureKeep the original words or image.
- CheckValidate the proposed records against their source.
- ConnectSave linked Markdown records in the vault.
- CorrectEdit the result or undo the capture.
The engineering work centres on the boundary between interpretation and action: what did the input actually say, what can safely be inferred, and what should remain unresolved?
A few decisions that matter
The source stays with the record
Original inputs are retained in History. Evidence checks help constrain dates, events, and measured values to what the source supports.
Automation should be reversible
Duplicate checks, staged saves, corrections, and edit-aware Undo help protect the vault when interpretation fails or a record changes later.
Ordinary files, native interfaces
The records remain Markdown in an Obsidian vault. Mac and iPhone interfaces share capture logic and SwiftUI features, with platform-specific input where needed.
Where it is now
A working native prototype, developed with AI-assisted coding and automated regression checks. Device-specific capture, cross-device behaviour, and Calendar/Reminders integrations are still being validated.
Interpretation currently uses a model provider. Local model experiments are exploratory; this is not yet a fully offline assistant.