Aura Translation

Problem

Translation is usually treated as a destination: open a website, move text into a form, wait, then move the result back. That interruption is expensive when translation is part of reading or writing rather than the main task.

Aura Translation explores a quieter model. It stays in the system tray, works with text already on the clipboard, and presents the result in a small window near the current point of attention.

Approach

The product is built around two explicit entry points. A configurable global hotkey translates the current clipboard content on demand. Optional Aura mode watches for fresh copied text and starts translation automatically when the platform supports it.

The result streams into a floating bubble instead of a full application window. It can disappear after use, remain pinned for comparison, or become a compact draft composer for revising and translating the source again.

Architecture

Svelte owns the translation and settings interfaces while Tauri exposes a narrow desktop command boundary. The Rust core manages global shortcuts, clipboard behavior, streaming provider requests, capabilities, profiles, history, and window placement.

Provider integrations use an OpenAI-compatible interface, allowing the same workflow to target hosted providers or a local Ollama endpoint. Preferences and recent results stay on the device.

Product Decisions

• Translation is initiated by an explicit hotkey or an optional, visible clipboard mode.
• Paste-back into the source application was removed in favor of an explicit copy action.
• Named profiles keep provider and language combinations easy to switch without turning the bubble into a settings surface.
• Token usage is shown when a provider supplies it, making the cost of a request observable.
• Aura Guard can skip clipboard content that resembles credentials before an automatic request is sent.

Current Boundaries

macOS is supported for current development and Windows is covered by project release workflows. Linux is not yet productized. Aura has no hosted backend; translation text is sent directly from the desktop app to the provider selected by the user.

API keys use the operating system credential store where supported. Plaintext configuration remains an explicit fallback, not the default path.

Next Steps

The next product phase is less about adding providers and more about release confidence: clearer onboarding, stronger cross-platform verification, and a small set of workflows that remain dependable while the app runs quietly in the background.