Feature Request: System-Wide Text Autocomplete and Autocorrection

[siyamsec-lab]

Summary

Add an optional, system-integrated text suggestion and autocorrection feature to AnduinOS.

Problem

Linux desktop environments currently provide inconsistent text prediction and autocorrection across applications. Some applications implement their own suggestion systems, while others provide little or no assistance.

This creates an inconsistent typing experience, particularly for users who frequently work with technical terminology, commands, documentation, and other specialized vocabulary.

Proposed Feature

Introduce an optional system-wide text assistance service that can provide:

• Autocomplete suggestions after entering 2–3 characters
• Context-aware word suggestions
• Basic spelling correction
• Fuzzy matching for common typing errors
• User-specific word learning
• Custom vocabulary support
• Keyboard navigation through suggestions
• Quick acceptance using Tab or Enter
• Per-application enable/disable controls
• A privacy-focused local mode that does not require sending typed text to an external service

For example:

User types:

sec

Suggestions:
security
secure
section
secondary

For a typo:

pythn

Suggestion:
python

Integration

The feature should ideally work across supported GTK/Libadwaita applications and, where technically feasible, other Linux applications.

The implementation should use an appropriate Linux text-input/accessibility mechanism rather than requiring each application to implement its own autocomplete system.

Possible integration points could include the existing desktop input stack, accessibility APIs, or an input-method architecture, depending on what is technically appropriate for AnduinOS.

Privacy

Privacy should be a core requirement.

The default implementation should preferably operate locally, with:

• No cloud dependency
• No transmission of typed text by default
• Local storage for learned vocabulary
• Clear controls for disabling learning
• Transparent configuration for users who choose to use external language/AI services in the future

Performance

The service should have minimal impact on:

• CPU usage
• RAM usage
• Application startup time
• Typing latency

Suggestions should appear asynchronously so that text input is never blocked while the prediction engine is processing.

Configuration

A simple settings interface could provide options such as:

• Enable/disable suggestions
• Enable/disable autocorrection
• Suggestion threshold
• Maximum number of suggestions
• Personal dictionary
• Learned vocabulary
• Per-application settings
• Reset learned data

Suggested Initial Implementation

A lightweight first version could use a local dictionary and frequency-based ranking with fuzzy matching.

More advanced prediction models could potentially be added later without changing the user-facing interface.

The feature could also be implemented as an independent package/service so that users who do not need predictive text do not have to run it.

Expected Benefit

This would provide a more consistent typing experience across the AnduinOS desktop while keeping the feature optional, lightweight, and privacy-respecting.

It could also provide a foundation for future local language-assistance features without requiring AnduinOS to depend on proprietary cloud services.

Acceptance Criteria

A minimal first release would be considered successful if:

1. Suggestions appear after 2–3 typed characters.
2. Suggestions can be accepted using the keyboard.
3. Common spelling mistakes can produce corrections.
4. Users can add custom words.
5. User vocabulary can persist between sessions.
6. The feature can be disabled globally.
7. The implementation does not require a network connection.
8. Typing remains responsive while suggestions are generated.
9. The feature integrates with the supported AnduinOS desktop applications without requiring application-specific modifications where possible.