Urik

Privacy-focused Android keyboard with swipe typing, custom layouts, and password manager support. No tracking, 100% on-device, and fully open source.

Status

Beta Software - In open beta testing.

Features

Input

• Swipe typing with geometric path matching and vertex detection
• One-handed and split modes with coordinate transformation
• Adaptive layout for tablets and foldables
• Long-press for character variations and custom key mapping
• Long-press comma key to switch keyboards (IME picker)
• Spacebar swipe for cursor control
• Backspace swipe to delete previous word
• Optional number row
• QWERTZ, AZERTY, Dvorak, Colemak, Workman, Bds and Hcesar layouts
• Hardware keyboard detection

Intelligence

• Local bigram model for next-word prediction
• On-device spell checking using URIK (Ultra-compressed Ranked Input Korpus) — a custom binary dictionary format with Levenshtein automaton traversal
• Autocorrect with configurable pause-on-misspell; skips URLs, emails, and punctuation
• Word learning with encrypted SQLCipher database
• User-specific word frequency tracking
• 19 supported languages with dedicated language toggle button
• Merged dictionaries mode for multilingual typing sessions
• Emoji search with keyword support

Integration

• Inline autofill support for password managers (Android 11+)
• Clipboard history with encrypted storage

Customization

• 15+ built-in themes with favorites and Material You dynamic theming (Android 12+)
• Custom long-press symbol and key assignments
• Haptic feedback strength slider
• Auto-capitalization toggle
• Configurable key size, key label size, spacebar size, and cursor speed
• Manage learned words: browse, search, and delete entries
• WCAG AA contrast and TalkBack support

Privacy

• No telemetry, analytics, or network permissions
• Local on-device processing
• Encrypted local storage for user data (AES-256 + Android Keystore)
• Biometric authentication for accessing learned words
• Dictionary export and import for backup or migration
• Bigram predictions and word frequencies remain local-only (not exported)

Privacy

All processing happens on-device. No data leaves your phone. User dictionary and learned words are stored in an encrypted local database.

We answer to our users, not investors.

Requirements

• Android 8.0 or higher (API level 26+)

Building from Source

git clone https://github.com/urikdev/Urik.git
cd Urik
./gradlew assembleDebug

Requirements: Android Studio Panda+, JDK 17+, Android SDK 36+

Contributing

See CONTRIBUTING.md for guidelines.

Security

Report vulnerabilities via SECURITY.md.

Dictionary Data

Spell checking uses word frequency lists from FrequencyWords by hermitdave, derived from the OpenSubtitles corpus.

• License: CC-BY-SA-4.0
• Modifications: Filtered top-N by frequency per language, compiled to URIK binary format at build time

Emoji Annotations

Emoji search uses multilingual keyword annotations from Unicode CLDR (Common Locale Data Repository).

• License: Unicode License V3
• Modifications: Extracted emoji annotations, converted to searchable JSON format

Japanese Conversion Data

Kana-to-kanji conversion uses dictionary data from Mozc by Google LLC.

• License: BSD 3-Clause
• Modifications: Extracted reading/surface/frequency triples; converted cost values to frequency scores; filtered single-character hiragana entries (covered by the spell-check dictionary)

License

GNU General Public License v3.0