v0.9.0

It's been a while!

This is a cumulative update for all betas released in the past few months. There aren't many major features this update, but many small improvements and fixes have been made app-wide.

I wasn't happy with the state of the app for good while due to many edge cases, crashes and SDK changes causing new bugs, but now most seem to be resolved. If you run into any problems, feel free to submit a new Issue!

Features:

• Updated llama.rn to support modern models up to Gemma 4
• Added NPU and OpenCL acceleration for Snapdragon 8 devices
• Added reasoning_content parser for Chat Completions
• Added Chat Text Size and Weight customization (modify in Settings > Chat Styling)
• Added option for separate dark and light mode themes
• Added chat import from JSON
• Added “Ask in ChatterUI” option, which adds ChatterUI as a text selection action in other apps. (enable in Settings > Chat > Ask In ChatterUI)
• Added Reasoning Effort, Reasoning Max Tokens, and Exclude Reasoning options in Samplers for OpenRouter compatibility
• Added “Fork” button in Quick Actions to clone a chat from a specific point
• Added “Keep Awake” setting to prevent screen from sleeping while the app is in the foreground
• Updated Quote parsing to support curly quotes
• Updated think tags to use more common thinking formats

Changes:

• Improved multiple markdown rendering styles
• Local Mode now automatically sets CPU affinity to use the fastest cores
• Most edit dropdowns are now accessible via long press instead of a dedicated edit button
• Migrated to a new PNG parser for character import and export.
• Character cards can now use more common image formats
• Tokenization is now debounced to reduce UI lag
• Notifications are dismissed when manually returning to the app
• Removed Presence Penalty from Google AI Studio API
• Unified bottom sheet components across the app
• Replaced TextboxModals with InputSheets
• Updated all context menus with a cleaner, modern design
• Chats now animate layout changes when messages are added, edited, or deleted

Fixes:

• Fixed crash caused by quote parsing
• Fixed issue where editing API entries did not automatically load the model list