Closer integration with the OS
This update brings Layla closer with the operating system! After you install Layla, you can find in context menus when you select any text in Android. You can choose to ask Layla to remember that text, putting it in her long-term memory, or send the text to Layla to chat with her. Layla can then explain, summarise, or remind you of it.
This feature will be expanded in the coming updates, giving you more control on what Layla does with the text. Some great ideas from our community:
Check grammar of selected text
Re-write selected text
Translate selected text (while loading a suitable multi-lingual LLM)
Layla will give you a new app where you can write your own instructions to utilise this feature as best it suits you!
Siri Integration (iOS Only)
Layla has been integrated with Siri!
Create your own shortcuts and send text to Siri! This feature will be expanded in the coming updates to give more features such as Layla replying, customising how the text is processed in Layla, and much more!
Background Inference (Android Only) redesign
The Background Inference widget has been removed. Now it integrates directly with the OS, showing up as a persistent notification. This should help with issues where the OS is killing the floating widget due to it using too much memory.
Redesigned RMI app
As more users are experimenting with trying out different models, the RMI (Raw Model Instructions) app has been redesigned to make it easier to switch between different models and prompts.
Your models are moved up to the top
Create custom prompts, name them to remember them easier
Popular models have their own logos!
Attach your custom prompts to your models so they are automatically applied when you switch
Full Changelog
New features:
Deeper integration into Android OS! You can select any text in your phone and ask Layla to remember it, explain it, summarise it, or remind you of it!
Siri shortcut support! Send any text to Layla and chat with her about it, more are coming soon!
Improvements:
removed need of floating widget in Background Inference
Background Inference now uses Android Foreground tasks, providing consistent notifications
Welcome screen now uses your default character images and name
added more configurable options for repetition control
added ability to share conversation history
added ability to edit & regen last message from loaded conversations
added ability to delete chat messages
added pagination and sorting in Personalities Hub
redesigned RMI app to streamline model management
character import now supports TavernPNG v1 cards
app screen now blocks you from installing incompatible apps (such as OpenAI + Claude)
Executorch now supports emojis
Bug fixes:
fixed bug where Preset boxes were not displaying correctly in Context Extend app
fixed bug where LTM is creating duplicate memories of the same conversation
fixed bug where voice configs were not showing if you only have GB voices installed
fixed bug where todos from imported data are not marked as completed
fixed bug where LTM was not rescheduling deleted tasks
Notes
The LTM bug did not affect your daily conversations. Since LTM only takes the first most relevant conversation, it did not matter if it was duplicated. However, it could be taking up unnecessary space on your phone. You can run Diagnostics in the settings screen to automatically clean up the LTM. You only have to do this once.