(New) Use DeepSeek locally on your device (New) Updated llm engine code

(Fixed) Custom imported models cannot be used in chat view (Improved) Improved UI design by moving model availabilities indicators and custom model definitions into Models tab (New) Sharing the same model files across my other indie apps, including SoraSNS

(New) Upgraded to Phi 3.5 and Llama to 3.1 (New) Added feature to verify downloaded model file to check if it is complete (New) Added feature to check for local AI model file updates (Revised) Improved the UI for downloading local AI models

- Fixed issue with voice speak button not intractable for voice chat mode - Fixed intent not using custom server endpoints (when provided)

(New) Download local AI LLM models like Llama 3.1, Phi 3, LLM-jp to run them locally on your device and chat with them. You can download LLM models in the "Models" tab, chat with them either using the quick chat box, or by creating a chat room. (New) Import your own LLM model file and experiment with chatting with them, you can also configure different ML parameters, prompt formats, and other data.

- Fixed UI issue - Added quick chat feature - Choose to discard or save the history of quick chat

(New) Talk to AI in voice chat mode, speak your question, and hear your answer. Use a variety of supported languages.

- (new) new Apple Watch app allows you to chat with AI on your wrist - (Improved) continued chat responses are automatically combined with the previous message - (fixed) fixed an issue when the first chat request might not succeed if used app extensions - (fixed) loading is slow for chat rooms with a lot of messages

- Quickly retry a request after a failure - Updated screenshots to reflect the newest UI - Fixed an issue with continuing chat not including previous user prompts

Fixed an issue with UI layout when creating a new chat