Serious bug fix, highly recommend to upgrade.

1. Added portraits for hexagram interpretations. 2. Added hexagram ratings and mapped them into the Life K-line chart (tap to toggle between Today’s Question and the Life K-line chart).

Added a “Today’s Question” card that auto-generates a sharp daily prompt from history, current time, and personal notes, with one-tap insert/undo.

- Improved Alter experience for a smoother flow - Polished hexagram interpretation layout

1. Something has changed? Tap Alter at the bottom of the card to view the relating hexagram. 2. Added sound effects and optimized animations.

Now is satisfying to just view the card.

1. Polished the card display and it finally looks great. 2. Fixed the issue where the last card drawn could differ from the target card.

1. Simpler Q&A with sharp, counter-intuitive titles. 2. Much higher Q&A cap for more asks. 3. Answer screen polish with improved display

1. Added background processing and notification features - you can now wait for results in the background without staying on the APP interface (Note: please keep the APP running in background) 2. Comprehensively optimized interface animations for enhanced visual experience 3. Improved AI analysis algorithms for more accurate interpretation results

Speed imporve for normal model

1. Updated to DeepSeek R1 0528 model as the enhanced model, with normal model also using deep thinking mode. 2. Added current time provision, enabling AI to have better time awareness 3. Added personal preset memory, now you can select common background knowledge when asking questions 4. Updated card drawing animation effects, now more engaging 5. Quick re-ask base on history

Fix the issue of default model not responding

Reduce abnormal English output

Now you can upload images to assist with your inquiries

Fix the issue of default model not responding

1. Increate the limit of daily usage for deep models 2. Optimize the default model, now faster

- Optimized streaming transmission - Updated the interface for the history record - QwQ-32B is also a native inference model, now updated to a standard model. It has reasoning capabilities comparable to DeepSeek R1, while being lighter and faster. - The API for DeepSeek R1 has also been updated to a higher-quality pipeline

Mic bug fixed

It's now available in China

1. Add Refresh button 2. Optimize model switching

1. Added model switching functionality 2. Optimized input field blur handling 3. Lowered minimum iOS version requirement to iOS 16