UI/UX Improvements:
- Assistant Memory: Replaced status dots with clear "EMBEDDED" / "NO EMBEDDING" pills.
- Dynamic Regeneration: "Regenerate Embeddings" button now only appears when memories are actually missing embeddings.
- Consolidation: Removed "Avg. Significance" metric from statistics.
- Translator: Removed legacy translation UI from the main chat interface.
Reasoning Model Support:
- Updated background services (MemoryConsolidation, SpontaneousWorker, ScheduledMessage, ChatService) to strip reasoning/thinking content using toContentText(), ensuring only the final response is used for summaries and automation.
Embedding Persistence (Per-Model Cache):
- Implemented EmbeddingCacheEntity and EmbeddingCacheDAO to store embeddings keyed by model ID.
- Updated MemoryRepository to check the cache before generating new embeddings, enabling seamless switching between embedding models without losing previous work.
- Updated retrieval logic to prioritize cached embeddings for the current active model.