Hetairos now remembers you — introducing Episodic Memory.
- Hetairos can now recall past conversations, your preferences, and personal details across sessions — she builds a persistent picture of who you are over time
- Memory retrieval is powered by a state-of-the-art semantic vector search engine running entirely on your device, with no data ever leaving your machine
- Automatically extracts and organizes facts about you from your conversations, so you never have to repeat yourself
- Smart multi-query retrieval: when you ask about multiple topics at once, each question is matched independently to the most relevant memory
- Memory naturally fades for old, unused information and reinforces what matters — inspired by how human long-term memory works
- Full privacy: the entire memory system runs locally using on-device AI models, zero cloud involvement
- Memory fact embeddings are now cached persistently — semantic deduplication no longer recomputes embeddings for already-known facts, making each extraction significantly faster as your memory grows
- Fixed an issue where the app would freeze on close after a memory-enabled conversation
- Minor fixes to model download logic
