Development of Sophia v2 is well underway and will be a general-purpose NLU engine, far superior in all aspects:

• Flawless intent recognition, plus multi-turn dialog handling across any domain.
• Knowledge compression into relational and time-series searchable data stores (E.G. RAG but deterministic and without preprocessing / chunking)
• Ability to surface buried relations and links within large volumes of text data.
• Deterministic, so 100% reliable with zero hallucinations or probabilistic mismatches.
• Adhere to core tenets of on-device, low compute, deterministic with high degree of precision and accuracy, all from the compute of a smartwatch with no internet required.

End goal is to become the Proton of NLP / NLU. Private, self-hosted, deterministic and accurate natural language processing free from big tech. Read product Cicero's manifesto.

Home Assistant edition

On top of the core upgrade, you can expect the following specifically for the Home Assistant edition:

• Persistent memory, so when it requests clarification it will remember your answer for future queries.
• Ability to update settings and train Sophia via voice. When Sophia gets something wrong, correct it via voice and Sophia will remember for future interactions.
• And much more....