Last released Jul 4, 2026
13.7M trainable parameter neural pre-reasoning engine for grounding LLMs before they answer.
Supported by