3 projects
modelmri
Chrome DevTools for AI models and agents — see attention, features, and agent steps inside any local model.
modelmri-record
Record agent runs — LLM calls, tool calls, subagents — and view them on a ModelMRI timeline. Stdlib only.
verdikt-eval
A CPU-only decision layer for robot-policy evaluation: reads the eval JSON you already have and refuses to let you draw a conclusion the data does not support.
