Systems Research
Optic-Trigeminal / ACmK
Zero-dependency C++ clinical patient simulation and nursing decision engine backed by Groq AI, Argon2id cryptographic security, and a 5-plane cognitive microkernel architecture designed to make cognitive state, reasoning traces, uncertainty, and human intervention completely inspectable.
Engineering objective
Can AI inference and clinical simulation be structured into a zero-dependency C++ microkernel where cognitive state, reasoning traces, vital telemetry, and cryptographic boundaries remain completely inspectable and controllable?
Architecture
Optic-Trigeminal implements an inspectable clinical decision substrate in zero-dependency modern C++ (C++17/20). It simulates a 6-patient intensive care cohort with physiological vital telemetry (HR, BP, SpO2, temp), handles clinical interventions via Groq Llama 3.3 70B inference, protects medical records with Argon2id cryptographic password hashing, and exposes state through an ACmK 5-plane inspectable architecture.
- Zero External C++ Dependencies: Implements native HTTP/HTTPS client, JSON parsing, and clinical simulation without third-party C++ libraries.
- Five Distinct Planes: Strict vertical isolation between Control, State, Trace, Inference, and Environment/IO.
- Cryptographic Clinical Security: Argon2id password hashing and session tokens protecting patient data and nursing records.
- Cognitive Debugging: Every inference step emits an immutable snapshot enabling step-through debugging and rollback.
Implementation evidence
Inspectable Cognitive Control
SUPPORTEDDecoupling inference from execution control permits interactive human-in-the-loop intervention and rollback.
Trace Storage Growth
OBSERVEDContinuous deep state snapshotting requires ring-buffer pruning for extended operation.