Crate agdr_aki 

AgDR-AKI: Atomic Kernel Inference SDK

Cryptographically-sealed, court-admissible AI decision records. Developed by the Genesis Glass Foundation (Fondation Genèse Cristal), a federally incorporated Canadian not-for-profit.

Feature Flags

• python (enabled by default): Enables PyO3 Python bindings via the pyo3 crate. Required for all Python-exposed types (AKIEngine, PPPTriplet, HumanDeltaChain, SealedRecord).

Core Concepts

Concept	Description
PPP Triplet	Provenance, Place, Purpose — immutable legal anchor recorded at inference time
AKI	Atomic Kernel Inference — sub-microsecond sealed decision capture
HumanDeltaChain	Fiduciary link tracking human oversight or escalation
Coherence Score	Normalized cosine similarity against the agent’s historical spine (0.0 – 1.0)
Reputation Scalar	Exponentially weighted moving average of past coherence
Sealed Record	Tamper-evident object with BLAKE3 hash and Ed25519 signature

Performance

Operation	Latency
Atomic Kernel Inference (hot path)	0.6 – 2 µs
Full sealed record (WAL + Merkle)	10 – 50 µs
ZK proof generation	< 10 ms
ZK verification (Merkle root only)	< 10 ms

See: https://accountability.ai

Structs

AKIEngine

Core execution container managing real-time calculations and integrity verification sequences.

CoreInsightToken

A real-time evaluation token logging historical data convergence patterns.

DeltaEmbedding

A localized, low-overhead embedding vector tracking contextual micro-shifts.

Functions

load_or_generate_signing_key

Retrieves an existing Ed25519 signing token or initializes a new key pair if missing.

monotonic_raw_nanos

Retrieves a raw, high-precision nanosecond timestamp from the system clock.