kanon-sentry 0.2.0

A deterministic, local-first safety layer for agentic AI.

Kanon Sentry

Kanon Sentry is a deterministic, local-first safety layer for agentic AI. It is designed for developers who prioritize cryptographic accountability and on-device enforcement over cloud-based alignment.

This repository is the Phase 1: Sovereign release. The focus is a minimal, inspectable kernel with an auditable trail of every enforcement decision. Expect sharp edges; the goal is correctness and traceability first.

What It Does

• Validates actions against explicit intent constraints.
• Emits signed audit log entries for every decision.
• Keeps enforcement logic deterministic and transparent.

Quick Start

1. Set the signing seed:
   • KANON_SEED must be a 64-hex-character value (32 bytes).
2. Install locally:
   • pip install -e .
3. Run the example:
   • python test_enforcement.py

Philosophy

• Local-First Sovereignty: The kernel should run and be verifiable on your machine.
• Deterministic Safety: The rules are explicit; outcomes are reproducible.
• Cryptographic Accountability: Audit logs are signed and verifiable.

Status

Phase 1: Sovereign. Built for builders who want a practical, auditable safety kernel today, not a promise of future alignment.