Skip to main content

AI-Centred Engineering · Public Framework · 1.1.0

AIRepo — The AI-Ready Repository Framework

An open, provider-neutral framework for making repository authority, ownership, lifecycle, provenance, compatibility, and validation explicit so humans and AI agents can work with clearer boundaries and stronger evidence.

This page is a navigation and explanation layer. Canonical framework semantics and release authority remain in the public GitHub repository and tagged release.

CURRENT RELEASE

AIRepo 1.1.0

Backward-compatible minor release with 29 active Rule IDs.

LICENSE

Apache-2.0

Open framework, technology agnostic and provider neutral.

PUBLISHER

MILIUTIN

Creator and publisher attribution; not a framework runtime or semantic authority.

What it solves

Authority is ambiguous

AIRepo helps make it explicit which artifacts may define meaning, which are derived evidence, and which systems remain authoritative for concrete project state.

Ownership is hard to resolve

It provides a framework for clarifying semantic ownership, lifecycle responsibility, compatibility boundaries, and change authority.

Evidence is mistaken for truth

AIRepo 1.1.0 adds claim-relative Evidence Fitness so tests, validators, evaluations, and generated outputs prove only what they can actually observe.

AI work needs bounded execution

The framework supports proportionate assessment and remediation without turning the repository into an agent runtime, orchestration platform, or universal technology stack.

Start here

01 · Quick Start

Understand how to approach an AIRepo assessment and identify the smallest relevant scope.

Read Quick Start →

02 · Canonical framework

The semantic owners live in framework/SO-01…SO-10 in the public repository.

Open framework sources →

03 · Release record

Use the tagged release and release manifest for exact public identity, compatibility, and provenance.

Inspect release manifest →

AIRepo 1.1.0

Evidence Fitness and simpler routine use

Release 1.1.0 preserves all 28 active Rule IDs from 1.0.0 and adds VAL-003 — Evidence fitness for claim. It also adds explicit stale, incomparable, insufficient, and unobservable evidence handling and a Class A/B Fast Path for routine low-risk work.

Compatibility

  • 29 active Rule IDs total.
  • EVAL-001 remains reserved.
  • Machine Projection schema remains 1.0.0.
  • No consumer migration is forced by publication.

Practical entry points

Creator & advisory

AIRepo is open. Applying it well can still require judgment.

AIRepo is created by Andrei Miliutin and published under the MILIUTIN brand. For repository assessment, AI-centred engineering operating-model review, or bounded adoption work, use the existing advisory/contact path.

Start a Conversation