# CoreAI Architecture

## Purpose

This document describes the logical architecture of CoreAI v1.

CoreAI is a context foundation, not a full application platform. Its architecture is built from readable files, manifests, bundles, snapshots, and exports.

## Logical Layers

### Documentation Layer

Markdown documents describe the project, decisions, models, runtime assumptions, and domain knowledge.

### Manifest Layer

Manifests describe documents and assets in a structured way. They provide identity, metadata, visibility, trust level, tags, and relationships.

### Bundle Layer

Bundles select and order context for a specific purpose, such as onboarding, forensic analysis, runtime loading, or publication.

### Runtime Layer

The runtime loader reads manifests and bundles, resolves files, applies visibility rules, and produces exports or snapshots.

### Forensic Layer

Snapshots and timelines preserve state, events, evidence, and causal links. They make context changes traceable over time.

### Domain Layer

Domains hold operational knowledge. HAS84 is the first validated domain, but the CoreAI core must remain generic.

## Context Flow

```text
Documents -> Manifests -> Bundles -> Runtime Loader -> Exports
                                      |
                                      v
                                  Snapshots
                                      |
                                      v
                                  Timelines
```

## Design Boundary

CoreAI should define how context is represented, selected, published, and preserved. It should not decide which AI provider, hosting model, or application UI a project must use.