Architecture

Ecosystem Overview

The Joyint ecosystem consists of multiple products that work together. At the center is joy-core, a shared Rust library that provides the data model, YAML I/O, status logic, and Git integration used by the CLI tools, the desktop app, and the platform. Planned parts are drawn dashed.

Products at a Glance

Product	License	Purpose
Joy	MIT	Git-native product management for teams
Jyn	MIT	Personal task management with recurrence
Jon (planned)	MIT	Natural language interface for Joy and Jyn
Platform	Commercial	Forge sync, web app, AI jobs, chats, budgets
App	Commercial	Desktop and web apps, Tauri and SolidJS
Judge (planned)	MIT	Post-hoc audit, compliance, anomaly detection
Crypt	MIT	Encryption and key management library
JOYC (idea)	Commercial	IOTA-based settlement layer

Data Flow

Data flows from user input through the CLI into local files, then through Git to team members and the platform.

How It Works

The user runs CLI commands to create and manage items.
Changes are written as YAML files in the .joy/ or .jyn/ directory.
Standard Git operations (commit, push, pull) distribute changes.
The platform provides optional sync through your forge for the web app.
Every Joy action is recorded in the event log. Independent audit by Judge is planned.

Trust Model

Joy's AI Governance architecture is built on five pillars that together form the Trust Model. The model scales from implicit trust (solo developer, no configuration) through explicit trust (team with capabilities) to verified trust (enterprise with gates, cost limits, and audit trails).

The Five Pillars

Pillar	Question	What it covers
Trustship	Who do I trust?	Authentication, identity, delegation tokens, capabilities, generated agent definitions
Guardianship	What do I protect against?	Gates, `allow_ai` flags, Guard (runtime protection), Crypt (encryption), job containers on the platform
Orchestration	How do I steer work?	Jobs and interaction levels today; Dispatcher and bring-your-own-executor planned
Traceability	What happened?	Event log and audit trail today; Judge (post-hoc audit, compliance, anomaly detection) planned
Settlement	What did it cost?	Job costs, per-job and monthly budgets today; JOYC on-chain settlement as an idea

Trust Levels

The Trust Model supports three levels that projects grow into naturally:

Implicit trust - Solo developer, no governance configuration needed. Everything works out of the box.
Explicit trust - Team with defined members, capabilities, and delegation. AI tools have identity and limits.
Verified trust - Enterprise with gates, mandatory reviews, cost controls, and audit trails. Independent verification via Judge is planned.