AIYRA is not a chatbot. It is not a prompt wrapper. It is not a coding assistant.
AIYRA is a Universal Autonomous Work Platform — a desktop-first operating system for AI agents, capable of turning high-level goals into verified professional outcomes through a sovereign, reusable execution kernel.
Think of it as an OS for AI — where agents are processes, the kernel manages their lifecycle, security enforces their boundaries, and the boot system guarantees cryptographic trust before any agent runs.
AIYRA Kernel (Rust + Tauri)
├── Layer 0: Boot System ← Cryptographic trust chain, DAG orchestration [FROZEN v1.0]
├── Layer 1: Runtime Foundation ← Identity, Contracts, Lifecycle FSM, API [FROZEN v1.0]
├── Layer 2: Discovery Pipeline ← Directory→Registry→Compatibility→Resolver [IN PROGRESS]
├── Layer 3: Intelligence System ← 14 AI cognitive sub-agents [30% BUILT]
├── Layer 4: Execution Engine ← Saga orchestrator, action layer, graph engine [50% BUILT]
├── Layer 5: Security Platform ← Vault, sandbox, zero-trust enforcement [40% BUILT]
└── Layer 6: Observability ← Health, metrics, distributed tracing [25% BUILT]
| Principle | Description |
|---|---|
| Downward Dependency Only | Upper layers can only call downward into lower layers. Never upward. |
| Immutable Identity | A runtime's UUID, owner, and creation timestamp never change. |
| Zero Trust | Every runtime must present a signed capability manifest before execution. |
| Certified Layers | No layer is built until the layer below is architecturally frozen. |
| No Primitives for Structure | Every concept is a typed struct/enum. No bare strings. |
| Layer | Technology |
|---|---|
| Desktop shell | Tauri v2 |
| Kernel & all agents | Rust |
| Async runtime | Tokio |
| Frontend | React + TypeScript + Vite |
| Storage | SQLite (via rusqlite) |
| Serialization | serde / serde_json |
| IPC | Tauri commands + event streams |
| Testing | Rust built-in + custom certification suites |
aiyra/
├── aiyra-core/
│ ├── src/ ← React/TypeScript frontend
│ └── src-tauri/
│ ├── src/
│ │ ├── kernel/ ← THE KERNEL (boot, runtime, registry, supervisor)
│ │ ├── intelligence/ ← 14 AI cognitive sub-agents
│ │ ├── execution/ ← Saga orchestrator + workflow engine
│ │ ├── runtime/ ← Action layer (browse, file, terminal, graph)
│ │ ├── security/ ← Vault, sandbox, tokens, trust
│ │ ├── agent_os/ ← OS-layer interface for agents
│ │ ├── workspace/ ← Workspace isolation
│ │ └── db/ ← SQLite abstraction
│ ├── tests/ ← Certification + integration test suites
│ └── tools/aiyra-audit/ ← Self-auditing compliance engine
Codebase stats:
- 343 Rust source files
- ~827 KB of kernel source code
- 5 test suites (boot, certification, intelligence, quality, kernel)
- 1 self-auditing compliance tool (
aiyra-audit)
| Document | Purpose |
|---|---|
AIYRA_PLATFORM_CONSTITUTION.md |
The 10-article supreme law governing all architectural decisions |
ADR-001-kernel-identity.md |
Architectural decision: Identity vs Descriptor separation |
ADR-002-runtime-discovery.md |
Architectural decision: 7-component discovery pipeline |
| Layer | Status | Completion |
|---|---|---|
| Layer 0: Boot System | 🟢 FROZEN v1.0 | 95% |
| Layer 1: Runtime Foundation | 🟢 FROZEN v1.0 | 95% |
| Layer 2: Discovery Pipeline | 🟡 Planned (ADR-002) | 5% |
| Layer 3: Intelligence System | 🟡 In Progress | 30% |
| Layer 4: Execution Engine | 🟡 In Progress | 50% |
| Layer 5: Security Platform | 🟡 In Progress | 40% |
| Layer 6: Observability | 🔴 Scaffolded | 25% |
| Frontend (React) | 🔴 Early | 10% |
# Install frontend dependencies
cd aiyra-core
npm install
# Run the desktop app in development mode
npm run tauri devcd aiyra-core/src-tauri
cargo test --lib kernel::runtime::foundation
cargo test --libcd aiyra-core/src-tauri/tools/aiyra-audit
cargo run -- --path ../../srcLETSCOOKTECH Studio
Building the infrastructure for autonomous AI that actually works.
MIT — See LICENSE
