FINAL FIOLET ENGINE

FINAL FIOLET ENGINE is a safety-critical systems research project focused on building a deterministic, fail-closed safety interlock for generative AI systems.

At the center of the repository is fiolet-core — a minimal Rust safety kernel whose behavior is formally constrained, test-enforced, and treated as normative.

This is not an application repository.
It defines what safe behavior means, not how to build features around it.

---

Project Status

• Fiolet Core and AgeOfDarkness integration is fully functional.
• Normative tests pass:
  • fiolet_conformance_halt.rs ✅
• Core safety invariants are enforced.
• CI validates conformance automatically.

---

Core component: fiolet-core

fiolet-core is a deterministic, monotonic, fail-closed safety kernel written in Rust.

Properties:

• deterministic behavior
• monotonic halt (halt is irreversible)
• explicit safety decisions (Continue | AtomicHalt)
• no hidden state
• no_std compatible
• panic = abort semantics in kernel context

---

Safety decisions

• Continue
• AtomicHalt

No recovery or partial halt is possible.

---

Normative safety standard

Defined in:

FIOLET_SAFETY_STANDARD.md

This is a binding safety contract specifying:

• mandatory invariants
• kernel behavior
• compliance tests (executed in fiolet-core/tests/)
• reference to formal model (SafetyKernel.tla)

---

Repository Structure

FINAL_FIOLET_ENGINE/
├── Cargo.toml
├── fiolet-core/
│   ├── Cargo.toml
│   ├── src/
│   │   ├── lib.rs
│   │   └── fiolet_conformance/
│   │       ├── mod.rs
│   │       └── adapter.rs
│   └── tests/
│       └── fiolet_conformance_halt.rs
├── .github/workflows/rust-ci.yml
├── FIOLET_SAFETY_STANDARD.md
├── SafetyKernel.tla
├── SafetyKernel.cfg
├── HOW_TO_READ_THIS_REPO.md
├── CONTRIBUTING.md
└── README.md

---

An implementation is FIOLET-compliant if:

It conforms to FIOLET_SAFETY_STANDARD.md.

Passes all normative tests.

Preserves all defined invariants.

License

MIT License

Author

Adrian Maliszewski Independent research on safety-critical systems and deterministic control of generative AI.