Skip to content

RIGOR

AI Constraint Protocol. Formal boundaries for AI-generated systems.

Why RIGOR?

Modern software systems increasingly rely on Large Language Models to assist in code generation. However, natural language is:

  • Ambiguous: Interpretable in multiple ways.
  • Context-dependent: Meaning changes with the prompt.
  • Non-deterministic: Produces inconsistent structural outputs.
  • Difficult to validate: Statically verifying narrative intent is impractical.

RIGOR exists to replace narrative ambiguity with:

  • Typed declarative structures: Formal process definitions.
  • Deterministic state transitions: Explicitly allowed paths.
  • Explicit mutation rules: No hidden side effects.
  • Static validation before execution: Structural compliance is a precondition of existence.

RIGOR is not a framework. It is not a runtime engine. It is a formal specification language designed for precision-first generation.


Core Invariants

The following invariants are formally defined in RIGOR Core v0.1 and are non-negotiable:

1. Event-Driven Mutation Only

All state mutation MUST occur exclusively inside event-triggered transitions. No state changes are allowed outside transitions, and no implicit side effects are permitted. This ensures determinism, traceability, and replayability.

2. One Event = One Transaction

Each processed event represents an independent transactional boundary. The Engine guarantees atomic state transitions and atomic context mutations. If any step fails, the entire transaction is rolled back.

3. Typed Context is Mandatory

Every process MUST define a typed context_schema. No dynamic properties or implicit field creation are allowed. All mutations must conform to the declared schema, enabling static validation and generator reliability.

4. Deterministic Transitions

For any (state, event) pair, at most one transition is allowed. Guards must be pure and must not mutate context.

5. Core Semantic Freeze

RIGOR Core v0.1 defines the process model, event model, transition model, mutation rules, and transaction boundaries. Future versions may extend, but must not break these foundational invariants.


Standard First, Implementation Second

RIGOR is designed as an open specification. It is engine-agnostic and independently implementable. While the official RIGOR Engine serves as a normative reference implementation, the standard remains valid and useful independently of any specific runtime.


Explicit Design Constraints

RIGOR deliberately excludes:

  • UI description and Layout.
  • Infrastructure orchestration (Terraform/IaC territory).
  • Runtime scheduling and retries.
  • Distributed consensus algorithms.
  • Direct database modeling.

RIGOR focuses exclusively on process semantics, state transitions, and event-driven mutation.


Long-Term Vision

While RIGOR begins with backend process modeling, its semantic model enables derivative layers such as API contract derivation, frontend state synchronization, and cross-service orchestration. The core remains backend-first to provide a stable foundation for AI-aligned system generation.

AI Constraint Protocol