Rivière

Appearance

Deterministic Extraction ​

Same code always produces the same architecture graph. No AI variability.

What It Is ​

Deterministic extraction uses a language-agnostic configuration DSL to define detection rules. A language-specific extractor (like TypeScript) parses your code and identifies components matching those rules.

Fast — Runs in seconds. No AI, no network calls.

Deterministic — Same code always produces same results.

CI-Ready — Integrate into build pipelines and gates.

When to Use ​

Use CaseAI-AssistedDeterministic
Initial extraction from unknown codebase✓
Codebase without architectural conventions✓
CI/build pipeline integration✓
Consistent, repeatable extraction✓
Speed (seconds vs minutes)✓
Codebases following conventions✓

Best practice: Use AI-assisted for initial extraction. Move to deterministic once you've established architectural conventions.

How It Works ​

  1. Define extraction config — Specify how to detect APIs, UseCases, DomainOps, Events, EventHandlers, UI components
  2. Run extractor — Language-specific tool (TypeScript, Java, Python) parses code using config rules
  3. Output draft components — Components with type, name, location, domain (no connections yet)
  4. Feed to CLI — Use CLI commands to build full Rivière graph

The Config DSL ​

The config format is language-agnostic — defined in JSON Schema, works across TypeScript, Java, Python, etc.

Current implementation: TypeScript (@living-architecture/riviere-extract-ts)

You can build extractors for other languages using the same config format.

yaml
modules:
  - name: "orders"
    path: "src/orders/**/*.ts"
    api:
      find: "methods"
      where:
        inClassWith:
          hasDecorator:
            name: "APIContainer"
    useCase:
      find: "classes"
      where:
        hasDecorator:
          name: "UseCase"
    # ... other component types

See Config Reference →

Modular Architecture ​

You can swap any piece:

  • Use our config + your extractor — Implement AST parsing for your language
  • Use our extractor + your config — Custom detection rules for TypeScript
  • Use our predicates + your parser — Reuse predicate logic in different tools
  • Build entirely custom — Use as reference implementation

TypeScript Implementation ​

We provide a complete reference implementation for TypeScript:

  • Decorators for annotating code (@APIContainer, @UseCase, etc.)
  • Default config ready to use
  • ESLint enforcement to ensure coverage
  • Extractor using ts-morph for AST parsing

Get Started with TypeScript →

Enforcement ​

Deterministic extraction works best with enforcement. Without enforcement, extraction becomes unreliable as code evolves.

Learn about Enforcement →

See Also ​