infer-logical-architecture-from-code

You reverse-engineer a logical architecture by deriving bounded contexts, components, and integration points from codebase scan signals and prior enriched capability proposals. This produces a technology-agnostic topology using repository structure and co-change signals without inventing physical implementation details.

Bounded contexts are logical system boundaries derived from repository structure, idioms, co-change signals, and entry points. During brownfield codification, you identify them by analyzing scan-index data to keep architectural boundaries grounded in actual codebase evidence.

You generate a technology-agnostic architecture artifact by scanning codebase signals and applying an abstraction-layer scrub to block physical, protocol, SDK, and schema-token leakage. This yields a logical architecture containing data entities, capability APIs, and integration points.

Yes, you need scan-index.json and prior scope, enriched-capabilities, and features proposals to infer a logical architecture. These inputs validate component roles and capability serving relationships, while optionally leveraging an architecture-inference stepping stone.

Logical architecture inference leaks physical details when the abstraction-layer scrub fails to enforce strict validation against knowledge-patterns. Blocking physical, protocol, SDK, and schema-token leakage ensures the derived bounded contexts remain technology-agnostic.