Landscape Audit v0.1.0 is the first public release of an exact, state-centered auditor for discrete local search.
Highlights:
- deterministic strict-descent closure and neutral-plateau BFS;
- separate point-local and plateau-local optimality conclusions;
- exact versus state-limited certificate semantics with atomic checkpoint/resume;
- deterministic parallel move evaluation with incremental/reference consistency checks;
- replayable JSON certificates and GraphML export;
- Max-SAT, graph-coloring, scheduling, and minimal neutral-trap examples;
- dependency-free header-only C++20 core under the MIT License.
Validation:
- local CMake/CTest: 3/3 tests passed;
- GitHub Actions: Windows, Ubuntu, and macOS passed on commit c651aad.
Scope note: this release extracts the general audit mechanism learned from the Erdős #617 project. It does not claim a mathematical result on Erdős #617 and does not yet ship the specialized high-performance K26 adapter.