rmc-semantic-overlaps

Semantic similarity detects duplicate or near-duplicate Rust logic across a workspace by building a hypergraph, indexing item-level vector embeddings, and returning ranked pairs or clusters for refactor planning and deduplication.

Chunk-level semantic search supports free-text queries to locate similar Rust code fragments via vector similarity, returning ranked matches for single-item neighbor discovery and workspace-wide overlap audits.

Workspace-wide overlap auditing clusters similar Rust items using cached vector embeddings, identifying extraction candidates and deduplication targets while supporting verification via call-site checks.

Semantic similarity identifies duplicate Rust logic even when code is renamed, moved across crates, or exists in multiple variants, relying on item-level vector embeddings rather than exact text matching.

Vector-backed search returns ranked pairs or clusters of similar items, and you verify the top deduplication candidates by performing call-site checks using call-graph verification.

Semantic overlap detection requires hypergraph building and indexing for item resolution before performing vector-backed search, meaning it depends on thresholded matching and may require call-site verification to confirm true duplicates.