Elamur GroupantsāŒ•
☰

mova-whale-originality

The groupant beside tools/whale/src/originality.rs in the Mova repository: what the code computes, what it never does, how to go back from its outputs, its invariants, cost and checks.

Facts

Implementation material. A Mova code model describes this code at one source revision: partial models, finite checks and their stated scope. It is not a reviewed general exposition, and its labels are shown as written, never raised.

Components

componentlabelwhere
approximantpartialForward approximant
astpartialsyntax.json
compositionpartialComposition
costpartialCost
descriptantpresentDescriptant
invariantpartialInvariant
testsfinite evidencetests.md

Obernants

The typed relations around the forward model: the behaviour complement (what the code never does), the reverse reconstruction (every input compatible with an output) and specialised views.

idtypelabelwhere
behavior_complementbehavior_complementpartial source derived modelbehavior_complement
reverse_reconstructionreverse_reconstructionpartial source derived modelreverse_reconstruction

Model text

Descriptant

Configurable bounded local lexical originality gate against primary documents and a configured corpus. Source: tools/whale/src/originality.rs. Public, domain-agnostic: concrete prompts, schemas, gate parameters and deployments come from pinned configuration.

Forward approximant

Tokenize (Unicode alphanumeric/combining-mark runs, lowercase); select narrative strings except configured exempt keys; index exact-run, shingle and short-gram windows of each document (primary full text plus configured .md/.txt corpus files, optionally through a hash-verified SQLite text cache); mark exact runs, shingle coverage union and near-copy windows; decision blocked_overlap if any exact run, near window >= near_percent, or shingle union >= aggregate bounds; pending_incomplete if any coverage reason; else clear_in_checked_corpus.

Invariant

publication_allowed iff decision=clear_in_checked_corpus; any incompleteness (missing roots, bound exhaustion, symlink, unreadable file, incomplete primary) yields at least pending_incomplete; thresholds lie in documented bounds and the three gram sizes are distinct; the cache returns text only if its hash equals the source hash.

Cost

O(D * (T_doc + T_out * sizes)) BTreeMap operations for D documents with T tokens; windows add O(T_out * |windows| * width); bounded by max_entries/max_bytes/max_document_bytes/max_output_bytes.

Composition

Called by whale::publish on the raw reply before projection with the specification thresholds merged with deployment corpus keys (spec::originality_config).

behavior_complement

Domain: narrative fields and configured documents. Excluded from 'clear': any exact run, near-copy window or aggregate overlap; any incomplete corpus. Not detected: semantic paraphrase, translation, copying from outside the configured corpus. Counterexample: substitutions breaking every exact run still block via near windows (lexical_near_copy_without_exact_run_blocks). Evidence status: source-derived model with finite checks, not a proof.

reverse_reconstruction

A receipt records matched spans (capped at 100), document hashes and thresholds; compatible narratives are all texts with the same token coverage pattern; the original text is not reconstructible from counts. A clear decision admits every narrative with no configured overlap in the checked corpus only. No unique inverse is claimed.

tests.md

8 originality tests: exact/case, near copy, configured thresholds incl. invalid sizes, aggregate union with negative control, exempt fields, incomplete scope, four concurrent cold cache writers, cache tamper.

Command: cargo test --manifest-path tools/whale/Cargo.toml --offline --locked (74 tests, fake teachers/executables; two identical runs on 2026-11-03). These are finite fixtures; no live run, model call or deployment was exercised.

This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/tools/whale/src/originality.rs.groupant/