mova-whale-english-tests
The groupant beside tools/whale/tests/english.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
| component | label | where |
|---|---|---|
| approximant | partial | Forward approximant |
| ast | partial | syntax.json |
| composition | partial | Composition |
| cost | partial | Cost |
| descriptant | present | Descriptant |
| invariant | partial | Invariant |
| tests | finite evidence | tests.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.
| id | type | label | where |
|---|---|---|---|
behavior_complement | behavior_complement | partial source derived model | behavior_complement |
reverse_reconstruction | reverse_reconstruction | partial source derived model | reverse_reconstruction |
Model text
Descriptant
End-to-end English tests of the Whales with Mova's 0.4 model: planner table (18 prompts, 10 refusals) and a served run through the real mova-whale child with a fake Codex executable and a 4,000..5,000-word fixture spec. Source: tools/whale/tests/english.rs.
Forward approximant
Planner: Deputy::review per prompt -> capability list and first action's arguments, or failure stage with a review candidate. Served: fixture spec/catalog/fake codex -> animal::serve thread -> ask prompts over the socket -> assert answers, journal order, calls.log order and bundle results.
Invariant
No model or network is called: the teacher is a fake executable bound by the config's codex path; skipped without MOVA_OCTOPUS_TEST_MODEL.
Cost
About 60 s with the debug build (model load twice).
Composition
Parent: whale-workers; exercises animal.rs, answer.rs, whale.rs, contract.rs and the octopus deputy.
behavior_complement
Not covered: live teachers, concurrent workers > 1 in the served run, restart of a served process, every phrasing. Evidence status: source-derived model with finite checks, not a proof.
reverse_reconstruction
Failures name the prompt, plan and failure stage; passing runs leave no artifacts (temporary directory removed). No unique inverse is claimed.
tests.md
Command: cargo test --manifest-path tools/whale/Cargo.toml --offline --locked with MOVA_OCTOPUS_TEST_MODEL=$MOVA_DATA/download/0/0.4/en.bin (unit tests incl. 5 service tests, 2 realizer tests, 3 length-gate tests; 2 English tests: 18 prompts, 10 refusals, a served end-to-end run with the real mova-whale child and a fake Codex executable; two identical runs on 2026-11-04)
Negative controls 2026-11-04: arrival-only ordering, recovery that ignores started lines, length bounds widened by one word, and dropping not/if from the deputy's scope check each turn the corresponding tests red; restored afterwards. These are finite fixtures; no live teacher or Whale run; not a proof.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/tools/whale/tests/english.rs.groupant/