Offer the complete reference in Simple mode, when it would change something

[JamesKane]

Realignment shipped in alpha.17 reachable only from Advanced mode, on a card among alignment internals. The reader most likely to be holding vendor data on an unfinished reference — and least likely to ever find that card — is the one in Simple mode.

What this adds

An offer under Your test, pitched at someone who should never need to know what a reference build is, with a confirmation step before a four-hour job starts.

Eligibility is a field on SubjectBrief, computed in the app layer beside needs_analysis, rather than a UI decision. It appears only when it would change something:

• A paternal line to improve. Realignment's payoff today is Y discovery; ancestry, IBD and the autosomes already handle GRCh37/38 and return the same answer either way.
• No CHM13 alignment already, by any route.
• An alignment the job would actually act on — realign::realignable_for_subject, extracted from the project-batch path rather than reimplemented. If the offer and the job disagreed about what is realignable, a user would be promised hours of work the job then refuses.

The bug a screenshot caught

The first version rendered correctly and was still wrong. On a donor holding six CHM13 alignments it offered to realign an old GRCh37 file nobody had touched, telling them their father's line had nowhere to be read from — while four finished alignments of theirs said otherwise.

The per-alignment rule was answering "would the job act on this file?". The card was making a claim about "is part of this person's line unreadable?". Those are not the same question, and only the second one is the offer's promise. Hence the subject-level guard.

Copy, corrected twice

Both corrections came from review, and both were substantive rather than stylistic:

1. It described the gaps as a shortcoming of the reader's test. They are a property of every human reference before T2T — saying so protects a reader from concluding their sequencing was bad.
2. It framed the finished genome as a paternal-line feature. That misdescribes what T2T delivered: a full autosomal sequence end to end and the first complete Y. Y discovery is what Navigator puts it to work on today, not the scope of what was built.

The card also now states whose genome the reference is — a donor of European ancestry, carrying a J1a Y — because a "complete" reference is still one person's, and complete-for-whom is exactly the sort of thing that goes unsaid.

README, User Guide and the published alpha.17 release notes carry the same framing.

Validation

• 983 tests, 0 failed; clippy clean under -D warnings; cargo fmt clean; locale parity (every_es_key_exists_in_en) passes.
• Six new tests on the eligibility rule, including the set-level one — the offer disappears once a realigned alignment exists — and that the masked CHM13 variant is still worth realigning to plain CHM13.
• Rendered in the running app and screenshotted; that is how the gating bug surfaced.

Not covered

The corrected gating has not been re-checked in the GUI. It follows deterministically from the subject's alignment rows, but it has not been seen.

🤖 Generated with Claude Code