v2026.8.12

Lungfish 2026.8.12

Channel: Preview

Previous versioned release: v2026.8.11

Stable baseline: v2026.8.7

Dependency set: 2026.2

Preview builds are under rapid iterative development. Features may be
incomplete, change quickly, or require additional feedback.

Lungfish 2026.8.12 is an expedited preview fixing consensus generation
from classifier alignment evidence, which failed for every whole-contig
request. It also adds direct numeric entry to every slider and a
rescalable coverage track. The dependency set is unchanged at 2026.2.

Changes since 2026.8.11

Consensus generation works again for whole-contig scope

Generating a consensus from classifier BAM evidence failed with
"Consensus output does not project exactly onto the requested reference
interval", and no consensus was produced. samtools consensus writes a
FASTA header of >chrom:start-end for a sub-region but a bare >chrom
when the requested region spans the whole contig. The header parser only
recognized the coordinate form, so it reported no start position and the
projection check rejected the caller's output.

Because whole contig is the default consensus scope, this affected every
default extraction, which made a consistent failure look intermittent.
Adjusting the depth, MAPQ, or base-quality thresholds appeared to do
nothing for the same reason: the sliders were applied correctly, but the
run failed afterward at any value.

A bare header is now read as the contig origin. The region suffix is
parsed from the final colon-delimited field and must be a numeric range,
so a contig whose own name contains colons (for example HLA:A*01:01)
is not misread as coordinates. Regression tests cover the bare header,
the colon-in-name case, and a full whole-contig extraction.

Any slider value can be typed directly

Every slider in the app now pairs with an editable numeric field, so an
exact threshold can be entered instead of dragged. Typed values are
clamped into the slider's range and snapped to its step; text that names
no number leaves the current value unchanged. This covers the consensus
evidence thresholds, read display settings, annotation dimensions,
assembly and classification wizard options, and genotype dropout
thresholds.

Coverage tracks can be rescaled to log10 or square root

A viral genome routinely mixes a few very deep peaks with long shallow
stretches. On a linear axis the peak sets the scale and low-coverage
regions collapse onto the baseline, which is exactly where dropouts need
to be visible. A new "Coverage scale" control in the Inspector's View tab
offers linear (the default), log10, and square root, and applies to every
coverage track.

Uncovered positions stay on the baseline in every mode, and compressed
modes annotate the track legend (for example "Depth (log10)") so a
rescaled track is never misread as linear. The setting is display only:
it changes how depth is drawn, never the depth values, the consensus
evidence, or any exported output.

Dependency versions

Dependency set 2026.2 (2026-08-18), unchanged from 2026.8.11. No tool,
pipeline, database, or bootstrap pins changed in this release.