Milab-6847: vj usage abundance dtype

[erohinaelena]

Greptile Summary

The PR fixes abundance normalization for empty or all-null input by preserving the analysis table’s declared schema, while also updating CI identity, Node, and publication tooling.

• Abundance — Numeric clonotype weight used for weighted gene-usage calculations. Its declared numeric type is now preserved even when no parseable values are present.
• PTabler schema — Structural metadata describing table column and axis types. The analysis table now uses buildForPT() to carry this schema into the calculation.
• Analysis table — Intermediate sample/clonotype/V-gene/J-gene/abundance table consumed by the usage calculation. Its construction changes from an inferred TSV to a typed PTabler artifact.
• Usage normalization — Division of grouped abundance or row counts by per-sample totals. It now receives a schema-defined abundance dtype rather than relying on CSV inference.
• V/J gene usage — Per-sample weighted and unweighted frequencies for V genes, J genes, and paired V/J combinations. The calculation itself is unchanged; only typed input loading changes.
• @platforma-sdk/block-tools — Build, packaging, publication, and stable-marking tooling. It is upgraded from 2.12.6 to 2.14.3.
• CI runtime — Node environment used by build and mark-stable workflows. It changes from Node 20.x to 22.x, and workflow labels are corrected for V/J Gene Usage.

Confidence Score: 5/5

The PR appears safe to merge, with no concrete blocking or independently actionable non-blocking defects identified.

The paired typed-table construction and frame-loading changes preserve existing headers and calculation contracts while preventing empty or all-null abundance data from being inferred as strings; the CI and tooling updates show no established incompatibility.

Important Files Changed

Filename	Overview
workflow/src/main.tpl.tengo	Replaces inferred TSV construction with a typed PTabler artifact while preserving existing analysis-table headers.
workflow/src/usage.lib.tengo	Loads the structural table directly so its embedded format and schema drive the usage frame.
.github/workflows/build.yaml	Corrects application identity and moves the build workflow from Node 20.x to 22.x.
.github/workflows/mark-stable.yaml	Corrects the stable-marking application label and moves its runtime to Node 22.x.
pnpm-workspace.yaml	Upgrades the shared block-tools catalog entry to 2.14.3.
pnpm-lock.yaml	Resolves block-tools 2.14.3 and its updated build-time transitive dependency graph.
.changeset/vj-usage-abundance-dtype.md	Documents the schema-inference failure, typed-table fix, and associated tooling release.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[Resolved sample, clonotype, V, J, and abundance columns] --> B[Build typed PTabler table]
  B --> C[Load frame with declared schema]
  C --> D[Compute per-sample totals]
  C --> E[Group V, J, and V/J usage]
  D --> F[Normalize weighted and unweighted measures]
  E --> F
  F --> G[Publish typed usage results]

Loading

_{Reviews (1): Last reviewed commit: "MILAB-6847: run CI on node 22, fix copy-..." | Re-trigger Greptile}

Context used (3)

• Knowledge Base — Gene usage workflow
• Knowledge Base — Workflow orchestration
• Knowledge Base — Gene usage calculations