Milab-6847: vj usage abundance dtype

.changeset/vj-usage-abundance-dtype.md

@@ -0,0 +1,17 @@
+---
+'@platforma-open/milaboratories.vj-usage.workflow': patch
+'@platforma-open/milaboratories.vj-usage': patch
+---
+
+Fix "division with 'String' datatypes is not allowed" when a dataset produces no
+numeric abundance rows.
+
+The input table was built with `tsvFileBuilder.build()`, which emits an untyped
+TSV, so ptabler let polars infer the column types. An `abundance` column with no
+parseable numeric value in the exported table — an empty join result, or all-null
+abundance — is inferred as `String`, and the per-sample normalisation then failed
+during query planning.
+
+The table is now built with `buildForPT()`, so column types come from the
+p-column and axis specs instead of inference. This also bumps
+`@platforma-sdk/block-tools` to 2.14.3 to satisfy the CI publication gate.

.github/workflows/build.yaml

@@ -22,9 +22,9 @@ jobs:
       - init
     uses: milaboratory/github-ci/.github/workflows/node-simple-pnpm.yaml@v4
     with:
-      app-name: 'Block: CDR3 SpectraType'
-      app-name-slug: 'block-cdr3-spectratype'
-      node-version: '20.x'
+      app-name: 'Block: V/J Gene Usage'
+      app-name-slug: 'block-vj-gene-usage'
+      node-version: '22.x'
       build-script-name: 'build'
       pnpm-recursive-build: false
       test: false

.github/workflows/mark-stable.yaml

@@ -14,8 +14,8 @@ jobs:
       - init
     uses: milaboratory/github-ci/.github/workflows/block-mark-stable.yaml@v4
     with:
-      app-name: 'Block: CDR3 SpectraType - Mark Stable'
-      node-version: '20.x'
+      app-name: 'Block: V/J Gene Usage - Mark Stable'
+      node-version: '22.x'
       npmrc-config: |
         {
           "registries": {