[nathan-stender]

Summary

Vi-Cell BLU's per-sample detail export currently fails with Expected non-null value for Viability (%). The parser only handles the flat summary export (one row per sample), but the detail export has:

1. A sample settings header and its values
2. A reagent table (part number, lot number, expiration, ...)
3. A table with one row per image, ending in a Total row that holds the sample-level results

pandas read every row against the settings header, so the results columns came out empty. Sniff also missed the file, because Viability (%) isn't on its first line.

Changes

• Reader: detects the detail export (first cell is Sample ID and there is an Image# table) and turns each sample into one summary-style row: settings, then reagent info, then the Total row's results. Files with several samples are supported. Summary exports take the same path as before.
• Sniff: looks for the results columns on the Image# line when it exists, and on the first line otherwise.
• New fixture: Beckman_Vi-Cell-BLU_detail_export.csv, an anonymized customer file cut to 3 image rows.

Notes

• Per-image rows are not included in the output, matching what the summary export gives you.
• Reagent fields and ROW/COLUMN/Concentration Adjustment Factor (%) go into custom information. If a sample lists more than one reagent, the reagent info is left out.
• The lot number is converted to a float by the shared custom-info coercion (SeriesData._get_custom_key). This is existing behavior across all parsers and is not changed here.

Testing

• Existing expected JSON is unchanged.
• The new fixture fails without the fix, for the original reason.
• Beckman Vi-Cell BLU tests and tests/discover_vendor_test.py pass (342 tests), and lint passes.

🤖 Generated with Claude Code