Claude Code daft-udf-tuning skill for vectorized batch processing

Choose a Daft UDF based on state, vectorization, and resources, then verify the output and runtime behavior.

Install in Claude Code

A Claude Code project with Daft and PyArrow. The example compares rowwise and vectorized batch output on local numeric columns.

Review `.claude/skills/daft-udf-tuning/SKILL.md` and its requirements at the pinned GitHub commit.
Create `.claude/skills/daft-udf-tuning/` in your project and save the reviewed file as `.claude/skills/daft-udf-tuning/SKILL.md`.
Start Claude Code from that project's root, then select `/daft-udf-tuning` in the slash menu.

Example task

Invoke /daft-udf-tuning and design a batch comparison from these fictional numeric columns.
task `museum attendance totals`
column `visitors=[5, 7, 9]` | column `guides=[2, 4, 6]`
operation `add paired values` | expected `combined=[7, 11, 15]`
CPU-only local fixture

What to check

Suggest `@daft.func.batch` for paired addition. Combine `visitors=[5, 7, 9]` and `guides=[2, 4, 6]` to get `combined=[7, 11, 15]`. Compare against the rowwise result first.

Original SKILL.md on GitHub · Updated 2026-11-14

Frequently asked question

Does adding two numeric columns need a GPU?

Start by checking this vectorized numeric operation on a CPU. Choose resources using the operation and observed evidence.