Skills
A skill is a package of structured files that teaches an AI coding agent how to work with a specific tool or framework. The skill below was generated by Great Docs from this project’s documentation. Install it in your agent and it will be able to run commands, edit configuration, write content, and troubleshoot problems without step-by-step guidance from you.
Any agent — install with npx:
npx skills add https://polars-umap.jeroenjanssens.com/Codex / OpenCode
Tell the agent:
Fetch the skill file at https://polars-umap.jeroenjanssens.com/skill.md and follow the instructions.Manual — download the skill file:
curl -O https://polars-umap.jeroenjanssens.com/skill.mdOr browse the SKILL.md file.
SKILL.md
--- name: polars-umap description: > UMAP dimensionality reduction as a Polars plugin, powered by Apple's embedding-atlas Rust implementation. Use when writing Python code that uses the polars_umap package. license: MIT compatibility: Requires Python >=3.9. --- # polars-umap UMAP dimensionality reduction as a Polars plugin, powered by Apple's embedding-atlas Rust implementation ## Installation ```bash pip install polars-umap ``` ## When to use what | Need | Use | |------|-----| | Embed an Array/List embedding column | `pl.col("emb").umap.fit_transform()` | | Embed several numeric columns | `polars_umap.umap("a", "b", "c")` | | Add umap_0, umap_1 columns to a frame | `df.umap.fit_transform("emb")` | | One embedding per group | `pl.col("emb").umap.fit_transform().over("group")` | | Approximate nearest neighbors | `pl.col("emb").umap.knn(k=15)` | | UMAP on a precomputed kNN graph | `polars_umap.umap_from_knn("indices", "distances")` | ## API overview ### Functions Build expressions that run UMAP or nearest neighbor search. They take column names, expressions or selectors. - `umap` - `knn` - `umap_from_knn` ### Namespaces Importing `polars_umap` registers a `umap` namespace on `pl.Expr`, `pl.DataFrame` and `pl.LazyFrame`. - `UmapExprNamespace` - `UmapDataFrameNamespace` - `UmapLazyFrameNamespace` ## Gotchas 1. Import `polars_umap` before using the `.umap` namespace; the import registers it. 2. UMAP fits on all rows the expression sees. Use `.over(...)` for one fit per group. 3. There is no `transform()` for new points; refit on the combined data instead. 4. Rows with null, NaN or infinite values raise unless `null_policy="ignore"`. ## Best practices - Pass `random_state` for reproducible embeddings; the default 0.6 layout is then bit-identical across runs and thread counts. - Use `metric="cosine"` for text embeddings. - Keep `gpu=False` unless the data is large (around a million rows) or you use `layout="0.5"`. ## Resources - [Full documentation](https://polars-umap.jeroenjanssens.com/) - [llms.txt](llms.txt) — Indexed API reference for LLMs - [llms-full.txt](llms-full.txt) — Comprehensive documentation for LLMs - [Source code](https://github.com/jeroenjanssens/polars-umap)