skrub-api

name	skrub-api
description	Discover the public API of the skrub package — top-level estimators, joiners, the DataOps lazy-pipeline framework, column selectors, datasets, and configuration. Use when you need to know what skrub exposes, what a class/function does, its signature, and where it lives, without grepping the source.

skrub-api

Reference index for skrub's public API. Each topic below has a dedicated file under references/ listing every public symbol with its import path, signature, and one-line description.

Only public API is documented (anything reachable from skrub, skrub.selectors, skrub.datasets, or skrub.core via __all__). Do not infer the existence of private helpers from this skill — read the source for those.

When to load a reference file

Read the file that matches the user's question. Don't preemptively load all of them.

Topic	File	Load when the user mentions…
Top-level overview	references/overview.md	"what does skrub export", "which transformer should I use", first contact with the API
Single-column encoders	references/encoders.md	DatetimeEncoder, GapEncoder, MinHashEncoder, SimilarityEncoder, StringEncoder, TextEncoder, ToCategorical, ToDatetime, ToFloat, SquashingScaler, DropUninformative
Joiners	references/joiners.md	Joiner, fuzzy_join, InterpolationJoiner, AggJoiner, MultiAggJoiner, AggTarget
Vectorizers & pipelines	references/vectorizers.md	TableVectorizer, Cleaner, tabular_pipeline
Column selection / wrapping	references/column_ops.md	SelectCols, DropCols, Drop, ApplyToCols
skrub.selectors module	references/selectors.md	s.numeric(), s.glob(...), building selector expressions
DataOps (lazy plans)	references/data_ops.md	DataOp, var, X, y, .skb namespace, deferred, as_data_op, eval_mode
Learners & tuning	references/learners.md	SkrubLearner, ParamSearch, OptunaParamSearch, cross_validate, choose_*, optional
Reporting	references/reporting.md	TableReport, patch_display, unpatch_display, column_associations
Configuration	references/config.md	get_config, set_config, config_context
Datasets	references/datasets.md	fetch_*, toy_orders, toy_products, make_deduplication_data, get_data_dir
Misc utilities	references/misc.md	to_datetime (function), deduplicate (function)
skrub.core	references/core.md	Building a custom SingleColumnTransformer, RejectColumn

How to use this skill

1. Identify the topic from the user's question and read the matching file (or overview.md if you're unsure).
2. Quote the signature and one-line description from the reference. If the user needs more (parameter semantics, return type, examples), fall back to the source under skrub/ or call help(...) in the dev environment (pixi run -e dev python -c "...").
3. The reference files are auto-generatable from __all__ and inspect.signature — if a symbol is missing here but present in skrub.__all__, the reference is stale; update it rather than improvising.

Conventions used in reference files

• Each entry: ### name → import line → Signature: → one-line description.
• class vs function is noted in the heading.
• Keyword-only markers (*) and defaults are preserved verbatim from inspect.signature.
• Long sklearn-object defaults (e.g. Pipeline(...), OneHotEncoder(...)) are abbreviated as <...> — check the source for the exact default if it matters.