Skip to content

Release History

v0.12.2 (2026-10-13)

Misc

  • Validation errors now carry liken-meaning exception classes.
  • Validation messages reworded.
  • Values that are bool are now rejected wherever a number or integer is required (threshold, frac, thresholds items, max_distance).

v0.12.1 (2026-10-13)

Misc

  • Adds profilers as part of developer tooling

v0.12.0 (2026-10-12)

Breaking changes

  • Dataframe backend floors generally changed, and all upper caps are removed
    • pandas floor increased from 1 to 1.5.3.
    • polars floor lowered from 1.24.0 to 0.20.4.
    • pyarrow floor lowered from 23.0.1 to 13.0.0.
    • pyspark floor is unchanged at 4.
    • ray floor lowered from 2.55.0 to 2.50.0.
    • dask floor lowered from 2026.3.0 to 2024.8.0.
    • modin floor lowered from 0.37.1 to 0.22.0.

Features

  • Python 3.15 support.

v0.11.0 (2026-10-10)

Breaking changes

  • Deduper bounding convention is now uniformly inclusive:
    • str_len matches lengths in the closed interval [min_len, max_len]; a value whose length equals min_len now matches (it previously did not), and min_len == max_len is the exact-length idiom, matching the new num_range's one-value interval.
    • The threshold family (fuzzy, cosine, jaccard, tfidf) matches a pair whose similarity is exactly equal to the threshold (previously strictly above).
  • str_len validates bounds at construction and raises ValueError for malformed calls. Previously str_len accepted inverted bounds and silently matched nothing.

Features

  • New edit_distance deduper: matches two values when their Levenshtein distance is at most max_distance (default 2).
  • New num_range deduper: matches numeric values inside the inclusive interval [min, max]. Available in it's negated form (~num_range).
  • Two new preprocessors: collapse_whitespace replaces every run of whitespace with a single space; regex_replace applies a regular-expression replacement.

Fixes

  • True null handling: a missing value (None or NaN) groups only with another missing value. "na" placeholder machinery is removed (so nulls no longer collide with literal "na" values)
  • Core type contracts annotated: the backend protocol, DataFrame wrapper interfaces and collections.
  • fuzzy's 100 * threshold scaling no longer drifts at the boundary; scores within a 1e-9 tolerance below the boundary also match.

v0.10.3 (2029-10-08)

Features

  • Installation guide get's it's own page
  • AI Agents documentation relegated to "In Practice"
  • General documentation revision

v0.10.2 (2029-10-07)

Fixes

  • Typing related imports relegated to checking only
  • General Python conventions review

v0.10.1 (2029-09-30)

Fixes

  • Missing typing in backends
  • Module docstrings, typos fixed
  • Validation of keep arg fixed

v0.10.0 (2029-09-30)

Features

  • PyArrow Table input support: deduplicate, canonicalize and explore pyarrow Tables natively

v0.9.2 (2026-09-24)

Features

  • Documentation overhaul: new documentation pages and added detail on tutorials.
  • Additional tests

Fixes

  • Tooling dependencies pinned
  • Pre-commit included

v0.9.1 (2026-09-23)

Features

  • Documentation overhaul: positioning now covers entity resolution and canonicalization
  • New "Use Cases" section
  • Tip on installing Liken in a uv-managed Python project
  • Reworked "AI Agent Skills" section, with a fixed tessl registry link
  • Docs tooling: zensical 0.0.39 to 0.0.51, rich added to the docs extras, termynal extension config removed from zensical.toml

Fixes

  • Fixed API reference links in the deduper table in the First Steps tutorial
  • CI badge now points at the ci.yml workflow

v0.9.0 (2026-06-04)

Features

  • Agent skills bundle for agentic workflows, installed from the tessl registry with tessl install victoraut/liken-skills
  • New .explore function for exploratory duplicate-rate profiling
  • Skill workspace renamed from liken/liken-skills to victoraut/liken-skills

Fixes

  • CI badge in README now points at the ci.yml workflow
  • Removed the under-development announcement banner from the docs

v0.8.0 (2026-05-05)

Features

  • Added support for Modin dataframes
  • Added support for Dask dataframes
  • Added support for Ray datasets
  • Pandas and Polars as default installation, other backends as optional
  • Pandas "affordances" added as Pandas API extension
  • Documentation on features, installation and first steps updated

v0.7.4 (2026-04-14)

Fixes

  • Fix release history

v0.7.3 (2026-04-14)

Features

  • Added documentation on "Common Recipes"
  • Revised documentation on "Tutorials"

Fixes

  • Added unit tests for .synthesize and .canonicals

v0.7.2 (2026-04-12)

Features

  • Release history added to documentation About page

v0.7.1 (2026-04-12)

Fixes

  • Docs build and deploy workflow

v0.7.0 (2026-04-12)

Features

  • .canonicals method returns canonical ids with more than one record

  • .synthesize method returns a "golden" record

  • Python 3.14 support

  • Migrated from mkdocs to zensical

v0.6.1 (2026-04-08)

Fixes

  • Removed .cache dir

Features

  • Optimised predicate dedupers

v0.6.0 (2026-03-24)

Features

  • Pipelines for deduplicating with several steps and AND semantics

  • Fluent interfaces

  • Pipelines support built-in Preprocessors

v0.5.0 (2026-03-19)

Features

  • Fuzzy deduper to accept rapidfuzz scorer

v0.4.5 (2026-03-18)

Fixes

  • Broken image links

Features

  • Updated badges

v0.4.4 (2026-03-18)

Features

  • Optimised Fuzzy deduper

  • Vectorized str predicate dedupers

  • Optimised cosine

  • Added performance explanation

v0.4.3 (2026-03-06)

Fixes

  • Removed fuzzy caching

v0.4.2 (2026-03-03)

Features

  • Updated logo

v0.4.1 (2026-03-02)

Features

  • Updated logo

v0.4.0 (2026-03-01)

Features

  • Rules predication allows for improved performance when combining a predicate deduper with a similarity deduper

  • Added icon and favicon

  • Added images

v0.3.1 (2026-02-22)

Fixes

  • general uv fixes

Features

  • Migrated build backend to uv

v0.3.0 (2026-02-22)

Fixes

  • Merge conflicts

v0.2.2 (2026-02-17)

Fixes

  • Added validation of input dataframe and updated type aliases

Features

  • Backend migrated from Numpy to Pyarrow

v0.2.1 (2026-02-15)

Fixes

  • Removed not implemented docstring warnings

v0.2.0 (2026-02-15)

Features

  • Added synthetic datasets

v0.1.3 (2026-02-14)

Fixes

  • Removed action release creation

v0.1.2 (2026-02-14)

Fixes

  • Types in Dedupe class

Features

  • Fixed readme and docs typos

v0.1.1 (2026-01-31)

Fixes

  • Broken link in README

v0.1.0 (2026-01-30)

Fixes

  • drop_duplicates now also drops canonical_id column

Features

  • New Jaccard deduper

  • New Cosine deduper

  • New LSH deduper

  • Custom deduper implementation

  • Added fake datasets generator

  • Update dataframe add canonical id method for spark

  • Added warnings

  • Docs migrated to mkdocs

  • Implemented Rules class with and semantics strats

  • Added tuple strat validation

  • Added negation on On instances

  • Added isna binary deduper

  • Added str_len binary deduper

v0.0.0 (2025-12-31)

Fixes

  • Updated dependencies