v0.3.3

What's Changed

• feat: add entity coverage as the new Nemo Anonymizer feature by @memadi-nv in #195
• chore: update telemetry, add inputTokens and internal deployType by @mckornfield in #200
• ci: make ty type checking blocking by @binaryaaron in #223
• docs: fix gen_ref_pages.py test-module substring filter by @andrewwhitecdw in #234
• feat(measurement): add W&B benchmark identity config by @lipikaramaswamy in #236
• docs: add devnote for evaluation replace by @memadi-nv in #239
• fix: restore NVSkills workflow startup by @andreatnvidia in #242
• refactor: rename LLM alignment score to Judge Agreement by @memadi-nv in #244
• chore: bump ty from 0.0.58 to 0.0.69 by @dependabot[bot] in #245
• chore(deps): require data designer 0.9 by @lipikaramaswamy in #248

Full Changelog: v0.3.1...v0.3.3

v0.3.1

What's Changed

• docs: polish self-hosted B300 devnote by @lipikaramaswamy in #222
• ci: add NVSkills CI request workflow by @binaryaaron in #217
• fix(measurement): allow rewrite W&B records and project dirs by @lipikaramaswamy in #221
• ci: use the DCO GitHub App by @binaryaaron in #224
• chore: add anonymizer skill publication artifacts by @binaryaaron in #214
• chore: Bump data-designer to 0.8.x series by @mikeknep in #228
• feat(measurement): add RAT-Bench reidentification W&B metrics by @lipikaramaswamy in #225
• chore: bump deps by @dependabot[bot] in #230

New Contributors

• @dependabot[bot] made their first contribution in #230

Full Changelog: v0.3.0...v0.3.1

v0.3.0

What's Changed

• docs: add docs for anonymizer replace evaluation by @memadi-nv in #181
• chore: Update indirect deps for vulns by @kendrickb-nvidia in #188
• feat(detection): export detection workflow for distributed/SLURM exec… by @mvansegbroeck in #182
• feat: anonymizer measurement instrumentation and benchmark tooling by @binaryaaron in #177
• fix(detection): pass single chunk validation flag to exports by @lipikaramaswamy in #190
• refactor(evaluation): consolidate judge workflows behind a base class by @memadi-nv in #173
• feat: load bundled model providers by default by @lipikaramaswamy in #189
• feat: add detection workflow column plugins by @andreatgretel in #192
• feat: rewrite evaluation improvements by @memadi-nv in #186
• chore: Bump data-designer to 0.7.x series by @mikeknep in #202
• fix: evaluate rewrite_judge not checked by @memadi-nv in #199
• feat(measurement): add W&B benchmark support by @binaryaaron in #191
• fix: increase latent entity rationale maxLength from 150 to 250 by @asteier2026 in #205
• docs: update contribution guidelines by @lipikaramaswamy in #204
• fix: keep rewrite final judge in evaluate by @fallintoplace in #196
• docs: add query flow diagram to home page by @alexahaushalter in #216
• docs: add self-hosted B300 devnote by @lipikaramaswamy in #213
• feat: expose Data Designer run config by @lipikaramaswamy in #203
• docs: update notebooks for release by @lipikaramaswamy in #218

New Contributors

• @mikeknep made their first contribution in #202
• @fallintoplace made their first contribution in #196

Full Changelog: v0.2.1...v0.3.0

v0.2.1

What's Changed

• docs: update README by @lipikaramaswamy in #167
• chore: Bump urllib by @kendrickb-nvidia in #169
• chore: Use uv and pip ecosystem for dependabot by @kendrickb-nvidia in #170
• feat: Add LLM-as-a-judge for Anonymizer replace evaluation by @memadi-nv in #158
• docs: self-hosting GLiNER locally by @mvansegbroeck in #132
• chore: collapse greptile diagram by default by @memadi-nv in #172
• chore: bump data-designer by @lipikaramaswamy in #178

New Contributors

• @mvansegbroeck made their first contribution in #132

Full Changelog: v0.2.0...v0.2.1

v0.2.0

What's Changed

• chore: fix data designer version by @lipikaramaswamy in #135
• docs: updating build notice by @alexahaushalter in #138
• fix: add visibility to silent failures in NDD adapter by @memadi-nv in #136
• fix(docs): publish release docs from release workflow by @lipikaramaswamy in #140
• feat: enhance repair by @asteier2026 in #137
• feat: re-export ModelProvider from anonymizer public API by @memadi-nv in #139
• chore: bump pytest to 9.0.3 by @kendrickb-nvidia in #147
• fix(replace): drop workflow-internal columns and use COL_* constants by @lipikaramaswamy in #154
• docs: add AGENTS.md, STYLEGUIDE.md, agent-assisted contribution (#114) by @lipikaramaswamy in #149
• feat: unify rewrite domain metadata into a single source by @memadi-nv in #143
• docs: add anonymizer Claude Code skill and supporting concept docs by @lipikaramaswamy in #153
• docs: add dataset license files and readme by @asteier2026 in #159
• feat: improve sensitivity disposition calibration by @asteier2026 in #150
• docs: define sensitivity and protection method by @asteier2026 in #151
• refactor: replace df.attrs with typed ResolvedInput context by @memadi-nv in #141
• feat: add anonymous run-level telemetry by @lipikaramaswamy in #155
• feat: enhance meaning units and utility by @asteier2026 in #156
• chore: upgrade data-designer to 0.6.0 by @lipikaramaswamy in #160
• fix(telemetry): flush events when called from a running event loop by @lipikaramaswamy in #161
• refactor: derive needs_protection from prot method and fix latent entity by @asteier2026 in #163
• docs(devnotes): add Introducing NeMo Anonymizer post by @lipikaramaswamy in #157
• refactor(display): consume structured rewrite columns in native shape by @memadi-nv in #144
• docs: update tutorials by @lipikaramaswamy in #166

New Contributors

• @kendrickb-nvidia made their first contribution in #147

Full Changelog: v0.1.1...v0.2.0

v0.1.1

What's Changed

• docs: update readme file by @memadi-nv in #113
• fix: Fix max_repair_iterations to be 3 by @asteier2026 in #108
• fix: register ipykernel before jupytext notebook execution by @lipikaramaswamy in #104
• feat: make preview input loading lightweight by @memadi-nv in #116
• feat: add strict_entity_protection by @asteier2026 in #123
• fix(cli): model config file paths and flag fixes by @lipikaramaswamy in #124
• fix: Input Error when duplicate columns in output data frame by @memadi-nv in #122
• feat: enable detection pipeline for async engine by @andreatgretel in #119
• docs: add strict_entity_protection by @asteier2026 in #131
• docs: replace placeholder patient_data.csv with real sample dataset by @lipikaramaswamy in #133
• feat(detect): chunked validation with validator pools by @lipikaramaswamy in #126

New Contributors

• @memadi-nv made their first contribution in #113

Full Changelog: v0.1.0...v0.1.1

v0.1.0

NeMo Anonymizer v0.1.0

First public release of NeMo Anonymizer — detect and anonymize sensitive entities in text using LLM-powered workflows.

Features

• Entity detection using GLiNER-PII with LLM-based augmentation and validation
• 4 replace strategies — Substitute (LLM-generated), Redact, Annotate, Hash
• Rewrite mode — transforms entire documents to reduce explicit and inferable identifiers, with customizable privacy goals, utility preservation, and automated repair
• Preview mode — inspect results on a small sample before full runs with display_record() visualization

Install

pip install nemo-anonymizer

Quick start

from anonymizer import Anonymizer, AnonymizerConfig, AnonymizerInput, Redact

anonymizer = Anonymizer()
config = AnonymizerConfig(replace=Redact())
data = AnonymizerInput(source="data.csv", text_column="text")

result = anonymizer.preview(config=config, data=data, num_records=3)
result.display_record()

Documentation

https://nvidia-nemo.github.io/Anonymizer/

Requirements

Python 3.11+

v0.1.0rc1

What's Changed

• chore: update github files by @lipikaramaswamy in #1
• chore: add repository boilerplate and contributor guidelines by @lipikaramaswamy in #2
• feat: implement replace mode (detection, replacement strategies, visualization) by @lipikaramaswamy in #3
• chore: add entity class examples to detection and augment prompts by @lipikaramaswamy in #11
• refactor: restructure AnonymizerConfig, rename strategies, add filter_labels by @lipikaramaswamy in #16
• chore: assign anonymizer-reviewers to docs/ in CODEOWNERS by @lipikaramaswamy in #19
• fix: resolve replaced highlight drift by @lipikaramaswamy in #17
• fix: harden io path validation and error handling by @lipikaramaswamy in #18
• fix: surface JSON parse failures for observability by @lipikaramaswamy in #20
• feat: improve validation id alignment with template-guided decisions by @lipikaramaswamy in #21
• fix: correct GLiNER model name and skip health check by @andreatgretel in #24
• chore: remove unnecessary skip_health_check for GLiNER by @andreatgretel in #25
• refactor: improve replacement mapping prompt by @lipikaramaswamy in #23
• refactor: update notebook source to use defaults by @lipikaramaswamy in #26
• feat: update latent entity detection prompt with demographic specifics by @lipikaramaswamy in #38
• feat: add preflight config validation for active workflows by @lipikaramaswamy in #27
• feat: add user-facing pipeline logging with progress signals by @andreatgretel in #22
• fix: filtered-label preview mismatch in replace mode by @lipikaramaswamy in #41
• fix: update detection prompts and entity examples by @lipikaramaswamy in #42
• feat: add rewrite foundation - schemas, constants, and model selection by @lipikaramaswamy in #40
• fix: align filtered display and harden detection prompt notation by @lipikaramaswamy in #43
• refactor: unify detection and replacement label scope by @lipikaramaswamy in #44
• fix: tighten substitute replacement-map generation by @lipikaramaswamy in #47
• feat: domain classification and sensitivity disposition workflows by @lipikaramaswamy in #45
• fix: drop age duration tags and first_name initials in validation prompt by @lipikaramaswamy in #50
• fix: cap preview_num_records to entity row count in replace workflow by @lipikaramaswamy in #59
• feat: add QA generation workflow for rewrite pipeline by @lipikaramaswamy in #48
• feat: add rewrite generation workflow by @lipikaramaswamy in #49
• fix: enforce entity_labels scope on augmented entities in detection output by @lipikaramaswamy in #57
• ci: upgrade GitHub Actions for Node.js 24 compatibility by @ko3n1g in #62
• feat: add evaluate and repair workflows by @lipikaramaswamy in #61
• chore: address a handful of cve versions with overrides/updates by @mckornfield in #69
• feat: add final judge and RewriteWorkflow orchestrator by @lipikaramaswamy in #64
• feat: port developer tooling and build from Safe-Synthesizer repo by @binaryaaron in #51
• feat: add Cyclopts-based CLI with run/preview/validate subcommands by @binaryaaron in #63
• chore: bump pygments and cryptography for security fixes by @lipikaramaswamy in #75
• feat: interface wiring + display by @lipikaramaswamy in #68
• chore: add lower-bound pins for pygments and cryptography by @lipikaramaswamy in #76
• docs: add tutorial datasets by @lipikaramaswamy in #81
• docs: Update README.md by @nina-xu in #85
• feat: composite actions, matrix testing, and release workflows by @binaryaaron in #54
• refactor: extract shared row-partitioning helpers by @andreatgretel in #79
• feat: improve privacy judge prompt by @asteier2026 in #86
• feat: add confidence-aware privacy re-answers and weighted leakage metric by @lipikaramaswamy in #88
• refactor: rename sensitivity disposition paraphrase+left_as_is actions by @lipikaramaswamy in #91
• refactor: simplify rewrite config and rename risk tolerance levels by @lipikaramaswamy in #92
• refactor: migrate prompt .replace() chains to substitute_placeholders by @andreatgretel in #89
• chore(ci): line up CI with fixes in safe-synthesizer repo by @mckornfield in #93
• docs: documentation site content by @lipikaramaswamy in #84
• docs: add step to activate venv by @nina-xu in #90
• feat: cli: expose rewrite mode on the CLI by @lipikaramaswamy in #78
• fix: remove stale substitution in detection workflow by @lipikaramaswamy in #97
• docs: update notebook csv inputs by @lipikaramaswamy in #96
• docs: add endpoint guidance and language coverage notes by @lipikaramaswamy in #99
• feat: Improve Sensitivity Disposition supplements by @asteier2026 in #94
• docs: clarify standard vs latent entity example by @lipikaramaswamy in #100
• fix: docs link fixes and notebook validation for release by @lipikaramaswamy in #102
• chore: fix stale package name in uv.lock by @lipikaramaswamy in #103
• test: clean up e2e smoke test and make target by @lipikaramaswamy in #101

New Contributors

• @lipikaramaswamy made their first contribution in #1
• @andreatgretel made their first contribution in #24
• @ko3n1g made their first contribution in #62
• @mckornfield made their first contribution in #69
• @nina-xu made their first contribution in #85
• @asteier2026 made their first contribution in #86

Full Changelog: https://github.com/NVIDIA-NeMo/Anonymizer/commits/v0.1.0rc1