Add TCK for deterministic recognizers

.github/workflows/ci.yml

@@ -204,6 +204,30 @@ jobs:
           pip install --require-hashes -r ${{ github.workspace }}/.github/pipelines/requirements-build.txt
           python -m build --wheel
 
+  tck-validation:
+    name: TCK Validation
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.0
+        with:
+          persist-credentials: false
+
+      - name: Set up Python
+        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: '3.13'
+
+      - name: Install uv
+        run: python -m pip install uv==0.11.6
+
+      - name: Generate and validate TCK
+        working-directory: tck
+        run: |
+          uv sync --locked --group dev
+          uv run python extract_tck.py
+          uv run python validate_tck.py
+
   build-platform-images:
     name: Build ${{ matrix.image }} (${{ matrix.platform }})
     runs-on: ${{ matrix.runner }}

.github/workflows/release.yml

@@ -46,22 +46,38 @@ jobs:
 
   github-release:
     name: Create GitHub Release
-    runs-on: ubuntu-slim
+    runs-on: ubuntu-latest
     needs: get-version
     permissions:
       contents: write
-    continue-on-error: true
     steps:
       - name: Checkout code
         uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.0
         with:
           persist-credentials: false
 
-      - name: Create GitHub Release       
+      - name: Set up Python
+        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: '3.13'
+
+      - name: Install uv
+        run: python -m pip install uv==0.11.6
+
+      - name: Generate and package TCK data
+        working-directory: tck
+        run: |
+          uv sync --locked --group dev
+          uv run python extract_tck.py
+          tar -czf "../presidio-tck-${{ needs.get-version.outputs.version }}.tar.gz" data
+
+      - name: Create GitHub Release
+        continue-on-error: true
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
         run: |
           gh release create "${{ needs.get-version.outputs.version }}" \
+            "presidio-tck-${{ needs.get-version.outputs.version }}.tar.gz" \
             --title "Release ${{ needs.get-version.outputs.version }}" \
             --notes "" \
             --draft \

CHANGELOG.md

@@ -4,6 +4,10 @@ All notable changes to this project will be documented in this file.
 
 ## [unreleased]
 
+### TCK
+#### Added
+- Added Technology Compatibility Kit (TCK) for deterministic recognizers as a portable YAML test suite (74 recognizers, 1478 test cases)
+
 ### Analyzer
 #### Added
 - Added `NoOpNlpEngine` for configurations that do not require NLP engine artifacts, enabling standalone recognizers such as `HuggingFaceNerRecognizer` to run without a spaCy or Stanza model (#2071) (Thanks @ultramancode)

tck/.gitignore

@@ -0,0 +1 @@
+data/

tck/README.md

@@ -0,0 +1,118 @@
+# Presidio TCK (Technology Compatibility Kit)
+
+Portable test suite for Presidio PII recognizers. Extracts test data from
+`@pytest.mark.parametrize` annotations in `presidio-analyzer/tests/` and
+produces YAML files that any compatible implementation can validate against.
+
+## Quick start
+
+```bash
+cd tck
+uv sync --group dev
+uv run python extract_tck.py    # generate YAML from Python tests
+uv run python validate_tck.py   # validate YAML against live Presidio
+```
+
+## How it works
+
+**Source of truth** stays in the Python test files. The YAML is a derived
+artifact, regenerated on demand.
+
+```
+presidio-analyzer/tests/test_*_recognizer.py
+        |
+        |  extract_tck.py (pytest import + AST + tokenize)
+        v
+tck/data/*.yaml  (one file per recognizer, with comments)
+        |
+        |  validate_tck.py (instantiate recognizer, run test cases)
+        v
+pass / fail
+```
+
+- `extract_tck.py` imports each test module via pytest to resolve
+  `@pytest.mark.parametrize` data (handles computed values, f-strings,
+  list multiplication). Comments are extracted from source via AST + regex.
+- `validate_tck.py` loads each YAML file, instantiates the named recognizer,
+  and runs every test case against the live Presidio implementation.
+- Both scripts are **idempotent** — safe to run repeatedly.
+
+## YAML format
+
+```yaml
+recognizer: AuAbnRecognizer
+entity_type: AU_ABN
+language: en
+country_code: au
+
+test_cases:
+  # Valid formatting and valid ABNs
+  - input: "51 824 753 556"
+    expected:
+      - start: 0
+        end: 14
+        score: 1.0
+
+  # Valid formatting but invalid ABNs
+  - input: "52 824 753 556"
+    expected: []  # should NOT match
+```
+
+See `schema/tck-recognizer-tests.schema.json` for the full JSON Schema.
+
+## Implementing validation in another language
+
+Download the TCK data from a GitHub release (or generate it with `extract_tck.py`)
+and validate your recognizer implementation against each YAML file. The algorithm
+is the same regardless of language:
+
+```
+for each YAML file:
+  instantiate the recognizer named in `recognizer`
+  for each test_case:
+    run the recognizer on test_case.input for entity_type
+    sort results by start position
+
+    if test_case has `expected_count`:
+      assert len(results) == expected_count
+    else:
+      assert len(results) == len(expected)
+      for each (result, exp) pair:
+        assert result.start == exp.start
+        assert result.end   == exp.end
+        if exp has `score`:     assert result.score == exp.score
+        if exp has `score_min`: assert result.score >= exp.score_min
+        if exp has `score_max`: assert result.score <= exp.score_max
+```
+
+A failing test should report the file, test index, what was expected vs. what the
+recognizer returned, and the input text. For example:
+
+```
+FAIL [0]: span[0]: got (0,16), expected (0,10) | '4012888888881881 ...'
+```
+
+See `validate_tck.py` for a complete Python reference implementation.
+
+## CI integration
+
+```yaml
+- name: Generate and validate TCK
+  run: |
+    python tck/extract_tck.py
+    python tck/validate_tck.py
+```
+
+The generated `data/` directory is gitignored — it is a build artifact,
+not committed. On releases it can be packaged as a `presidio-tck.tar.gz`.
+
+## Scope
+
+**Included** (~74 recognizers, ~1478 test cases):
+all deterministic `PatternRecognizer` subclasses (regex, checksum, deny-list)
+across 19 countries.
+
+**Excluded** (16 test files):
+NLP/ML model-dependent recognizers (spaCy, Stanza, Transformers, GLiNER,
+HuggingFace NER), remote/LLM services (Azure AI, LangExtract), and
+phone recognizers requiring NLP artifacts or custom config.