glam/docs/plan/specificity_score/02-tdd.md

Specificity Score System - Test-Driven Development Strategy

Overview

This document outlines the TDD approach for implementing the specificity scoring system. All components will be developed test-first to ensure correctness and maintainability.

INTEGRATION NOTE: Tests in this document should verify integration with existing infrastructure:

• The TemplateClassifier at backend/rag/template_sparql.py:1104 - DO NOT mock in integration tests
• The TemplateClassifierSignature at backend/rag/template_sparql.py:634 - verify compatibility
• SPARQL templates in data/sparql_templates.yaml - use as test fixtures

---

Existing Component Testing Strategy

Wrapper Pattern Tests

Since we're wrapping the existing TemplateClassifier rather than replacing it, add these tests:

# tests/integration/test_classifier_wrapper.py
import pytest
from backend.rag.template_sparql import TemplateClassifier
from specificity.wrapper import SpecificityAwareClassifier
from specificity.mapper import SPARQLToContextMapper

class TestSpecificityAwareClassifier:
    """Tests for wrapper around existing TemplateClassifier."""
    
    @pytest.fixture
    def wrapper(self):
        """Create wrapper with real TemplateClassifier."""
        base_classifier = TemplateClassifier()
        mapper = SPARQLToContextMapper()
        return SpecificityAwareClassifier(base_classifier, mapper)
    
    def test_delegates_to_base_classifier(self, wrapper):
        """Wrapper delegates classification to existing classifier."""
        result = wrapper.classify("What archives are in Drenthe?")
        
        # Verify it got a SPARQL template from base classifier
        assert result.sparql_template in [
            "list_institutions_by_type_city",
            "list_institutions_by_type_region",
        ]
    
    def test_maps_sparql_to_context_template(self, wrapper):
        """Wrapper maps SPARQL template ID to context template."""
        result = wrapper.classify("What archives are in Drenthe?")
        
        # SPARQL template → context template mapping
        assert result.context_template in ["archive_search", "location_browse"]
    
    def test_institution_type_refinement(self, wrapper):
        """Refines context template based on institution_type slot."""
        result = wrapper.classify("What museums are in Amsterdam?")
        
        # institution_type = M → museum_search
        if result.slots.get("institution_type") == "M":
            assert result.context_template == "museum_search"
    
    def test_returns_specificity_scores(self, wrapper):
        """Wrapper returns specificity scores for filtered classes."""
        result = wrapper.classify_with_scores("What archives are in Drenthe?")
        
        assert hasattr(result, "class_scores")
        assert "Archive" in result.class_scores
        assert result.class_scores["Archive"] > 0.8  # High for archive_search

SPARQL → Context Mapper Tests

# tests/unit/test_sparql_to_context_mapper.py
import pytest
from specificity.mapper import SPARQLToContextMapper

class TestSPARQLToContextMapper:
    """Tests for SPARQL template → context template mapping."""
    
    @pytest.fixture
    def mapper(self):
        return SPARQLToContextMapper()
    
    @pytest.mark.parametrize("sparql_template,expected_context", [
        ("list_institutions_by_type_city", "location_browse"),
        ("list_institutions_by_type_region", "location_browse"),
        ("find_institution_by_identifier", "identifier_lookup"),
        ("find_institutions_by_founding_date", "organizational_change"),
        ("none", "general_heritage"),
    ])
    def test_maps_sparql_to_context(self, mapper, sparql_template, expected_context):
        """Maps SPARQL template IDs to context template IDs."""
        context = mapper.map(sparql_template)
        assert context == expected_context
    
    @pytest.mark.parametrize("institution_type,expected_context", [
        ("A", "archive_search"),
        ("M", "museum_search"),
        ("L", "library_search"),
        ("G", "general_heritage"),  # Gallery falls back to general
    ])
    def test_refines_by_institution_type(self, mapper, institution_type, expected_context):
        """Refines context template based on institution_type slot."""
        context = mapper.map(
            "list_institutions_by_type_city",
            slots={"institution_type": institution_type}
        )
        assert context == expected_context
    
    def test_unknown_template_returns_general(self, mapper):
        """Unknown SPARQL templates map to general_heritage."""
        context = mapper.map("unknown_template_xyz")
        assert context == "general_heritage"

---

Testing Pyramid

                    /\
                   /  \
                  / E2E \        ← 5% - Full pipeline tests
                 /______\
                /        \
               / Integr.  \      ← 20% - Cross-component tests
              /____________\
             /              \
            /    Unit        \   ← 75% - Individual function tests
           /__________________\

---

Unit Tests

1. Score Repository Tests

# tests/unit/test_score_repository.py
import pytest
from specificity.repository import LinkMLScoreRepository, InMemoryScoreRepository

class TestInMemoryScoreRepository:
    """Tests for in-memory repository (used in other tests)."""
    
    def test_get_score_default(self):
        """Unknown classes return default score of 0.5."""
        repo = InMemoryScoreRepository()
        assert repo.get_score("UnknownClass") == 0.5
    
    def test_set_and_get_general_score(self):
        """Can set and retrieve general specificity score."""
        repo = InMemoryScoreRepository()
        repo.set_score("Archive", 0.75)
        assert repo.get_score("Archive") == 0.75
    
    def test_set_and_get_template_score(self):
        """Can set and retrieve template-specific score."""
        repo = InMemoryScoreRepository()
        repo.set_score("Archive", 0.95, template_id="archive_search")
        repo.set_score("Archive", 0.20, template_id="museum_search")
        
        assert repo.get_score("Archive", "archive_search") == 0.95
        assert repo.get_score("Archive", "museum_search") == 0.20
        assert repo.get_score("Archive") == 0.5  # General score unchanged
    
    def test_get_all_scores(self):
        """Can retrieve all scores for a template."""
        repo = InMemoryScoreRepository()
        repo.set_score("Archive", 0.95, "archive_search")
        repo.set_score("Museum", 0.30, "archive_search")
        repo.set_score("Library", 0.40, "archive_search")
        
        scores = repo.get_all_scores("archive_search")
        assert scores == {"Archive": 0.95, "Museum": 0.30, "Library": 0.40}
    
    def test_bulk_update(self):
        """Can update multiple scores at once."""
        repo = InMemoryScoreRepository()
        repo.bulk_update({
            "Archive": 0.90,
            "Museum": 0.85,
            "Library": 0.80
        }, template_id="general_heritage")
        
        assert repo.get_score("Archive", "general_heritage") == 0.90
        assert repo.get_score("Museum", "general_heritage") == 0.85
        assert repo.get_score("Library", "general_heritage") == 0.80


class TestLinkMLScoreRepository:
    """Tests for LinkML file-based repository."""
    
    @pytest.fixture
    def temp_schema_dir(self, tmp_path):
        """Create temporary schema directory with sample class files."""
        classes_dir = tmp_path / "modules" / "classes"
        classes_dir.mkdir(parents=True)
        
        # Create sample class file
        archive_yaml = classes_dir / "Archive.yaml"
        archive_yaml.write_text("""
classes:
  Archive:
    is_a: HeritageCustodian
    description: An archive institution
    annotations:
      specificity_score: 0.75
      template_specificity:
        archive_search: 0.95
        museum_search: 0.20
""")
        return tmp_path
    
    def test_load_existing_score(self, temp_schema_dir):
        """Loads existing score from YAML file."""
        repo = LinkMLScoreRepository(temp_schema_dir)
        assert repo.get_score("Archive") == 0.75
        assert repo.get_score("Archive", "archive_search") == 0.95
    
    def test_update_score_persists(self, temp_schema_dir):
        """Updated scores are persisted to file."""
        repo = LinkMLScoreRepository(temp_schema_dir)
        repo.set_score("Archive", 0.80)
        
        # Create new repository instance to verify persistence
        repo2 = LinkMLScoreRepository(temp_schema_dir)
        assert repo2.get_score("Archive") == 0.80
    
    def test_missing_class_raises_error(self, temp_schema_dir):
        """Raises ValueError for non-existent class."""
        repo = LinkMLScoreRepository(temp_schema_dir)
        with pytest.raises(ValueError, match="Class file not found"):
            repo.get_score("NonExistentClass")

2. Scoring Strategy Tests

# tests/unit/test_scoring_strategies.py
import pytest
from specificity.strategies import (
    ArchiveSearchStrategy,
    MuseumSearchStrategy,
    LocationBrowseStrategy,
    ScoringStrategyFactory
)

class TestArchiveSearchStrategy:
    """Tests for archive search scoring strategy."""
    
    @pytest.fixture
    def strategy(self):
        return ArchiveSearchStrategy()
    
    def test_archive_class_high_score(self, strategy):
        """Archive class gets high score for archive_search template."""
        score = strategy.calculate_score("Archive", {})
        assert score >= 0.90
    
    def test_museum_class_low_score(self, strategy):
        """Museum class gets low score for archive_search template."""
        score = strategy.calculate_score("Museum", {})
        assert score <= 0.30
    
    def test_location_class_moderate_score(self, strategy):
        """Location class gets moderate score (useful across templates)."""
        score = strategy.calculate_score("Location", {})
        assert 0.50 <= score <= 0.70
    
    def test_archival_properties_boost(self, strategy):
        """Classes with archival properties get score boost."""
        metadata_without = {"slots": ["name", "description"]}
        metadata_with = {"slots": ["name", "finding_aid", "extent"]}
        
        score_without = strategy.calculate_score("Collection", metadata_without)
        score_with = strategy.calculate_score("Collection", metadata_with)
        
        assert score_with > score_without
    
    def test_template_id(self, strategy):
        """Strategy returns correct template ID."""
        assert strategy.get_template_id() == "archive_search"


class TestScoringStrategyFactory:
    """Tests for strategy factory."""
    
    def test_get_registered_strategy(self):
        """Factory returns registered strategy."""
        strategy = ScoringStrategyFactory.get_strategy("archive_search")
        assert isinstance(strategy, ArchiveSearchStrategy)
    
    def test_get_unregistered_returns_default(self):
        """Factory returns default strategy for unknown template."""
        strategy = ScoringStrategyFactory.get_strategy("unknown_template")
        assert strategy.get_template_id() == "default"
    
    def test_register_custom_strategy(self):
        """Can register custom strategy."""
        class CustomStrategy(ScoringStrategy):
            def calculate_score(self, class_name, metadata):
                return 0.99
            def get_template_id(self):
                return "custom"
        
        ScoringStrategyFactory.register(CustomStrategy())
        strategy = ScoringStrategyFactory.get_strategy("custom")
        assert strategy.calculate_score("Any", {}) == 0.99

3. Score Calculator (Decorator) Tests

# tests/unit/test_score_calculator.py
import pytest
from specificity.calculator import (
    BaseScoreCalculator,
    CustodianTypeBoostDecorator,
    InheritanceDepthDecorator
)
from specificity.repository import InMemoryScoreRepository

class TestBaseScoreCalculator:
    """Tests for base score calculator."""
    
    @pytest.fixture
    def calculator(self):
        repo = InMemoryScoreRepository()
        repo.set_score("Archive", 0.75)
        repo.set_score("Archive", 0.95, "archive_search")
        return BaseScoreCalculator(repo)
    
    def test_returns_general_score(self, calculator):
        """Returns general score when no template specified."""
        score = calculator.calculate("Archive", None)
        assert score == 0.75
    
    def test_returns_template_score(self, calculator):
        """Returns template-specific score when template specified."""
        score = calculator.calculate("Archive", "archive_search")
        assert score == 0.95
    
    def test_fallback_to_general_for_unknown_template(self, calculator):
        """Falls back to general score for unknown template."""
        score = calculator.calculate("Archive", "unknown_template")
        assert score == 0.75


class TestCustodianTypeBoostDecorator:
    """Tests for custodian type boost decorator."""
    
    @pytest.fixture
    def decorated_calculator(self):
        repo = InMemoryScoreRepository()
        repo.set_score("Archive", 0.70)
        
        # Mock schema loader with custodian_types annotation
        schema_loader = MockSchemaLoader({
            "Archive": {"annotations": {"custodian_types": ["A", "O"]}},
            "Museum": {"annotations": {"custodian_types": ["M"]}},
            "Location": {"annotations": {"custodian_types": ["*"]}}  # Universal
        })
        
        base = BaseScoreCalculator(repo)
        return CustodianTypeBoostDecorator(base, custodian_type="A")
    
    def test_boost_matching_custodian_type(self, decorated_calculator):
        """Boosts score for matching custodian type."""
        score = decorated_calculator.calculate("Archive", None)
        assert score == 0.85  # 0.70 + 0.15 boost
    
    def test_no_boost_non_matching_type(self, decorated_calculator):
        """No boost for non-matching custodian type."""
        score = decorated_calculator.calculate("Museum", None)
        assert score == 0.50  # Default score, no boost
    
    def test_boost_universal_type(self, decorated_calculator):
        """Boosts score for universal (*) custodian type."""
        score = decorated_calculator.calculate("Location", None)
        assert score >= 0.65  # Gets boost because "*" matches all


class TestInheritanceDepthDecorator:
    """Tests for inheritance depth decorator."""
    
    @pytest.fixture
    def decorated_calculator(self):
        repo = InMemoryScoreRepository()
        repo.set_score("HeritageCustodian", 0.10)
        repo.set_score("Archive", 0.70)
        repo.set_score("NationalArchive", 0.80)
        
        # Mock schema loader with inheritance hierarchy
        schema_loader = MockSchemaLoader({
            "HeritageCustodian": {},  # Root class, depth 0
            "Archive": {"is_a": "HeritageCustodian"},  # Depth 1
            "NationalArchive": {"is_a": "Archive"}  # Depth 2
        })
        
        base = BaseScoreCalculator(repo)
        return InheritanceDepthDecorator(base, schema_loader)
    
    def test_root_class_no_boost(self, decorated_calculator):
        """Root class (depth 0) gets no depth boost."""
        score = decorated_calculator.calculate("HeritageCustodian", None)
        assert score == 0.10
    
    def test_child_class_small_boost(self, decorated_calculator):
        """Direct child (depth 1) gets small boost."""
        score = decorated_calculator.calculate("Archive", None)
        assert score == 0.73  # 0.70 + 0.03 (depth 1 * 0.03)
    
    def test_grandchild_class_larger_boost(self, decorated_calculator):
        """Grandchild (depth 2) gets larger boost."""
        score = decorated_calculator.calculate("NationalArchive", None)
        assert score == 0.86  # 0.80 + 0.06 (depth 2 * 0.03)

4. Command Tests

# tests/unit/test_commands.py
import pytest
from datetime import datetime
from specificity.commands import (
    UpdateScoreCommand,
    BatchScoreCommand,
    ScoreCommandInvoker
)
from specificity.repository import InMemoryScoreRepository

class TestUpdateScoreCommand:
    """Tests for single score update command."""
    
    def test_execute_updates_score(self):
        """Execute updates the score in repository."""
        repo = InMemoryScoreRepository()
        repo.set_score("Archive", 0.50)
        
        cmd = UpdateScoreCommand(repo, "Archive", 0.75)
        change = cmd.execute()
        
        assert repo.get_score("Archive") == 0.75
        assert change.old_score == 0.50
        assert change.new_score == 0.75
    
    def test_undo_restores_original(self):
        """Undo restores the original score."""
        repo = InMemoryScoreRepository()
        repo.set_score("Archive", 0.50)
        
        cmd = UpdateScoreCommand(repo, "Archive", 0.75)
        cmd.execute()
        cmd.undo()
        
        assert repo.get_score("Archive") == 0.50
    
    def test_execute_records_metadata(self):
        """Execute records change metadata."""
        repo = InMemoryScoreRepository()
        
        cmd = UpdateScoreCommand(
            repo, "Archive", 0.75,
            author="kempersc",
            rationale="User feedback indicated higher relevance"
        )
        change = cmd.execute()
        
        assert change.author == "kempersc"
        assert "feedback" in change.rationale
        assert isinstance(change.timestamp, datetime)


class TestBatchScoreCommand:
    """Tests for batch score update command."""
    
    def test_execute_all_updates(self):
        """Execute updates all scores in batch."""
        repo = InMemoryScoreRepository()
        
        batch = BatchScoreCommand([
            UpdateScoreCommand(repo, "Archive", 0.90),
            UpdateScoreCommand(repo, "Museum", 0.85),
            UpdateScoreCommand(repo, "Library", 0.80),
        ])
        changes = batch.execute()
        
        assert len(changes) == 3
        assert repo.get_score("Archive") == 0.90
        assert repo.get_score("Museum") == 0.85
        assert repo.get_score("Library") == 0.80
    
    def test_rollback_on_failure(self):
        """Rollback all changes if any update fails."""
        repo = InMemoryScoreRepository()
        repo.set_score("Archive", 0.50)
        repo.set_score("Museum", 0.50)
        
        # Create a command that will fail
        class FailingCommand(UpdateScoreCommand):
            def execute(self):
                raise ValueError("Simulated failure")
        
        batch = BatchScoreCommand([
            UpdateScoreCommand(repo, "Archive", 0.90),
            FailingCommand(repo, "Fail", 0.00),
            UpdateScoreCommand(repo, "Museum", 0.85),
        ])
        
        with pytest.raises(ValueError):
            batch.execute()
        
        # First update should be rolled back
        assert repo.get_score("Archive") == 0.50


class TestScoreCommandInvoker:
    """Tests for command invoker with undo/redo."""
    
    def test_undo_reverts_last_command(self):
        """Undo reverts the most recent command."""
        repo = InMemoryScoreRepository()
        repo.set_score("Archive", 0.50)
        
        invoker = ScoreCommandInvoker()
        invoker.execute(UpdateScoreCommand(repo, "Archive", 0.75))
        invoker.execute(UpdateScoreCommand(repo, "Archive", 0.90))
        
        assert repo.get_score("Archive") == 0.90
        
        invoker.undo()
        assert repo.get_score("Archive") == 0.75
        
        invoker.undo()
        assert repo.get_score("Archive") == 0.50
    
    def test_redo_reapplies_undone_command(self):
        """Redo reapplies an undone command."""
        repo = InMemoryScoreRepository()
        
        invoker = ScoreCommandInvoker()
        invoker.execute(UpdateScoreCommand(repo, "Archive", 0.75))
        invoker.undo()
        
        assert repo.get_score("Archive") == 0.50
        
        invoker.redo()
        assert repo.get_score("Archive") == 0.75
    
    def test_redo_cleared_after_new_execute(self):
        """Redo stack is cleared after new command execution."""
        repo = InMemoryScoreRepository()
        
        invoker = ScoreCommandInvoker()
        invoker.execute(UpdateScoreCommand(repo, "Archive", 0.75))
        invoker.undo()
        invoker.execute(UpdateScoreCommand(repo, "Archive", 0.80))
        
        invoker.redo()  # Should have no effect
        assert repo.get_score("Archive") == 0.80

---

Integration Tests

1. RAG Pipeline Integration

# tests/integration/test_rag_integration.py
import pytest
from specificity.rag import SpecificityFilteredRetriever
from specificity.repository import InMemoryScoreRepository

class TestSpecificityFilteredRetriever:
    """Tests for RAG retriever with specificity filtering."""
    
    @pytest.fixture
    def retriever(self):
        repo = InMemoryScoreRepository()
        # Set up scores for archive_search template
        repo.set_score("Archive", 0.95, "archive_search")
        repo.set_score("Collection", 0.85, "archive_search")
        repo.set_score("Location", 0.70, "archive_search")
        repo.set_score("Museum", 0.20, "archive_search")
        repo.set_score("Gallery", 0.15, "archive_search")
        
        return SpecificityFilteredRetriever(repo, threshold=0.5)
    
    def test_filters_low_score_classes(self, retriever):
        """Filters out classes below threshold."""
        classes = retriever.get_relevant_classes("archive_search")
        
        assert "Archive" in classes
        assert "Collection" in classes
        assert "Location" in classes
        assert "Museum" not in classes
        assert "Gallery" not in classes
    
    def test_ranks_by_score(self, retriever):
        """Returns classes ranked by specificity score."""
        classes = retriever.get_relevant_classes("archive_search")
        
        assert classes.index("Archive") < classes.index("Collection")
        assert classes.index("Collection") < classes.index("Location")
    
    def test_threshold_adjustment(self, retriever):
        """Can adjust threshold at query time."""
        classes = retriever.get_relevant_classes("archive_search", threshold=0.80)
        
        assert "Archive" in classes
        assert "Collection" in classes
        assert "Location" not in classes  # 0.70 < 0.80

2. Template Classifier Integration

# tests/integration/test_template_classifier.py
import pytest
from specificity.classifier import TemplateClassifier

class TestTemplateClassifier:
    """Tests for question-to-template classification."""
    
    @pytest.fixture
    def classifier(self):
        return TemplateClassifier()
    
    @pytest.mark.parametrize("question,expected_template", [
        ("What archives are in Drenthe?", "archive_search"),
        ("Which museums have art collections?", "museum_search"),
        ("Find libraries in Amsterdam", "library_search"),
        ("Who is the director of the Rijksmuseum?", "person_research"),
        ("Where is the National Archive located?", "location_browse"),
        ("What is the ISIL code for Amsterdam Museum?", "identifier_lookup"),
        ("When did the museum merge with the archive?", "organizational_change"),
    ])
    def test_classifies_common_questions(self, classifier, question, expected_template):
        """Classifies common question patterns correctly."""
        result = classifier.classify(question)
        assert result.template_id == expected_template
    
    def test_returns_confidence_score(self, classifier):
        """Returns confidence score with classification."""
        result = classifier.classify("What archives are in Drenthe?")
        
        assert hasattr(result, "confidence")
        assert 0.0 <= result.confidence <= 1.0
    
    def test_fallback_for_ambiguous_questions(self, classifier):
        """Falls back to general_heritage for ambiguous questions."""
        result = classifier.classify("Tell me about heritage in the Netherlands")
        
        assert result.template_id == "general_heritage"

---

End-to-End Tests

1. Full Pipeline Test

# tests/e2e/test_full_pipeline.py
import pytest
from specificity.pipeline import SpecificityPipeline

class TestFullPipeline:
    """End-to-end tests for complete specificity scoring pipeline."""
    
    @pytest.fixture
    def pipeline(self, real_schema_dir):
        """Create pipeline with real schema files."""
        return SpecificityPipeline(real_schema_dir)
    
    def test_archive_question_filters_correctly(self, pipeline):
        """Archive question filters to archive-relevant classes."""
        question = "What archives in Noord-Holland have photo collections?"
        
        result = pipeline.process(question)
        
        # Should include archive-relevant classes
        assert "Archive" in result.relevant_classes
        assert "Collection" in result.relevant_classes
        assert "Location" in result.relevant_classes
        
        # Should exclude irrelevant classes
        assert "LinkedInConnectionExtraction" not in result.relevant_classes
        assert "PersonProfileExtraction" not in result.relevant_classes
    
    def test_person_question_includes_person_classes(self, pipeline):
        """Person research question includes person-relevant classes."""
        question = "Who are the curators at the Rijksmuseum?"
        
        result = pipeline.process(question)
        
        assert "PersonProfile" in result.relevant_classes
        assert "Staff" in result.relevant_classes
        assert "HeritageCustodian" in result.relevant_classes
    
    def test_context_reduces_class_count(self, pipeline):
        """Template-specific filtering reduces total class count."""
        # Without filtering
        all_classes = pipeline.get_all_classes()
        
        # With filtering
        question = "What archives are in Drenthe?"
        result = pipeline.process(question)
        
        # Filtered should be significantly smaller
        assert len(result.relevant_classes) < len(all_classes) * 0.5

---

Test Fixtures

Shared Fixtures

# tests/conftest.py
import pytest
from pathlib import Path
import yaml

@pytest.fixture
def temp_schema_dir(tmp_path):
    """Create temporary schema directory with sample classes."""
    classes_dir = tmp_path / "modules" / "classes"
    classes_dir.mkdir(parents=True)
    
    sample_classes = {
        "HeritageCustodian": {
            "description": "Base class for heritage institutions",
            "annotations": {"specificity_score": 0.10}
        },
        "Archive": {
            "is_a": "HeritageCustodian",
            "description": "An archive institution",
            "annotations": {
                "specificity_score": 0.75,
                "template_specificity": {
                    "archive_search": 0.95,
                    "museum_search": 0.20
                }
            }
        },
        "Museum": {
            "is_a": "HeritageCustodian",
            "description": "A museum institution",
            "annotations": {
                "specificity_score": 0.75,
                "template_specificity": {
                    "archive_search": 0.20,
                    "museum_search": 0.95
                }
            }
        },
        "Location": {
            "description": "Geographic location",
            "annotations": {
                "specificity_score": 0.30,
                "custodian_types": ["*"]
            }
        }
    }
    
    for class_name, class_def in sample_classes.items():
        yaml_path = classes_dir / f"{class_name}.yaml"
        yaml_path.write_text(yaml.dump({"classes": {class_name: class_def}}))
    
    return tmp_path


@pytest.fixture
def real_schema_dir():
    """Path to real schema directory for integration tests."""
    return Path("/Users/kempersc/apps/glam/schemas/20251121/linkml")


class MockSchemaLoader:
    """Mock schema loader for unit tests."""
    
    def __init__(self, classes: dict):
        self._classes = classes
    
    def get_class(self, class_name: str) -> dict:
        return self._classes.get(class_name, {})

---

Test Coverage Targets

Component	Target Coverage
Repository	95%
Strategies	90%
Calculator/Decorators	90%
Commands	95%
Classifier	85%
RAG Integration	80%
UML Integration	80%
Overall	85%

---

CI/CD Integration

# .github/workflows/specificity-tests.yml
name: Specificity Score Tests

on:
  push:
    paths:
      - 'src/specificity/**'
      - 'tests/**/*specificity*'
      - 'schemas/**/classes/**'

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      
      - name: Set up Python
        uses: actions/setup-python@v5
        with:
          python-version: '3.11'
      
      - name: Install dependencies
        run: |
          pip install -r requirements-dev.txt          
      
      - name: Run unit tests
        run: |
          pytest tests/unit -v --cov=src/specificity --cov-report=xml          
      
      - name: Run integration tests
        run: |
          pytest tests/integration -v          
      
      - name: Upload coverage
        uses: codecov/codecov-action@v4
        with:
          file: coverage.xml

---

References

• pytest documentation: https://docs.pytest.org/
• pytest-cov: https://pytest-cov.readthedocs.io/
• TDD best practices: https://martinfowler.com/bliki/TestDrivenDevelopment.html