Skip to main content

​
What is Coverage?

Coverage tracking measures how well your generated scenarios cover different aspects of your policy—like code coverage but for training data.

​
Coverage Report

result = synkro.generate(policy, traces=50, return_logic_map=True)
print(result.coverage_report)
Output: 
Coverage Report
========================================
Overall: 68%
Sub-categories: 2 covered, 1 partial, 1 uncovered

Gaps:
  - Receipt requirements [HIGH] (0% coverage)
  - Travel booking rules [MEDIUM] (40% coverage)

Suggestions:
  1. Add 3+ scenarios for 'Receipt requirements'
  2. Add edge_case scenarios for 'Travel booking'

​
Coverage Statuses

StatusCoverageDescription
Covered80%+Well-tested
Partial30-80%Needs more scenarios
Uncoveredless than 30%Gap in coverage

​
Improving Coverage

​
Interactive Mode

Use natural language commands in HITL mode: 
Enter feedback: improve coverage for receipt requirements
✓ Added 5 scenarios for 'Receipt requirements'
Coverage: 0% → 72%

Enter feedback: get travel booking to 80%
✓ Added 3 scenarios for 'Travel booking'
Coverage: 40% → 85%

​
Programmatic

# Access coverage data
report = result.coverage_report

# Check gaps
for gap in report.gaps:
    print(f"Gap: {gap.sub_category} ({gap.coverage_percent}%)")

# Export as dict/JSON
data = report.to_dict()
json_str = report.to_json()

​
Coverage Metrics

Each sub-category tracks:
MetricDescription
coverage_percent% of expected coverage achieved
scenario_countNumber of scenarios testing this
type_distributionBreakdown by positive/negative/edge

​
Best Practices

Aim for 80%+

High coverage ensures your model sees diverse examples

Check Gaps

Uncovered areas = potential model blind spots

Balance Types

Mix of positive, negative, and edge cases

Iterate

Generate → Check Coverage → Improve → Repeat