Skip to main content

​
The Cartographer

The Cartographer is Flightline’s scenario mapping engine. It identifies what needs to be tested before any data is generated.

​
What It Does

Before generating random test data, the Cartographer analyzes your schema to identify dimensions of variance: the axes along which your inputs can vary and potentially cause different behaviors. 
Input:  Pydantic Schema (e.g., LoanApplication)
        ↓
Action: Analyze schema to identify dimensions
        ↓
Output: Test Matrix of 10-20 distinct scenarios

​
Why Mapping Matters

Most test generation tools create random data. This leads to:
  • Redundant test cases (100 “normal” scenarios)
  • Missing edge cases (the one bankruptcy case you needed)
  • Wasted compute on uninteresting variations
The Cartographer takes a different approach: map first, then generate.

​
How It Works

​
Step 1: Dimension Detection

Given a schema, the Cartographer identifies dimensions: 
class LoanApplication(BaseModel):
    credit_score: int          # → Dimension: Credit
    annual_income: float       # → Dimension: Income
    loan_type: LoanType        # → Dimension: LoanType (enum)
    loan_amount: float         # → Dimension: Magnitude
    employment_status: str     # → Dimension: Employment
    bankruptcy_history: bool   # → Dimension: Risk Flag

​
Step 2: Edge Case Identification

For each dimension, the Cartographer identifies interesting regions:
DimensionRegions
Creditlow, fair, good, excellent
Incomezero, low, median, high
LoanTypeFHA, Conventional, Jumbo, VA
Risk Flagtrue, false

​
Step 3: Scenario Generation

Dimensions are combined to create targeted scenarios:
ScenarioDescription
1High credit, low income, Jumbo loan
2Bankruptcy within 7 years, FHA loan
3Zero income, employed status (inconsistency test)
4Maximum loan amount, minimum credit
Scenario #3 tests for internal inconsistency: a state that shouldn’t be valid but might slip through without proper validation.

​
The Latent Space

We call the space of all possible inputs the “latent space.” The Cartographer’s job is to identify:
  • Clusters: Groups of similar inputs (the normal cases)
  • Boundaries: Edges where behavior might change (credit score thresholds)
  • Voids: Untested regions that could hide bugs

​
What’s Next

Once the Cartographer has mapped the scenarios, the Fabricator generates the actual test data.

The Fabricator

Generate synthetic data from mapped scenarios.

The Fact-Checker

Validate outputs against generated inputs.