Skip to main content

​
flightline mimic

Generate synthetic variations from an existing sample file. This is the fastest way to create test data when you have an example but can’t share the original due to PII concerns.

​
Usage

flightline mimic <input-file>

​
The Problem It Solves

Developers at companies handling sensitive data face a blocking problem:
“I need to test my LLM feature against realistic credit reports, tax documents, and loan applications. But I can’t download production data due to PII and security policies.”
The mimic command analyzes the structure of your sample file and generates synthetic variations that:
  • Match the exact schema and field types
  • Follow inferred business rules
  • Contain zero PII, making them safe to share and commit

​
Example

$ flightline mimic customer_orders.json

► Analyzing schema structure...
► Detecting PII fields...
► Inferring business rules...

✓ LEARNING COMPLETE

  Profile saved: profile.json
  Fields: 12  |  PII detected: 9
  Business rules inferred: 13

► Generating 50 synthetic records...

✓ GENERATION COMPLETE

  Records: 50  |  Output: synthetic.json
  PII in output: 0

​
How It Works

  1. Schema Analysis: Flightline analyzes your sample to detect field types, PII fields, and nested structures
  2. Rule Inference: Business rules are inferred from the data (e.g., “if employed, income > 0”)
  3. Generation: Synthetic data is generated following all constraints

​
Use Cases

  • Local Development: Can’t access production data? Mimic one sample and generate hundreds of variations locally.
  • CI/CD Testing: Create a synthetic test suite that covers edge cases without exposing customer data.
  • Sharing Examples: Need to share a realistic file with a contractor or in documentation? Mimic creates shareable versions.

Learn More: The Fabricator

Deep dive into how synthetic data generation works.