flightline-ai 0.1.1

Flightline - Synthetic data generation from sample files

Flightline

Synthetic data generation from sample files.

The Problem

You can't download real production files to test locally due to security/compliance constraints. Flightline solves this by analyzing a sample file and generating N valid synthetic variations.

How It Works

1. Learn - Analyze a sample file to create a profile describing the schema, data types, business logic, and PII fields
2. Generate - Use the profile to generate N realistic synthetic records

Installation

pip install flightline-ai

For development:

# Clone the repo and install in editable mode
pip install -e .

Usage

Step 1: Learn from a sample file

flightline learn path/to/sample.json

This analyzes your file and creates a profile at ./flightline_output/profile.json.

Step 2: Generate synthetic data

flightline generate -n 10

This generates 10 synthetic records based on the profile, saved to ./flightline_output/synthetic_<timestamp>.json.

You can also use the shorthand:

flightline gen -n 10

Choosing a Model

Both commands support a --model flag to choose any model available on OpenRouter:

# Use Claude
flightline learn sample.json --model anthropic/claude-sonnet-4

# Use a cheaper model for generation
flightline gen -n 100 --model openai/gpt-4o-mini

See OpenRouter Models for all available models.

Environment Variables

• OPENROUTER_API_KEY - Your OpenRouter API key (required). Get one at https://openrouter.ai/keys

Output

All generated files are saved to ./flightline_output/:

• profile.json - The learned schema and rules
• synthetic_<timestamp>.json - The generated synthetic records