Synth Data Generation

When the user wants to generate synthetic data, do not run raw synth commands. Use the included Python script which handles generation, JSON processing, and database loading automatically.

Prerequisites

1. Synth CLI must be installed: Run synth version to verify.

   • If not installed, use @install-synth agent or run:

     python3 .github/skills/install-synth/install_synth.py
     

2. Database must be initialized: Run make db-init if tables don't exist.

Standard Generation Procedure

Run the following commands in order:

1. Navigate to Project Root:

   cd airline-discount-ml
   

2. Run the Generation Script: This script generates data using Synth, processes the JSON, and loads it into SQLite.

   python3 ../.github/skills/generate-dataset-synth/run_synth.py
   

   With custom record count (default is 100 per table):

   python3 ../.github/skills/generate-dataset-synth/run_synth.py --count 500
   

   Generate only (no database load):

   python3 ../.github/skills/generate-dataset-synth/run_synth.py --no-load
   

3. Verify Data:

   sqlite3 data/airline_discount.db "SELECT COUNT(*) FROM passengers;"
   sqlite3 data/airline_discount.db "SELECT COUNT(*) FROM routes;"
   sqlite3 data/airline_discount.db "SELECT COUNT(*) FROM discounts;"
   

Synth Models Location

The Synth schema files are located in:

airline-discount-ml/synth_models/airline_data/
├── passengers.json    # Passenger profiles with travel history
├── routes.json        # Flight routes with origin/destination/distance
└── discounts.json     # Discount records linking passengers to routes

Generated Data Schema

passengers

routes

discounts

Troubleshooting

synth: command not found

• Run @install-synth or install manually per SKILL.md in install-synth.

"no such table" error

• Run make db-init to create tables first.

JSON parse errors

• Ensure Synth version is 0.6.9 (synth version).
• Check that synth_models/*.json files are valid JSON.