Datacrafter

AI-powered, schema-driven synthetic data generation platform.

Design datasets using YAML or generate them using natural language, and produce realistic data in CSV / JSON / JSONL / XML / Parquet formats.

---

✨ Key Highlights

• Schema-driven generation using YAML
• AI-powered schema creation from natural language prompts
• Formula engine for dynamic, cross-field computations
• Deterministic output using seed control
• Multiple formats: CSV, JSON, JSONL, XML, Parquet
• Plugin architecture for extensibility
• CLI + Python API

---

🤖 AI Schema Generation

Generate dataset schemas directly from prompts and save to a file:

datacrafter ai --prompt "
Generate a banking transactions dataset with:

transaction_id (uuid),
account_number (integer between 100000000000 and 999999999999),
transaction_type (categorical: Debit, Credit, Withdrawal, Deposit, Transfer),
amount (float between 50 and 50000),
currency (categorical: USD, EUR, GBP, INR),
timestamp (datetime between 2023-01-01 and 2025-12-31 in '%Y-%m-%d %H:%M:%S'),
merchant_name (categorical: Amazon, Walmart, Starbucks, Uber, Apple Store, Shell Fuel Station, Best Buy, ATM Withdrawal, Bank Transfer).

Output format: xml.
Return ONLY valid Datacrafter YAML schema.
" --out examples/banking_transactions.yaml

✔ AI-generated schema will be saved to the specified output file.

---

🧮 Formula Engine

Create dynamic fields using expressions:

total_price:
  type: formula
  expr: "price * quantity"

Supports:

• Arithmetic operations
• Comparisons and boolean logic
• Ternary expressions
• String concatenation
• Cross-field access

---

📦 Installation

pip install datacrafter-ai

Requirements: Python 3.9+

---

🚀 Quickstart

1. Create Schema

version: 1
rows: 10

fields:
  price:
    type: float

  quantity:
    type: integer

  total:
    type: formula
    expr: "price * quantity"

output:
  format: csv
  path: ./output/data.csv

---

2. Generate Data

datacrafter generate --schema schema.yaml

---

🧩 Built-in Capabilities

Providers

• uuid, id.incremental
• integer, float, boolean
• person., text., string.regex
• datetime, categorical, geo.country
• formula

Features

• Unique constraints
• Null handling
• Regex validation
• Distributions
• Templating & dependencies
• Cross-field computation

---

🖥️ CLI Commands

datacrafter generate --schema schema.yaml
datacrafter validate --schema schema.yaml
datacrafter list providers
datacrafter list writers
datacrafter init --template minimal
datacrafter ai --prompt "..." --out schema.yaml

---

🔌 Extensibility

Datacrafter supports plugins for:

• Custom providers
• Custom writers

No core modification required.

---

⚙️ AI Configuration

Datacrafter’s AI features support multiple LLM providers and require API credentials.

1. Create a .env file

Copy the example configuration:

cp .env.example .env

---

2. Configure your provider and model

Edit .env and choose one provider:

# Choose one provider: openrouter / openai / gemini / groq / deepseek
LLM_PROVIDER=openai

# Choose the model supported by the provider
LLM_MODEL=gpt-4

---

3. Add the corresponding API key

Provide ONLY the API key for your selected provider:

OPENAI_API_KEY=your_api_key_here

Examples for other providers:

OPENROUTER_API_KEY=your_key
GEMINI_API_KEY=your_key
GROQ_API_KEY=your_key
DEEPSEEK_API_KEY=your_key

---

4. Run AI schema generation

datacrafter ai --prompt "..." --out schema.yaml

---

⚠️ Important:

• AI features will NOT work without valid API credentials
• Only one provider needs to be configured
• Ensure the selected model is supported by the chosen provider

---

📦 Development

python -m build
twine check dist/*
twine upload dist/*

---

🔒 License

MIT © 2026 Mahalakshmi Shanmuga Sundaram

---

🏢 About

Datacrafter is developed and maintained by DHS Tech Services.