Get Started

Get up and running with SDG Hub in two steps using Claude Code.

Prerequisites

Python 3.10+
Claude Code installed (instructions)

!!! tip "Prefer manual installation?" See the full Installation guide for pip/uv setup without Claude Code.

Step 1: Bootstrap SDG Hub

Run this command to install SDG Hub and clone the repository:

curl -fsSL https://raw.githubusercontent.com/Red-Hat-AI-Innovation-Team/sdg_hub/main/scripts/bootstrap.sh | claude --dangerously-skip-permissions

What this does

Installs the sdg-hub Python package (via uv or pip)
Clones the SDG Hub repository, which includes the synthetic-data-generation Claude Code skill
Verifies the installation by discovering available blocks and flows

Step 2: Start generating data

Navigate to the cloned repository and launch Claude Code:

cd ~/sdg_hub && claude

Claude Code automatically discovers the synthetic-data-generation skill bundled in the repository. Describe what data you need, and it will select the right blocks and flows.

Example prompts

What you need	Example prompt
Question-answer pairs	"Generate 50 QA pairs from my PDF at `./data/manual.pdf`"
Instruction tuning data	"Create instruction-response pairs for customer support using `./data/tickets.csv`"
Red-team evaluation set	"Build a red-team dataset to test my chatbot's safety guardrails"
RAG evaluation	"Generate evaluation questions from the documents in `./data/knowledge_base/`"
Multi-step pipeline	"Create a flow that extracts text, generates questions, then filters by difficulty"

What's next

Quick Start -- build a pipeline step by step in Python
Core Concepts -- understand blocks, flows, and registries
Built-in Flows -- browse ready-to-use pipelines
Custom Flows -- author your own YAML flow