SDG Hub plugin
A synthetic data Python framework with a Claude Code and Codex plugin: blocks chain into YAML flows that generate datasets
- 171 ★ on the repository
- Last commit
Extended access
Why this level
- Shell scripts run generation and spend your model provider paid tokens
- The plugin reads provider keys from the environment to call models
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/sdg-hubDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Shell scripts run generation and spend your model provider paid tokens
- The plugin reads provider keys from the environment to call models
This is third-party code. Review the repository files before installing.
What it does
SDG Hub builds synthetic data pipelines from blocks: LLM calls, parsing, transforms, filtering and agent steps. Blocks are chained into YAML flows, and a flow then runs generation on your dataset. The package installs with `pip install sdg-hub`, built-in flows are found through `FlowRegistry`, and a flow such as "MCP Server Distillation" gets its model with `flow.set_model_config`. The plugin brings this into the agent chat. For Claude Code it installs from the Red Hat AI Innovation Team marketplace, and Codex has its own manifest. Three skills ship inside: `setup-guide` configures the provider and model on first run, `flow-browser` lists and inspects available flows, and `data-generation` runs a flow and returns the result. A detailed reference on blocks, the YAML schema and patterns lives in the `synthetic-data-generation` skill. The scripts `sdg_detect.sh`, `sdg_flows.sh` and `sdg_generate.sh` are allowed to the skills only for their own tasks. Keys and model come from the environment, with a sample in `.env.example`.
Who it is for. Machine learning engineers who prepare training datasets and want to run generation flows from Claude Code or Codex.
Good fit when
- You need question and answer pairs from your documents and want to run the flow right from the agent chat
- First run: `setup-guide` configures the provider and model so a flow works
- A built-in flow for the task is picked and inspected through `flow-browser`
- You need to write your own YAML flow and want to ask the agent about blocks and the schema
Not a fit when
- You need real user data: the framework generates synthetic data and does not clean real data
- You want a quality audit of a finished dataset: flows generate data and the description has no separate audit of ready sets
- You need a single text or a couple of hand-made examples: a flow is overkill for that
Example request
Show the available SDG Hub flows, pick one for question and answer pairs from my documents and run it on the docs/ folderLimitations
Generation calls a language model under your key and is billed at the provider rate, and large datasets burn many tokens. The plugin carries the shell scripts `sdg_generate.sh`, `sdg_detect.sh` and `sdg_flows.sh`, so for Claude Code it allows running only those. The framework installs from PyPI separately, and some blocks need extra dependencies listed in the docs. The repository has hooks `commit-on-stop.sh`, `track-read.sh` and `verify-gate.sh` for developing the project itself, and they do not land in your project. For Codex, installation follows `.codex-plugin/INSTALL.md`.
How to disable. In Claude Code run `/plugin uninstall sdg-hub` and, if needed, remove the marketplace with `/plugin marketplace remove`. The Python package is removed with `pip uninstall sdg-hub`.
FAQ
How do I install SDG Hub in Claude Code?
Run `/plugin marketplace add Red-Hat-AI-Innovation-Team/plugins`, then `/plugin install sdg-hub@Red-Hat-AI-Innovation-Team/plugins`. You can also add the repository itself: `/plugin marketplace add Red-Hat-AI-Innovation-Team/sdg_hub`.
What do the plugin skills do?
`setup-guide` handles first-time setup, `flow-browser` finds and describes flows and `data-generation` runs the chosen flow. The reference skill `synthetic-data-generation` answers questions about blocks and the YAML schema.
Can I use the framework without the plugin?
Yes. Install `pip install sdg-hub`, load a flow with `Flow.from_yaml`, set the model with `flow.set_model_config` and call `flow.generate(dataset)`.
How do I connect Codex?
Run `codex plugin marketplace add Red-Hat-AI-Innovation-Team/plugins` and install the plugin from the marketplace. Manual installation is described in `.codex-plugin/INSTALL.md`.
What blocks make up flows?
LLM call, parsing, transform, filtering and agent blocks. A block reference is in the project docs and in the `synthetic-data-generation` skill.
Discussion
No questions or reviews yet.
Related
PluginsThe Dex plugin: an agent explores a database, builds dbt models and a semantic layer and tracks schema drift under cost control
A plugin with two skills over a semantic metrics layer: the agent writes models in YAML and queries them through SQL.
A Python library for typed, versioned datasets over S3, GCS, and Azure, with a datachain skill install command that teaches Claude Code, Cursor, and Codex your
DataHub skills for searching the catalog, enriching metadata, tracing lineage, managing data quality and building connectors
A template for data scientists: AGENTS.md, CLAUDE.md and skills for Python, SQL, Polars and model diagnostics.
A plugin of four skills that guides setup, GPU memory estimation and fine-tuning runs of LLMs through the Training Hub interface.
README in short
The README opens with the framework description, a Python example and links to the docs. A section on the plugin for two agents with install commands and a table of three skills follows. The rest holds links to installation, quick start, concepts, block and flow references and the contributor guide.