Repolode
Submit
NVIDIA-NeMo

NVIDIA-NeMo/DataDesigner

A Python framework from NVIDIA for generating synthetic datasets from scratch or from seed data, using samplers, LLMs, image models and validators.

View on GitHubWebsite
Stars
2,401
Forks
245
Open issues
67
Last commit
19 hours ago

What is it?

NeMo Data Designer is a framework for building synthetic datasets that go beyond simple prompting of a language model. Columns can be sampled from statistical distributions, generated by LLMs or image models, or drawn from existing seed datasets, and fields can be made to depend on one another so that relationships between them stay consistent.

Generated outputs can be checked with Python and SQL validators, custom validators, or LLM judges. The framework supports multimodal context with image, audio and video, and can connect generation to local or remote MCP servers while recording interaction traces. Users can preview a sample record before running a large job, and can resume and monitor long generation runs.

Models are configured through providers such as NVIDIA Build, OpenAI and OpenRouter, and a CLI handles provider and model setup. Plugin packages, maintained in a separate repository, add capabilities like GitHub seed readers and a retrieval question-answer generation toolkit. Python 3.10 to 3.14 is supported.

Key features

  • 01Generates text, structured data and images with samplers, LLMs, image models or seed datasets.
  • 02Controls relationships between fields through dependency-aware generation.
  • 03Validates outputs with Python, SQL, custom validators and LLM judges.
  • 04Connects generation to local or remote MCP servers and captures traces.
  • 05Supports plugins for additional data sources and processing steps.

Use cases

  • Creating synthetic customer reviews or product catalogs with controlled attributes for testing or training.
  • Generating question-answer pairs from documents to train retrievers.
  • Producing validated tool-use or agent interaction datasets.

Quick start

  1. 1

    Install Data Designer from PyPI

    pip install data-designer
  2. 2

    Set an API key for a default model provider

    export NVIDIA_API_KEY="your-api-key-here"
  3. 3

    Run the README Python example to preview generated data

Things to know

  • Generation needs an API key from a model provider such as NVIDIA Build, OpenAI, or OpenRouter.
  • Telemetry is collected by default and can be disabled with NEMO_TELEMETRY_ENABLED=false.
  • NVIDIA Build is intended for evaluation and testing only, not production use.

In collections

Open-source AI agent frameworks & appsMCP servers

Topics

#synthetic-data#data-generation#llm#multimodal#data-augmentation#tool-use#agentic-ai#mcp#nemo#nvidia

FAQ

What is DataDesigner?
A Python framework from NVIDIA for generating synthetic datasets from scratch or from seed data, using samplers, LLMs, image models and validators. NeMo Data Designer is a framework for building synthetic datasets that go beyond simple prompting of a language model. Columns can be sampled from statistical distributions, generated by LLMs or image models, or drawn from existing seed datasets, and fields can be made to depend on one another so that relationships between them stay consistent.
Is DataDesigner free to use? What license does it have?
DataDesigner is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is DataDesigner still maintained?
Actively maintained. The last commit to DataDesigner was on Nov 7, 2026.
What can I use DataDesigner for?
Creating synthetic customer reviews or product catalogs with controlled attributes for testing or training. Generating question-answer pairs from documents to train retrievers. Producing validated tool-use or agent interaction datasets.
How do I install DataDesigner?
Install Data Designer from PyPI: pip install data-designer Set an API key for a default model provider: export NVIDIA_API_KEY="your-api-key-here" Run the README Python example to preview generated data
Where can I get DataDesigner?
The source code and installation instructions are on GitHub: https://github.com/NVIDIA-NeMo/DataDesigner

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects