Holohub Overview
Holohub is the community collection of reference applications, reusable operators, tutorials, and benchmarks built on top of the Holoscan SDK. It complements the SDK by showing production-style patterns for real sensor pipelines, AI inference, visualization, and multi-operator graph design. The Holohub documentation website is the best place to browse applications, operators, tutorials, and benchmarks before running locally.
Holohub is an open, community-driven project and is not an officially supported NVIDIA product. It is not covered by the NVAIE (NVIDIA AI Enterprise) maintenance plan, so official support channels may not address Holohub-specific issues.
The Holohub repository is hosted on the NVIDIA GitHub organization and welcomes community participation. If you discover a bug, have a feature request, or want to suggest improvements, please open an issue on the Holohub issue tracker.
Interested in contributing code, documentation, or examples? Read the Contributing to Holohub guide for details on submitting pull requests, code style, and becoming an active contributor to the Holohub ecosystem.
Reference applications that demonstrate complete end-to-end pipelines.
Reusable operators that can be imported into your own projects.
Tutorials and examples that focus on specific developer workflows.
Benchmarks for evaluating performance and deployment configurations.
Command-line interface for building, running, and managing applications and operators.
The Holoscan SDK provides the core runtime, scheduling, resources, and built-in operators. Holohub builds on those SDK capabilities with higher-level, domain-oriented examples and components.
In practice:
Use the SDK docs to understand core APIs, runtime behavior, and operator development.
Use Holohub to discover working patterns you can run, inspect, and adapt.
Holohub separates reusable building blocks from complete solutions:
Operator: a modular processing component (for example, a pre-processing or post-processing stage) designed to be reused across graphs.
Application: a full Holoscan graph composed of operators, conditions, and resources that can be run as a complete pipeline.
This separation helps teams prototype quickly with existing operators, then package complete reference applications for deployment and benchmarking.
The Holohub CLI provides a unified interface for building, running, and managing applications and operators.
For most users, a single command is enough:
./holohub run <application_name>
This command orchestrates the end-to-end workflow by preparing and launching the required Docker environment, building the selected application, and running it with the application’s preconfigured arguments and parameters. For more details, see the Holohub CLI Guide.
The holohub repository is organized around discoverable components:
applications/: runnable reference applications.operators/: reusable custom operators.benchmarks/: performance and scaling evaluations.tutorials/: focused how-to examples.
Most applications and operators include a metadata.json file.
These metadata files describe tags, dependencies, supported platforms, and compatibility details, and are used by Holohub tooling to discover and manage components.
Holohub is especially useful when you want to:
Start from a working reference rather than building from scratch.
Learn best practices for composing operators into robust graphs.
Compare multiple implementation patterns for similar tasks.
Evaluate performance with realistic workloads.
Contribute to the community by sharing your own operators, applications, workflows, and tools.