############################## HOOPS AI 1.0 Release Note ############################## What’s New in HOOPS AI 1.0 ========================== Compared to the beta versions, HOOPS AI 1.0 introduces key improvements that make the framework more robust, scalable, and ready for broader adoption: * **Linux Support** - HOOPS AI now runs on Linux environments, enabling integration into a wider range of development and production workflows. * **Improved Memory Management** - Significant improvements in memory handling allow more efficient processing of large CAD datasets and more stable execution of machine learning pipelines. ---- Overview ======== HOOPS AI 1.0 introduces a new SDK designed to bring machine learning capabilities to 3D CAD data. It enables developers and data scientists to build AI-driven workflows using engineering geometry while integrating with the HOOPS ecosystem. For a detailed introduction, see the :doc:`/getting_started/technical_overview`. This first release provides: * A framework to build machine learning pipelines on CAD data * A set of pre-built models and workflows to accelerate common engineering use cases ---- HOOPS AI Framework ================== The HOOPS AI framework provides the infrastructure needed to develop machine learning workflows using CAD models. It enables developers to access CAD geometry, structure datasets, and build machine learning pipelines without implementing low-level geometry processing. CAD Data Access and Encoding ----------------------------- HOOPS AI provides structured access to geometric and topological CAD information, allowing developers to convert engineering models into representations suitable for machine learning workflows. This allows ML models to operate directly on engineering geometry. See the :doc:`/programming_guide/cad-data-access` and :doc:`/programming_guide/cad-data-encoding` guides, or try the :doc:`/tutorials/data_access` and :doc:`/tutorials/data_encode` tutorials. ---- Dataset Creation and Management -------------------------------- The framework includes tools to: * Ingest CAD models * Create structured datasets * Encode geometric features * Prepare datasets for training and evaluation This simplifies the preparation of engineering datasets for machine learning experiments. See the :doc:`/programming_guide/datasets` guide and the :doc:`/tutorials/data_management` tutorial. ---- Machine Learning Pipeline Integration --------------------------------------- HOOPS AI integrates with common Python-based machine learning environments and enables developers to build workflows including: * Dataset preparation * Model training * Evaluation and experimentation The framework works with standard ML tools while providing CAD-specific capabilities. See the :doc:`/programming_guide/data-flow-management` and :doc:`/programming_guide/ml-model` guides. ---- Visualization and Exploration ------------------------------ HOOPS AI includes visualization capabilities that allow developers to: * Inspect datasets * Explore model outputs * Validate machine learning results on 3D models Interactive visualization helps developers better understand how ML models interpret engineering geometry. See the :doc:`/programming_guide/cad-data-visualization` and :doc:`/programming_guide/explore-dataset` guides. ---- Pre-Built Models and Workflows ================================ HOOPS AI includes pre-built models and example workflows that demonstrate how machine learning can be applied to CAD data. Examples include: * :doc:`Part classification ` * :doc:`Shape similarity search ` * :doc:`Embedding generation ` for CAD models * :doc:`Feature extraction ` workflows These models can be used directly for experimentation or extended to build custom applications. ---- Example Use Cases and Workflows ================================= HOOPS AI enables developers to build machine learning workflows directly on CAD datasets. Typical workflow: #. Import CAD models and build a dataset #. Encode geometric information for ML processing #. Train or apply models to analyze data #. Visualize and validate results on 3D models Common use cases include: * Part classification and catalog organization * Shape similarity search and duplicate detection * CAD embedding generation for dataset exploration * Geometry-driven analytics and feature extraction Example applications across engineering platforms: * **PLM systems** – automated part classification, duplicate detection, and design reuse across product catalogs * **MES platforms** – identification and grouping of components to improve production planning and traceability * **CAM applications** – geometry-based analysis to support machining strategy selection or feature recognition * **Manufacturing / MaaS platforms** – automatic matching of CAD models with suitable manufacturing processes or suppliers These workflows enable software developers to integrate AI-driven engineering intelligence directly into CAD-centric applications. See the :doc:`/tutorials/index` for hands-on examples of these workflows. ---- Developer Experience ===================== Python-based SDK ----------------- HOOPS AI is delivered as a Python framework, enabling data scientists and developers to build AI workflows using familiar machine learning tools and libraries. See the full :doc:`/api` for details on available modules and classes. ---- Simplified Installation ----------------------- HOOPS AI provides an installer that automates environment setup and dependency installation. The installer: * Checks prerequisites * Creates a Conda environment * Installs ML dependencies (including PyTorch) * Installs HOOPS AI components * Registers a Jupyter environment for experimentation GPU acceleration is supported when CUDA is available. See :doc:`/getting_started/evaluate` for full installation instructions. ---- Evaluation =========== HOOPS AI can be evaluated through a 30-day evaluation license. Evaluation packages include: * HOOPS AI framework * Example datasets * Tutorial notebooks These resources allow users to quickly explore machine learning workflows on CAD data. See :doc:`/getting_started/evaluate` for details on how to get started. ---- Tutorials and Learning Resources ================================== The release includes :doc:`tutorial notebooks ` demonstrating: * :doc:`Dataset creation ` * :doc:`Model training workflows ` * :doc:`Similarity search using embeddings ` * :doc:`Applying machine learning to CAD datasets ` These tutorials are available through the HOOPS AI documentation. ---- Supported Platforms ==================== HOOPS AI runs in a Conda-based Python environment and supports: * Windows * Linux GPU acceleration is available using CUDA. See :doc:`/getting_started/supported_platforms` for full platform details. ---- Who Should Use HOOPS AI ========================= HOOPS AI is designed for: * Engineering software developers * Data scientists working with CAD data * Teams building AI-powered engineering applications * Organizations exploring ML workflows for design and manufacturing data ---- Getting Started ================ To begin evaluating HOOPS AI: #. Request access through the HOOPS AI product page #. :doc:`Install the HOOPS AI package ` using the provided installer #. Launch JupyterLab and explore the :doc:`tutorial notebooks ` #. Start building machine learning workflows on CAD data