In the realm of generative AI (GenAI), achieving ultrahigh relevance and producing high-quality results is the key to gaining a competitive advantage. It’s not just about having more data; it's about leveraging a combination of publicly available data, your proprietary data, and the unique insights exclusive to your organization. By harnessing this amalgamation, you can make learning and results more relevant, establishing an ever-expanding intelligent foundation.
Whether organizations choose to build using DIY tools or partner with system integrators, the integration of complex technologies can be simplified with a well-compiled toolkit. This enables organizations to extract greater value from their data and derive meaningful insights.
Your opportunity
NetApp provides an opportunity: AI-ready storage that enables retrieval-augmented generation (RAG) operations. By securely augmenting public data with your proprietary data, NetApp storage increases relevance and allows pinpoint accuracy for your large language model (LLM) workloads. Additionally, it lets you rapidly create intelligent, containerized modern applications.
The solution: NetApp GenAI Toolkit
The NetApp® GenAI Toolkit empowers you to take advantage of your private data stored on Google Cloud NetApp Volumes by using the foundation models provided by Google Cloud’s Vertex AI machine-learning platform. The GenAI Toolkit comes with a chatbot app that’s hosted on a NetApp artifact registry and a Terraform module (in a GitHub repository) that automates deployment of the chatbot app in your environment.
As depicted in the following figure:
- You provision a Google Cloud NetApp Volume instance.
- You configure API access and transfer documents of interest to a NetApp volume.
- You download the GenAI Toolkit Terraform module from GitHub and apply it.
- The Terraform module creates a Linux VM, launches the RAG framework, and connects to the NetApp volumes provided in the configuration.
- You interact with the chatbot endpoint to get answers grounded in your proprietary data.
The toolkit, along with the accompanying reference architecture, allows you to implement RAG operations more quickly while enabling secure, consistent, and automated workflows that connect data stored in Google Cloud NetApp Volumes with Vertex AI. The result is an enhanced ability to generate unique, high-quality, and ultrarelevant competitive insights.
Industry-leading capabilities
The NetApp GenAI Toolkit helps optimize RAG processes with industry-leading capabilities, including:
- Common data footprint everywhere. You can easily include data from any environment to power your RAG efforts. NetApp ONTAP® data management software lets you use common operational processes while reducing risk, cost, and time to results.
- Automated classification. The NetApp BlueXP™ classification service streamlines data categorization, classification, and cleansing for both the ingest and inferencing phases of the data pipeline. With this approach, the right data is used for queries, and sensitive data is protected according to your organization’s policy.
-
Fast, scalable Snapshot copies. NetApp Snapshot™ technology creates near-instant space-efficient, in-place copies of vector stores and databases for interval-based A/B testing and recovery. You can perform point-in-time analysis or, if data is inconsistent, immediately roll back to a previous version.
- Real-time cloning at scale. NetApp FlexClone® technology can create instant clones of vector index stores for parallel processing of A/B prompt testing and result validation. With cloning, you can safely make uniquely relevant data instantly available for queries from different users, without affecting the core production.
The GenAI Toolkit for Vertex AI is available for preview in May 2024, with general availability planned for the second half of 2024. We’re also planning support and updates for additional hyperscalers.
The integration of the GenAI Toolkit with Vertex AI represents a powerful synergy that empowers you to harness advanced language generation capabilities while maintaining data privacy and security. To learn more, read the press release.