You can now add a data source from a generic SMB or NFS share. This enables you to include files that are stored on volumes hosted by filesystems other than Amazon FSx for NetApp ONTAP.

Add data sources to a knowledge base

Add data sources to a connector

The Tracker monitoring capability is now available in GenAI. You can use Tracker to monitor and track the progress and status of pending, ongoing, and completed operations, review details for operation tasks and subtasks, diagnose any issues or failures, edit parameters for failed operations, and retry failed operations.

Monitor workload operations with Tracker in BlueXP workload factory

You can now increase the relevance of reranked query results by selecting a specific reranker model to use with a knowledge base. GenAI supports the Cohere Rerank and Amazon Rerank models.

Create a GenAI knowledge base

This release of GenAI introduces support for NetApp Connector for Amazon Q Business, enabling you to create connectors for Amazon Q Business. Quickly and easily take advantage of the Amazon Q Business AI assistant with less initial configuration than building a GenAI knowledge base for Amazon Bedrock.

Create a NetApp Connector for Amazon Q Business

GenAI now supports the following additional chat models for knowledge bases:

GenAI supports the models from each provider that Amazon Bedrock supports: Supported foundation models in Amazon Bedrock

Create a GenAI knowledge base

The workload factory user interface and documentation now use "read-only" to refer to read permissions and "read/write" to refer to automate permissions. == 02 March 2025

You can now copy questions and responses directly to the clipboard, adjust the size of the chat window, and change its title. Additionally, chat responses can now include tables, which are also copyable.

Test a GenAI knowledge base

Chat responses now include citations that list the files and chunks of data that were used to generate the response.

Test a GenAI knowledge base

This release of GenAI provides enhanced file support:

Supported data source file formats

GenAI now supports the Amazon Nova foundation models. Amazon Nova Micro, Amazon Nova Lite, and Amazon Nova Pro are supported.

GenAI requirements

GenAI now supports selecting specific file types to include in the data source scan when you add a data source.

Add data sources to the knowledge base

GenAI now supports filtering files to include in the data source scan by modification date when you add a data source. You can choose a modification date range for the included files.

Add data sources to the knowledge base

GenAI now supports enhancing responses to knowledge base queries with insights from images and graph descriptions, as well as document text, leading to richer and higher quality answers. GenAI can now scan image files and images within PDF files (also known as multi-modal file support). If you choose to scan images or PDF files, the text from the images (including images embedded in PDF documents) is scanned into the data source and insights from the scans are included in the responses to knowledge base queries.

Add data sources to the knowledge base

GenAI can now significantly enhances the relevance and accuracy of search results by using hybrid search and re-ranking the results. Hybrid search combines the strengths of traditional keyword-based search with advanced dense vector-based semantic search techniques. The standard keyword search results are augmented with close matches and linguistic nuance, enhancing relevance. GenAI then refines these results further by using advanced re-ranking models, such as Cohere Rerank and Amazon Rerank, and returns the most relevant results. This capability is available for newly created knowledge bases.

Learn about BlueXP workload factory for GenAI

You can now provide a snapshot name for an ad-hoc snapshot.

Protect a knowledge base with snapshots

You can now give a custom name to the AI engine instance during deployment.

Deploy the GenAI infrastructure

If your AI engine instance becomes corrupted or is somehow deleted, you can let workload factory rebuild it for you. Workload factory automatically reattaches your knowledge bases to the infrastructure after rebuilding is complete, so that they are ready to use.

Troubleshooting

BlueXP workload factory for GenAI now supports cloning a knowledge base from a snapshot. This enables quick recovery of knowledge bases and creation of new knowledge bases with existing data sources, and helps with data recovery and development.

Clone a knowledge base

Discover and replicate on-premises ONTAP cluster data to an FSx for ONTAP file system so that it can be used to enrich AI knowledge bases. All on-premises discovery and replication workflows are possible from the new On-Premises ONTAP tab in the Storage inventory.

Discover an on-premises ONTAP cluster

The Generative AI workload introduces the data guardrails feature, powered by BlueXP classification. The data guardrails feature identifies and masks Personal Identifiable Information (PII) helping you maintain compliance and strengthen security for your sensitive organizational data.

Create a GenAI knowledge base

Learn about BlueXP classification

You can now protect your Generative AI workloads data by taking a point-in-time copy of a knowledge base. This enables you to protect your data against accidental loss or test changes to the settings of the knowledge base. You can restore the previous version of the knowledge base volume at any time.

Take a snapshot of a knowledge base volume

Restore a snapshot of a knowledge base volume

You can now pause scheduled data source scans. By default, Generative AI workloads scans each data source daily to ingest new data into each knowledge base. If you don't want the latest changes to be ingested (during testing or while restoring a snapshot, for example) you can pause the scheduled scans and resume them at any time.

Manage knowledge bases

When selecting a knowledge base volume, you can now choose a data protection volume that is part of a NetApp SnapMirror replication relationship. This enables you to store knowledge bases on volumes that are already protected by SnapMirror replication.

Identify the data sources to integrate in your knowledge base

Generative AI workloads now supports multi-sentence chunking and overlap-based chunking for data sources.

Generative AI workloads now creates a dedicated Amazon FSx for NetApp ONTAP volume for each new knowledge base, enabling individual snapshot policies for each knowledge base and improved protection against failures and data poisoning.

Generative AI workloads is now integrated with Amazon CloudWatch Logs, enabling you to monitor Generative AI workloads log files.

The NetApp workload factory GenAI sample application enables you to test authentication and retrieval from your published NetApp workload factory knowledge base by interacting directly with it in a web-based chatbot application.

The initial release includes the capability to develop a knowledge base that is customized by embedding your organization's data. The knowledge base can be accessed by a chatbot application for your users. This capability ensures accurate and relevant responses to organization-specific questions, enhancing the satisfaction and productivity for all of your users.