17-08-2024

Blueprints simplify agent-based automation on Amazon Bedrock

Amazon Bedrock Customers can now leverage Blueprints to jumpstart their use of Agents. Blueprints are a collection of pre-built templates optimized for popular customer use cases. These templates allow Amazon Bedrock customers to quickly experiment with agent-based applications without the need for complex configuration or lengthy development cycles.? Agent Blueprints are open-source templates hosted on the AWS QuickStart GitHub repository. The pre-built templates come with sample actions, and Knowledge Bases, making it easy for customers to experiment and learn more about the capabilities of agents. Furthermore, the templates include customized prompts tailored to common use cases, saving customers valuable time and effort on prompt engineering to optimize their automation solutions.? Blueprints are free to use, and customers only pay for the resources they create, such as the standard InvokeModel charges, when they deploy the Blueprint.

SageMaker Canvas unlocks no-code ML and data preparation at petabyte-scale

Amazon SageMaker Canvas now empowers enterprises to harness the full potential of their data by enabling support of petabyte-scale datasets. Starting today, you can interactively prepare large datasets, create end-to-end data flows and trigger AutoML experiments on petabytes – a substantial leap from the previous 5GB limit. With 50+ connectors, intuitive "chat with data" interface, and petabyte support, Canvas provides a scalable, low-code/no-code ML solution for handling real-world, enterprise use cases.? Starting today, Canvas empowers you with new sampling techniques like random and stratified, allowing samples up to 200K rows – a tenfold increase. This makes it easy to gather data quality insights and understand the impact of your data transformations interactively before processing your entire dataset, leveraging our new seamless integration with EMR Serverless. Canvas automatically scales processing over 5GB data across sampling, preparation, model building and inference to EMR Serverless, unlocking your data's full predictive potential through an intuitive experience. EMR Serverless usage incurs additional EMR pricing costs.? The new petabyte support and improved interactive experience is available across all AWS Regions where SageMaker Canvas is offered.? To get started with no-code ML and data preparation of large datasets, enable "large data processing configuration" in your Canvas domain and user profile using our technical documentation, and learn how to use the new capability from the AWS Machine Learning blog. Existing users should update their SageMaker domain settings per the documentation, log out from the Canvas workspace, and log back in to access the latest version.?

Amazon Neptune now introduces support for PropertyGraphStore in Neptune to build more reliable GraphRAG applications

Starting today, you can build Graph Retrieval-Augmented Generation (GraphRAG) applications by enabling PropertyGraphIndex and combining knowledge graphs stored in Amazon Neptune with LlamaIndex, a popular open-source framework for building applications with Large Language Models (LLMs) such as those available in Amazon Bedrock. We are excited to introduce the capability to add natural language querying via the TextToCypher Retriever, knowledge graph retrieval via the Cypher Template Retriever and Knowledge Graph Enhanced RAG creation and querying via the supported extractors and retrievers.? Customers building Generative AI applications often use Retrieval-Augmented Generation (RAG) to ensure LLM output is relevant, accurate, and useful. While RAG enhances LLM capabilities by integrating specific domain knowledge without retraining the model, RAG applications may still face significant challenges when relevant information is dispersed across multiple sources or documents. Knowledge graphs consolidate and integrate an organization’s information, enabling GraphRAG to relate concepts and entities across the content. PropertyGraphIndex in GraphRAG applications allows efficient indexing and querying of node and relationship properties in knowledge graphs, enabling quick retrieval of relevant data based on specific attributes. With this launch, you can now effortlessly convert text into openCypher queries, making it easier to interact with and extract insights from your knowledge graphs. Additionally, you can utilize pre-defined templates for common openCypher queries, streamlining the query-building process and ensuring consistency across applications. Whether you are handling complex multi-hop retrievals or simple queries, PropertyGraphIndex significantly enhances the overall performance and capability of your GraphRAG solutions.