AWS AppSync, a fully managed API management service that connects applications to events, data, and AI models. Today, customers use AppSync as an AI gateway to trigger generative AI workflows and use subscriptions, powered by WebSockets, to return progressive updates from long-running invocations. This allows them to implement asynchronous patterns. However, in some cases, customers need to make short synchronous invocations to their models. AWS AppSync now supports Amazon Bedrock runtime as a data source for GraphQL APIs, enabling seamless integration of generative AI capabilities. This new feature allows developers to make short synchronous invocations (10 seconds or less) to foundation models and inference profiles in Amazon Bedrock directly from their AppSync GraphQL APIs.
The integration supports calling the converse and invokeModel APIs. Developers can interact with Anthropic models like Claude 3.5 Haiku and Claude 3.5 Sonnet for data analysis and structured object generation tasks. They can also use Amazon Titan models to generate embeddings, create summaries, or extract action items from meeting minutes.
For longer-running invocations, customers can continue using AWS Lambda functions in event mode to interact with Bedrock models and send progressive updates to clients via subscriptions.
This new data source is available in all AWS Regions where AWS AppSync is available. To get started, customers can visit the AWS AppSync console and refer to the AWS AppSync documentation for more information.
Source:: Amazon AWS