Post

First post on BlueSky. Let's go! I'm really excited about our 2 latest feature releases for AppSync 🎉 First, AppSync now supports Amazon Bedrock runtime as a data source for GraphQL APIs. Second, AppSync now supports resource sharing of AppSync GraphQL APIs across accounts. Let's dig into it. 1/7

Customers use AppSync today to for generative AI workloads powered by Bedrock. Often, developers use an async pattern that calls a Lambda function to invoke a model, and then returns results to clients via a subscription. Great for long-running invocations and getting progressive updates. 2/7

In some cases, customers just want to make short synchronous model invocations directly. Think about getting an embedding or summarization from a Amazon Titan model or generating an object using a Claude 3.5 Haiku model. 3/7

With this new release we simplify config, monitoring and observability by introducing an Amazon Bedrock runtime data source that you can use to call the `invokeModel` and `converse` APIs. You can easily call these APIs from your JavaScript resolvers. 4/7

Next, let's talk about resource sharing of AppSync GraphQL APIs. Our recent launch of data plane logging to AWS CloudTrail might have been a hint this was coming. A lot of customers are using AppSync's private API feature today to enable private workloads inside their VPCs. 5/7

Before today, customers had to set up additional networking infrastructure to share their private GraphQL APIs between their organization accounts. Now, customers can centralize their GraphQL API management in a dedicated account and share access to these APIs with other accounts. 6/7