Build a Serverless Web Application using Generative AI

TUTORIAL

Task 4: Deploy the Backend API

In this task, you will use AWS Amplify to configure a custom data query to use Amazon Bedrock as data source.

Overview

In this task, you will configure a custom query that will reference the data source and the resolver you defined the previous model to produce a recipe based on a list of ingredients. This query will use a custom type to structure the response from Amazon Bedrock. 

What you will accomplish

  • Define a GraphQL Query that takes an array of strings
  • Define a custom type to be used to structure the response from the query

 

Implementation

 Time to complete

5 minutes

  • 1. On your local machine, navigate to the ai-recipe-generator/amplify/data/resource.ts file, and update it with the following code. Then, save the file.

    • The following code defines the askBedrock query that takes an array of strings called ingredients and returns a BedrockResponse. The .handler(a.handler.custom({ entry: "./bedrock.js", dataSource: "bedrockDS" })) line sets up a custom handler for this query, defined in bedrock.js, using bedrockDS as its data source.

     

    import { type ClientSchema, a, defineData } from "@aws-amplify/backend";
    
    const schema = a.schema({
      BedrockResponse: a.customType({
        body: a.string(),
        error: a.string(),
      }),
    
      askBedrock: a
        .query()
        .arguments({ ingredients: a.string().array() })
        .returns(a.ref("BedrockResponse"))
        .authorization((allow) => [allow.authenticated()])
        .handler(
          a.handler.custom({ entry: "./bedrock.js", dataSource: "bedrockDS" })
        ),
    });
    
    export type Schema = ClientSchema<typeof schema>;
    
    export const data = defineData({
      schema,
      authorizationModes: {
        defaultAuthorizationMode: "apiKey",
        apiKeyAuthorizationMode: {
          expiresInDays: 30,
        },
      },
    });

    2. Open a new terminal window, navigate to your apps project folder (ai-recipe-generator), and run the following command to deploy cloud resources into an isolated development space so you can iterate fast.

    npx ampx sandbox

    3. Once the cloud sandbox has been fully deployed, your terminal will display a confirmation message and the amplify_outputs.json file will be generated and added to your project. 

Conclusion

You have configured a GraphQL API to define a custom query to connect to Amazon Bedrock and generate recipes based on a list of ingredients. 

Build the Frontend

Was this page helpful?