Invoque Amazon Nova en Amazon Bedrock mediante la API Converse de Bedrock con un flujo de respuesta

Los siguientes ejemplos de código muestran cómo enviar un mensaje de texto a Amazon Nova mediante la API Converse de Bedrock y procesar el flujo de respuestas en tiempo real.

.NET

SDK para .NET

nota

Hay más información al respecto. GitHub Busque el ejemplo completo y aprenda a configurar y ejecutar en el Repositorio de ejemplos de código de AWS.

Envíe un mensaje de texto a Amazon Nova mediante la API Converse de Bedrock y procese el flujo de respuestas en tiempo real.

// Use the Converse API to send a text message to Amazon Nova
// and print the response stream.

using System;
using System.Collections.Generic;
using System.Linq;
using Amazon;
using Amazon.BedrockRuntime;
using Amazon.BedrockRuntime.Model;

// Create a Bedrock Runtime client in the AWS Region you want to use.
var client = new AmazonBedrockRuntimeClient(RegionEndpoint.USEast1);

// Set the model ID, e.g., Amazon Nova Lite.
var modelId = "amazon.nova-lite-v1:0";

// Define the user message.
var userMessage = "Describe the purpose of a 'hello world' program in one line.";

// Create a request with the model ID, the user message, and an inference configuration.
var request = new ConverseStreamRequest
{
    ModelId = modelId,
    Messages = new List<Message>
    {
        new Message
        {
            Role = ConversationRole.User,
            Content = new List<ContentBlock> { new ContentBlock { Text = userMessage } }
        }
    },
    InferenceConfig = new InferenceConfiguration()
    {
        MaxTokens = 512,
        Temperature = 0.5F,
        TopP = 0.9F
    }
};

try
{
    // Send the request to the Bedrock Runtime and wait for the result.
    var response = await client.ConverseStreamAsync(request);

    // Extract and print the streamed response text in real-time.
    foreach (var chunk in response.Stream.AsEnumerable())
    {
        if (chunk is ContentBlockDeltaEvent)
        {
            Console.Write((chunk as ContentBlockDeltaEvent).Delta.Text);
        }
    }
}
catch (AmazonBedrockRuntimeException e)
{
    Console.WriteLine($"ERROR: Can't invoke '{modelId}'. Reason: {e.Message}");
    throw;
}

• Para obtener más información sobre la API, consulte la Referencia de ConverseStreamla AWS SDK para .NET API.

Java

SDK para Java 2.x

nota

Hay más información al respecto GitHub. Busque el ejemplo completo y aprenda a configurar y ejecutar en el Repositorio de ejemplos de código de AWS.

Envíe un mensaje de texto a Amazon Nova mediante la API Converse de Bedrock y procese el flujo de respuestas en tiempo real.

import software.amazon.awssdk.auth.credentials.DefaultCredentialsProvider;
import software.amazon.awssdk.regions.Region;
import software.amazon.awssdk.services.bedrockruntime.BedrockRuntimeAsyncClient;
import software.amazon.awssdk.services.bedrockruntime.model.*;

import java.util.concurrent.ExecutionException;

/**
 * This example demonstrates how to use the Amazon Nova foundation models with an
 * asynchronous Amazon Bedrock runtime client to generate streaming text responses.
 * It shows how to:
 * - Set up the Amazon Bedrock runtime client
 * - Create a message
 * - Configure a streaming request
 * - Set up a stream handler to process the response chunks
 * - Process the streaming response
 */
public class ConverseStream {

    public static void converseStream() {

        // Step 1: Create the Amazon Bedrock runtime client
        // The runtime client handles the communication with AI models on Amazon Bedrock
        BedrockRuntimeAsyncClient client = BedrockRuntimeAsyncClient.builder()
                .credentialsProvider(DefaultCredentialsProvider.create())
                .region(Region.US_EAST_1)
                .build();

        // Step 2: Specify which model to use
        // Available Amazon Nova models and their characteristics:
        // - Amazon Nova Micro: Text-only model optimized for lowest latency and cost
        // - Amazon Nova Lite:  Fast, low-cost multimodal model for image, video, and text
        // - Amazon Nova Pro:   Advanced multimodal model balancing accuracy, speed, and cost
        //
        // For the latest available models, see:
        // https://docs.aws.amazon.com/bedrock/latest/userguide/models-supported.html
        String modelId = "amazon.nova-lite-v1:0";

        // Step 3: Create the message
        // The message includes the text prompt and specifies that it comes from the user
        var inputText = "Describe the purpose of a 'hello world' program in one paragraph";
        var message = Message.builder()
                .content(ContentBlock.fromText(inputText))
                .role(ConversationRole.USER)
                .build();

        // Step 4: Configure the request
        // Optional parameters to control the model's response:
        // - maxTokens: maximum number of tokens to generate
        // - temperature: randomness (max: 1.0, default: 0.7)
        //   OR
        // - topP: diversity of word choice (max: 1.0, default: 0.9)
        // Note: Use either temperature OR topP, but not both
        ConverseStreamRequest request = ConverseStreamRequest.builder()
                .modelId(modelId)
                .messages(message)
                .inferenceConfig(config -> config
                                .maxTokens(500)     // The maximum response length
                                .temperature(0.5F)  // Using temperature for randomness control
                        //.topP(0.9F)       // Alternative: use topP instead of temperature
                ).build();

        // Step 5: Set up the stream handler
        // The stream handler processes chunks of the response as they arrive
        // - onContentBlockDelta: Processes each text chunk
        // - onError: Handles any errors during streaming
        var streamHandler = ConverseStreamResponseHandler.builder()
                .subscriber(ConverseStreamResponseHandler.Visitor.builder()
                        .onContentBlockDelta(chunk -> {
                            System.out.print(chunk.delta().text());
                            System.out.flush();  // Ensure immediate output of each chunk
                        }).build())
                .onError(err -> System.err.printf("Can't invoke '%s': %s", modelId, err.getMessage()))
                .build();

        // Step 6: Send the streaming request and process the response
        // - Send the request to the model
        // - Attach the handler to process response chunks as they arrive
        // - Handle any errors during streaming
        try {
            client.converseStream(request, streamHandler).get();

        } catch (ExecutionException | InterruptedException e) {
            System.err.printf("Can't invoke '%s': %s", modelId, e.getCause().getMessage());
        }
    }

    public static void main(String[] args) {
        converseStream();
    }
}

• Para obtener más información sobre la API, consulte la referencia de ConverseStreamla AWS SDK for Java 2.x API.

JavaScript

SDK para JavaScript (v3)

nota

Hay más información. GitHub Busque el ejemplo completo y aprenda a configurar y ejecutar en el Repositorio de ejemplos de código de AWS.

Envíe un mensaje de texto a Amazon Nova mediante la API Converse de Bedrock y procese el flujo de respuestas en tiempo real.

// This example demonstrates how to use the Amazon Nova foundation models
// to generate streaming text responses.
// It shows how to:
// - Set up the Amazon Bedrock runtime client
// - Create a message
// - Configure a streaming request
// - Process the streaming response

import {
  BedrockRuntimeClient,
  ConversationRole,
  ConverseStreamCommand,
} from "@aws-sdk/client-bedrock-runtime";

// Step 1: Create the Amazon Bedrock runtime client
// Credentials will be automatically loaded from the environment
const client = new BedrockRuntimeClient({ region: "us-east-1" });

// Step 2: Specify which model to use
// Available Amazon Nova models and their characteristics:
// - Amazon Nova Micro: Text-only model optimized for lowest latency and cost
// - Amazon Nova Lite:  Fast, low-cost multimodal model for image, video, and text
// - Amazon Nova Pro:   Advanced multimodal model balancing accuracy, speed, and cost
//
// For the most current model IDs, see:
// https://docs.aws.amazon.com/bedrock/latest/userguide/models-supported.html
const modelId = "amazon.nova-lite-v1:0";

// Step 3: Create the message
// The message includes the text prompt and specifies that it comes from the user
const inputText =
  "Describe the purpose of a 'hello world' program in one paragraph";
const message = {
  content: [{ text: inputText }],
  role: ConversationRole.USER,
};

// Step 4: Configure the streaming request
// Optional parameters to control the model's response:
// - maxTokens: maximum number of tokens to generate
// - temperature: randomness (max: 1.0, default: 0.7)
//   OR
// - topP: diversity of word choice (max: 1.0, default: 0.9)
// Note: Use either temperature OR topP, but not both
const request = {
  modelId,
  messages: [message],
  inferenceConfig: {
    maxTokens: 500, // The maximum response length
    temperature: 0.5, // Using temperature for randomness control
    //topP: 0.9,        // Alternative: use topP instead of temperature
  },
};

// Step 5: Send and process the streaming request
// - Send the request to the model
// - Process each chunk of the streaming response
try {
  const response = await client.send(new ConverseStreamCommand(request));

  for await (const chunk of response.stream) {
    if (chunk.contentBlockDelta) {
      // Print each text chunk as it arrives
      process.stdout.write(chunk.contentBlockDelta.delta?.text || "");
    }
  }
} catch (error) {
  console.error(`ERROR: Can't invoke '${modelId}'. Reason: ${error.message}`);
  process.exitCode = 1;
}

• Para obtener más información sobre la API, consulte la referencia de ConverseStreamla AWS SDK para JavaScript API.

Kotlin

SDK para Kotlin

nota

Hay más información al respecto GitHub. Busque el ejemplo completo y aprenda a configurar y ejecutar en el Repositorio de ejemplos de código de AWS.

Envíe un mensaje de texto a Amazon Nova mediante la API Converse de Bedrock y procese el flujo de respuestas en tiempo real.

import aws.sdk.kotlin.services.bedrockruntime.BedrockRuntimeClient
import aws.sdk.kotlin.services.bedrockruntime.model.ContentBlock
import aws.sdk.kotlin.services.bedrockruntime.model.ConversationRole
import aws.sdk.kotlin.services.bedrockruntime.model.ConverseStreamOutput
import aws.sdk.kotlin.services.bedrockruntime.model.ConverseStreamRequest
import aws.sdk.kotlin.services.bedrockruntime.model.Message

/**
 * This example demonstrates how to use the Amazon Nova foundation models
 * to generate streaming text responses.
 * It shows how to:
 * - Set up the Amazon Bedrock runtime client
 * - Create a message with a prompt
 * - Configure a streaming request with parameters
 * - Process the response stream in real time
 */
suspend fun main() {
    converseStream()
}

suspend fun converseStream(): String {
    // A buffer to collect the complete response
    val completeResponseBuffer = StringBuilder()

    // Create and configure the Bedrock runtime client
    BedrockRuntimeClient { region = "us-east-1" }.use { client ->

        // Specify the model ID. For the latest available models, see:
        // https://docs.aws.amazon.com/bedrock/latest/userguide/models-supported.html
        val modelId = "amazon.nova-lite-v1:0"

        // Create the message with the user's prompt
        val prompt = "Describe the purpose of a 'hello world' program in a paragraph."
        val message = Message {
            role = ConversationRole.User
            content = listOf(ContentBlock.Text(prompt))
        }

        // Configure the request with optional model parameters
        val request = ConverseStreamRequest {
            this.modelId = modelId
            messages = listOf(message)
            inferenceConfig {
                maxTokens = 500 // Maximum response length
                temperature = 0.5F // Lower values: more focused output
                // topP = 0.8F // Alternative to temperature
            }
        }

        // Process the streaming response
        runCatching {
            client.converseStream(request) { response ->
                response.stream?.collect { chunk ->
                    when (chunk) {
                        is ConverseStreamOutput.ContentBlockDelta -> {
                            // Process each text chunk as it arrives
                            chunk.value.delta?.asText()?.let { text ->
                                print(text)
                                System.out.flush() // Ensure immediate output
                                completeResponseBuffer.append(text)
                            }
                        }
                        else -> {} // Other output block types can be handled as needed
                    }
                }
            }
        }.onFailure { error ->
            error.message?.let { e -> System.err.println("ERROR: Can't invoke '$modelId'. Reason: $e") }
            throw RuntimeException("Failed to generate text with model $modelId: $error", error)
        }
    }

    return completeResponseBuffer.toString()
}

• Para obtener más información sobre la API, consulte la referencia sobre la API de Kotlin ConverseStreamen el AWS SDK.

Python

SDK para Python (Boto3)

nota

Hay más información al respecto. GitHub Busque el ejemplo completo y aprenda a configurar y ejecutar en el Repositorio de ejemplos de código de AWS.

Envíe un mensaje de texto a Amazon Nova mediante la API Converse de Bedrock y procese el flujo de respuestas en tiempo real.

# Use the Conversation API to send a text message to Amazon Nova Text
# and print the response stream.

import boto3
from botocore.exceptions import ClientError

# Create a Bedrock Runtime client in the AWS Region you want to use.
client = boto3.client("bedrock-runtime", region_name="us-east-1")

# Set the model ID, e.g., Amazon Nova Lite.
model_id = "amazon.nova-lite-v1:0"

# Start a conversation with the user message.
user_message = "Describe the purpose of a 'hello world' program in one line."
conversation = [
    {
        "role": "user",
        "content": [{"text": user_message}],
    }
]

try:
    # Send the message to the model, using a basic inference configuration.
    streaming_response = client.converse_stream(
        modelId=model_id,
        messages=conversation,
        inferenceConfig={"maxTokens": 512, "temperature": 0.5, "topP": 0.9},
    )

    # Extract and print the streamed response text in real-time.
    for chunk in streaming_response["stream"]:
        if "contentBlockDelta" in chunk:
            text = chunk["contentBlockDelta"]["delta"]["text"]
            print(text, end="")

except (ClientError, Exception) as e:
    print(f"ERROR: Can't invoke '{model_id}'. Reason: {e}")
    exit(1)

• Para obtener más información sobre la API, consulta ConverseStreamla AWS Referencia de API de SDK for Python (Boto3).