Bedrock Runtime Converse SDK Document Block does not recognize Uint8Array() when reading PDF files

[Ceebig]

Checkboxes for prior research

• I've gone through Developer Guide and API reference
• I've checked AWS Forums and StackOverflow.
• I've searched for previous similar issues and didn't find any solution.

Describe the bug

Based on the AWS Bedrock Runtime SDK documentation, if a file is present in a message, it should be passed in as an Uint8Array, and the SDK will convert the Uint8Array into based 64 encoded string.

File upload is handled by middleware multer, I used the readFileSync method to read the PDF file as raw bytes and then convert it to an Uint8Array as shown in the code. However, it keeps emitting the error message Error: @smithy/util-base64: toBase64 encoder function only accepts string | Uint8Array. I did a console.log() of the Uint8Array that I created, and I can confirm it is a Uint8Array; the SDK does not seem to recognize it for some reason. Thank you.

Error: @smithy/util-base64: toBase64 encoder function only accepts string | Uint8Array.
    at toBase64 (/node_modules/@smithy/util-base64/dist-cjs/toBase64.js:15:15)
    at Object.bytes (/node_modules/@aws-sdk/client-bedrock-runtime/dist-cjs/index.js:1810:41)

Regression Issue

• Select this option if this issue appears to be a regression.

SDK version number

@aws-sdk/client-bedrock-runtime@3.716.0

Which JavaScript Runtime is this issue in?

Node.js

Details of the browser/Node.js/ReactNative version

v20.8.1

Reproduction Steps

const fs = require("fs");
const { BedrockRuntimeClient, ConverseCommand } = require("@aws-sdk/client-bedrock-runtime")
const config = {
  region: "us-east-1",
  credentials: {
    accessKeyId: awsKeyID,
    secretAccessKey: awsKeySecret
  }
}
const bedrockRuntimeClient = new BedrockRuntimeClient(config)

let files = req.files
files.forEach(file => {
      // Read the file synchronously
      const filePath = path.resolve(file.path);
      const rawBytes = fs.readFileSync(filePath);

      userPrompt.content.push({
        document: {
          format: file.originalname.split('.').pop(),
          name: new StringService().sanitizeFileName(file.originalname),
          source: {
            bytes: new Uint8Array(rawBytes),
          }
        }
      })
    })

messages.push(userPrompt)

const input = { // ConverseRequest
     modelId: modelID, // required
     messages: messages,
     inferenceConfig: inferenceConfig,
};

const command = new ConverseCommand(input);
return await bedrockRuntimeClient.send(command)

Observed Behavior

Error: @smithy/util-base64: toBase64 encoder function only accepts string | Uint8Array.
at toBase64 (/node_modules/@smithy/util-base64/dist-cjs/toBase64.js:15:15)
at Object.bytes (/node_modules/@aws-sdk/client-bedrock-runtime/dist-cjs/index.js:1810:41)

Expected Behavior

The AWS SDK should accept the document block with bytes in Uint8Array without any errors

Possible Solution

No response

Additional Information/Context

No response

[zshzbh]

Hey @Ceebig,

Can you check what is the type of userPrompt ? I used a single pdf and it works, I can share my code below -

// Use the Conversation API to send a text message to Anthropic Claude.

import {
  BedrockRuntimeClient,
  ConverseCommand,
} from "@aws-sdk/client-bedrock-runtime";
import fs from "fs";
const rawBytes = fs.readFileSync('./DogPolicy.pdf')
// Create a Bedrock Runtime client in the AWS Region you want to use.
const client = new BedrockRuntimeClient({ region: "us-west-2",
  credentials: {
    accessKeyId: "XXX",
    secretAccessKey: "XXXX",
  },
});

  const uint8 = new Uint8Array(rawBytes);

// Set the model ID, e.g., Claude 3 Haiku.
const modelId = "anthropic.claude-3-haiku-20240307-v1:0";

// Start a conversation

const conversation = [
  {
    role: "user",
    content: [
      {
        document: {
          format: "pdf",
          name: "DogPolicy",
          source: {
            bytes: uint8,
          },
        },
      },
      {
      text: "Describe this pdf",
      }
    ],
  },
];

// Create a command with the model ID, the message, and a basic configuration.
const command = new ConverseCommand({
  modelId,
  messages: conversation,
  inferenceConfig: { maxTokens: 512, temperature: 0.5, topP: 0.9 },
});

try {
  // Send the command to the model and wait for the response
  const response = await client.send(command);

  // Extract and print the response text.
  const responseText = response.output.message.content[0].text;
  console.log(responseText);
} catch (err) {
  console.log(`ERROR: Can't invoke '${modelId}'. Reason: ${err}`);
  process.exit(1);
}

And the result I got
This PDF document outlines Amazon's policy for allowing employees to bring their dogs to work at select corporate offices. The key points are: ....

I think the root cause is the type of rawBytes or the field in content is not string or Uint8Array Can you console log the rawBytes and userPrompt and their type for further investigation?

Thanks

[github-actions]

This issue has not received a response in 1 week. If you still think there is a problem, please leave a comment to avoid the issue from automatically closing.

[github-actions]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs and link to relevant comments in this thread.