Using Claude 3 Haiku Vision with Amazon Bedrock Locally

 This video is a hands-on guide as how to use vision features of Anthropic's Claude 3 Haiku AI model with Amazon Bedrock.

Code Used:

import boto3

import json

import base64

from botocore.exceptions import ClientError

bedrock = boto3.client(service_name="bedrock-runtime",region_name='us-east-1')

modelId = "anthropic.claude-3-haiku-20240307-v1:0"

accept = "application/json"

contentType = "application/json"

# prompt = "What is written in this image?"

# image_path = "./images/ab55.png"

# prompt = "How many faces are there in this image and what are the expressions of those faces?"

# image_path = "./images/expression.png"

# prompt = "Tell me a short story about this image."

# image_path = "./images/hiking.png"

prompt = "What's the location in this image?"

image_path = "./images/whereisthis.png"

with open(image_path, "rb") as image_file:

    image = base64.b64encode(image_file.read()).decode("utf8")

request_body = {

    "anthropic_version": "bedrock-2023-05-31",

    "max_tokens": 2048,

    "messages": [

        {

            "role": "user",

            "content": [

                {

                    "type": "text",

                    "text": prompt,

                },

                {

                    "type": "image",

                    "source": {

                        "type": "base64",

                        "media_type": "image/png",

                        "data": image,

                    },

                },

            ],

        }

    ],

}

try:

    response = bedrock.invoke_model(

        modelId=modelId,

        body=json.dumps(request_body),

    )

    # Process and print the response

    result = json.loads(response.get("body").read())

    input_tokens = result["usage"]["input_tokens"]

    output_tokens = result["usage"]["output_tokens"]

    output_list = result.get("content", [])

    # print("Invocation details:")

    # print(f"- The input length is {input_tokens} tokens.")

    # print(f"- The output length is {output_tokens} tokens.")

    # print(f"- The model returned {len(output_list)} response(s):")

    for output in output_list:

        print(output["text"])

except ClientError as err:

    print(

        "Couldn't invoke Claude 3 Haiku Vision. Here's why: %s: %s",

        err.response["Error"]["Code"],

        err.response["Error"]["Message"],

    )

    raise