bedrock-inference

_{Keywords: generative AI, Amazon AWS, Bedrock, API, wrapper, LLM, inference, NLP.}

Bedrock Inference is a simple python package to handle 2FA and on-demand calls to AWS Bedrock foundation models.

Requirements

• An AWS IAM account
• An Access key ID (es: AKIAIOSFODNN7EXAMPLE) and its correspondent secret access key (es: wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY) for API requests
  • These credentials can be created through Home --> Security credentials --> Access Keys --> Create access key
  • If you have AWS CLI installed and set up on your machine, chances are your credentials are already stored at ~/.aws/credentials and/or in the AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY environment variables. In this case, the package will retrieve them automatically.
    • Note: AWS CLI is not required to use bedrock-inference.
• A Multi-factor authentication device ARN (es: arn:aws:iam::012345678910:mfa/example_username)
  • The ARN of your MFA device can be located at Home --> Security credentials --> Multi-factor authentication
• The 6 digits temp token generated by the MDA device (es: 366 712)

Installation

1. Make sure you have the latest version of pip installed

   pip install --upgrade pip
   

2. Install araucanaxai through pip

   pip install bedrock_inference
   

Usage

import from bedrock_inference import bedrock

# Authenticates through MFA and establish a session
session = bedrock.aws_login_mfa(arn="MFA_device_arn", token="MFA_token", aws_access_key_id="your_ID", aws_secret_access_key="your_key")
#session =  bedrock.aws_login_mfa(arn="MFA_device_arn", token="MFA_token") #use this if you already have AWS credentials set up for API requests on your machine

# Instantiate youre Bedrock caller
caller = bedrock.Bedrock(session=session, region="your_aws_region")

#Check available models in your region
print(caller.list_models())

#Invoke the target AWS model and retrieve the generated answer
answer = caller.invoke_model(prompt="What's the conceptual opposite of 'Hello World'?", model_id="target_aws_model_id")
print(answer)

Contacts and Useful Links

• Repository maintainer: Tommaso M. Buonocore

• Project Link: https://github.com/detsutut/bedrock-inference

• Package Link: https://pypi.org/project/bedrock-inference/

License

Distributed under MIT License. See LICENSE for more information.