Gen AI Training Day Workshop

Beginner Workshop

• 001.ipynb (10 mins)

  • Boto3 Client
  • Zero-Shot
  • Single Turn Conversation
  • Bedrock Runtime and InvokeModel
  • Comparing Models
    • Amazon Titan Text Express V1
    • Claude Haiku 3
  • Load Local Files
    • Use Amazon Q Developer to generate function

• 002.ipynb (10 mins)

  • Few-Shot
  • In-Context Learning
  • Packaging Function

• 003.ipynb (30 mins)

  • Multi-turn Conversation
  • Multi-turn with Invoke Model
  • Multi-turn with Converse API
  • Session or Chat History Management (Knowledgebase)
  • NOTES:
    • Message History (how is the conversation being chained)
    • Summarization (what happens if hte message history gets too long?)

• 004.ipynb (30 mins)

  • reAct Reasoning
  • Tool Use
  • Gaurd Rails

• 005.ipynb (10 mins)

  • Generate a frontend using Gradio
  • Generate a frontend using Streamlit
  • Generate a frontend using v0

Setup

Install AWS CLI

We don't directly need the AWS CLI but its useful to have around for debugging or quickly creating on off resources like an S3 Bucket

chmod u+x ./bin/install_aws_cli
./bin/install_aws_cli

SageMaker StudioLabs JuypterLabs Workspace will have AWS CLI preinstalled

Configure your AWS Credentials

You can set your AWS Credentials two ways:

• Env Vars
• AWS Credentials Profile

Cloud developer enviroments will require you set env vars since they can't persist or store an ~/.aws/credentaisls file:

export AWS_ACCESS_KEY_ID=""
export AWS_SECRET_ACCESS_KEY=""
export AWS_DEFAULT_REGION="ca-central-1"

Local developer enviroments you can use AWS configure which will persist or store credentials in a ~/.aws/credentaisls file:

aws configure

Please consider there can be limitations of models or features based on selected region. In worst-case we will fallback to nearest best suppported region>

SageMaker StudioLabs JuypterLabs Workspace will use your user roll for permissions so you don't need to set credentinals

Refernece Materials

These are other workshops or repositories that cover similar topics:

• https://github.com/aws-samples/amazon-bedrock-workshop