Amazon Bedrock with AI Agents

Discover Amazon Bedrock, a serverless platform delivering foundation models via a single api for fast inference and easy data customization. Use Bedrock for text generation, chat, search, and image tasks.

Explain the difference between AI and GenAI, showing AI analyzes data and follows rules to interact with environments and achieve goals, while GenAI creates new content from prompts.

Join this hands-on module to set up Bedrock by provisioning your AWS account and CLI, enable programmatic access, access foundation models, and understand model pricing to manage costs.

Create and secure an AWS account by configuring a root user, then set up an IAM user with admin access, enable multi-factor authentication, and verify identities.

Install the AWS CLI, set up IAM user access keys, configure with aws configure, and verify access by listing S3 buckets in us-east-1 with JSON output.

Learn how to request access to foundation models in Amazon Bedrock, including managing model access, selecting models, submitting use case details for Anthropic models, and troubleshooting access errors.

Explore Amazon Bedrock pricing across on demand, batch mode, provision throughput, and model customization, with token costs and examples like Jurassic 2 Mid and Ultra.

Explore Amazon Bedrock playgrounds—text, image, and chat—learn to select models and run prompts, with examples comparing Python and Java and generating a festive Chihuahua image.

Explore the chat playground in Amazon Bedrock, run llama 2 chat 13b with step-by-step prompts, and adjust temperature and top_p to observe how inference parameters shape answers to math prompts.

Compare Bedrock text playground and Bedrock chat playground, highlighting interface, input methods, context handling, capabilities, and use cases for code testing versus dialogue.

Explore prompt engineering and how text and image prompts drive generative AI results. Highlight productivity gains: up to 17% for top performers and 43% for lower performers.

Learn to craft effective image prompts by defining clear objectives, applying image and artistic styles, emotional tone, and color, while adjusting abstraction and complexity in hands-on prompts.

Explore text prompting fundamentals, including defining purpose, giving the AI a role, specifying audience, providing context, setting length, and guiding tone through hands-on prompts and audience-tailored examples.

Explore how inference parameters shape AI predictions in Amazon Bedrock, including temperature, top k, and top p, along with length controls, stop sequences, and repetition penalties.

Explore the Amazon Bedrock examples pane by filtering by model, modality, category, and provider. View prompts and API request, then run Amazon Titan Text G1 express in playground.

Install Python from python.org, verify that pip is installed, then install libraries like boto3 with pip; run python -m webbrowser for the image example, completing the setup.

Master integrating Amazon Bedrock with Python by using boto3 to invoke Bedrock models with prompts, handle json responses, and write outputs to files in a hands-on workflow.

Integrates Bedrock with Python using boto3 to invoke the Bedrock runtime, process base64 image artifacts, and display results in an HTML page.

Explore Amazon bedrock agents and agentic AI, using action groups, knowledge bases, and memory to autonomously handle multi-step tasks like booking one-day tours in Amman via Rag.

Showcase building and managing agentic ai agents in amazon bedrock using agent builder, linking to python and lambdas, configuring versions and aliases, and managing knowledge bases, cost considerations, and guardrails.

Create DynamoDB tables, tour and tour booking, with tour name as the partition key and tour date as the sort key; populate the tour table from results.csv using a Python script.

Create read and upsert lambdas and connect them to a bedrock agent's action groups to enable tour bookings from available slots.

Develop and test a second action group, tour booking upsert, wiring it to a Lambda with an OpenAI API schema and DynamoDB storage, and prepare a Rag knowledge base.

Create a knowledge base with an S3 data source and vector store using Titan text embeddings v2 on Amazon Bedrock; index a PDF and enable question answering and booking flows.

Connect an Amazon Bedrock AI agent to Python using Boto3 to send messages with an agent id, alias id, and region. Demonstrate alias creation and Jerash tour bookings.

Clean up your AI agents project by deleting action groups, memory, and related data sources, then remove the Amazon OpenSearch Service collection to avoid ongoing charges.

Explore cost details for the hands-on demo, noting most resources stay within the free tier. OpenSearch vector knowledge base can cost $0.50 per hour, so clean up to avoid charges.

Conclude this course and keep it as a reference to revisit and apply the learnings, recognizing your investment in yourself as a path to success.