Amazon Bedrock Guide | Build AI Agents with Amazon Bedrock

Explore how foundation models drive generative AI and how Amazon Bedrock unifies multiple models, single API access, and knowledge base integration to simplify building AI agents.

Explore the AWS Bedrock console, including model catalog, serverless versus marketplace pricing, and model details. Learn about inference profiles for real-time and batch inference, regions, and cross-region failover.

Discover Bedrock Console agents, using route prompts and multi-model flows to build layered reasoning, knowledge bases, and guardrails, with no-code tools and model deployments.

Learn the major differences between Bedrock serverless and Bedrock marketplace, and how to configure real-time vs batch inference, prep and deploy AI agents, manage knowledge bases, and apply guardrails.

Select the right foundation model by balancing relevancy, reasoning, cost, and latency, using the Bedrock catalog to compare serverless and marketplace options.

Explore a detailed cheat sheet to select the best foundation model for your use case, comparing quality, latency, cost, context windows, and provider strengths for Amazon Bedrock.

Meet Unicornix, a fictional company used to prototype enterprise agent solutions on Amazon Bedrock, featuring unicorns with varied abilities and inventories to teach policies, operations, and use-case workflows.

Explore the unicorn care policy guide to understand facilities, locations, breed ranges, and magical abilities, and learn adoption, health care, grooming, and feeding guidelines.

Explore the unicorns training program guide overview, outlining safety rules, daily training schedules, and ability-specific exercises (healing, teleportation, time manipulation) with equipment needs and emergency contacts to train unicorns.

Create the Unicorn Care Policy Knowledge Base with vector store in Amazon Bedrock, attach an S3 data source, apply parsing, and use Teton embedding version two to power your agent.

Test the unicorn care policy knowledge base by syncing with the vector database, validating permissions, and evaluating a RAC model with targeted questions before attaching RAC to an LLM agent.

Attach a knowledge base to the unicorn care policy agent, follow step-by-step instructions, and observe how policy chunks and guardrails improve answer accuracy.

Explore guardrails for responsible AI, including input and output rules that prevent sensitive data leakage and enforce pricing controls, with filters and automated reasoning guiding safe agent behavior.

Learn to define guardrail rules for Amazon Bedrock agents by naming risks, choosing input/output scopes, selecting filters and actions, and testing prompts to ensure safe, compliant responses.

Create a new guardrail in the Amazon Bedrock console, applying eight verification steps for pricing filters. Block input and output for price tier questions and test with your agent.

Add guardrails to care policy agent by using cartrail versioning, knowledge base aliases, and resources policies. Test the setup to verify responses, traces, and blocked pricing questions for unicorns.

Define an inventory agent architecture that uses combined logical functions to query a unicorn inventory database, performing lookups, counts, and filters via predefined functions and action groups.

Define calculation rules for an agent group by building functions, attaching them to a lambda group, and triggering via action group. Apply read, filtering, and count operations on inventory data.

Create an inventory agent in Amazon Bedrock by building action groups and a Lambda function, configuring parameters, and testing inventory queries with explicit instructions.

Test agent calculation functions locally by invoking Lambda-based tests, load data from local drive or S3, parse parameters and filters, and filter inventory results through simple CSV processing.

Explore how multi-agent architecture routes questions to the right agents, using inventory and care policy agents, with optional training and reg agents, demonstrated end-to-end in parallel workflows.

Discover multi-agent systems and design a routing strategy with a super agent to orchestrate inventory and care policy agents, including combined routes and knowledge base ROC agent interactions.

Build and test ai agents with Amazon Bedrock by integrating knowledge bases and vector storage, processing text, image, and pdf data, and orchestrating multi-agent workflows in AWS.

Extend your Bedrock workflow by adding custom Lambda functions, diverse knowledge bases, and domain-specific data sets, then connect to apps and explore external triggering with the return control pattern.

Join The Intelligent Worker platform to access all current and future courses at a deep discount, with live Q&A sessions and workshops to help you build AI agents and automations.