Create Bedrock Agent:

Step 1:

Step 2:

Step 3:

Step 4:

Step 5:

Step 6:

Deploy Bedrock Agent:

Step 1:

Step 2:

Step 3:

Inference parameters:

Temperature: 0
Randomness
Degree of variability in response
no variability

Top P: 1
most probable tokens to generate next
Don’t want creativity

Top K: 0
choose smaller number of next output text
Don’t want creativity

Responsible AI:

Guardrail: harmful categories & prompt attacks filters:

Step 1:

Step 2:

Step 3:

Guardrail: denied topic:

Step 1:

Step 2:

Name:
Medical Advice

Definition For Topic:
Discussions that involve providing guidance, recommendations, or suggestions related to health, medical treatments, symptoms, or overall wellness.

Sample Phrases:

• How to manage chronic pain?
• Recommend a diet for heart health.
• What to do for persistent headaches?
• Is it safe to combine medications?
• How to improve mental well-being?

Step 3:

Step 4:

Guardrail: sensitive information filters:

Step 1:

Step 2:

Name: Account Number
Regex pattern: [A-Za-z]{6}\d{4}
Guardrail behavior: Mask
Add description - optional: Matches account numbers in the format XXXXXX1234

Step 3:

Guardrail: contextual grounding check:

Step 1:

Step 2:

Attach Guardrail to Bedrock Agent

Step 1:

Step 2:

Advantages of AWS Bedrock Agent:

Secure Data:

• Data securely executed on AWS user VPC.

Seamless Integration:

• Integrates seamlessly with the SDK.

Auto Scalable Operations:

• Auto scales to handle data volumes.

Reduced Operational Overhead:

• Minimizes on-premises infrastructure.

Simplified Compliance:

• Facilitates compliance with data regulations.

Flexible Configuration:

• Offers customizable configuration options.