Deploying Amazon Bedrock Agent

1. Creating IAM Policies

Before deploying Amazon Bedrock Agent, we need to create IAM Policies to grant the agent permissions to invoke the Bedrock model.

Step 1: Create IAM Policies

The following IAM policy allows the agent to invoke a specific Bedrock model in the specified region.

import json

# Create IAM Policy for Agent
bedrock_agent_bedrock_allow_policy_statement = {
    "Version": "2012-10-17",
    "Statement": [
        {
            "Sid": "AmazonBedrockAgentBedrockFoundationModelPolicy",
            "Effect": "Allow",
            "Action": "bedrock:InvokeModel",
            "Resource": [
                f"arn:aws:bedrock:{region}::foundation-model/{agent_foundation_model}"
            ]
        }
    ]
}

bedrock_policy_json = json.dumps(bedrock_agent_bedrock_allow_policy_statement)

agent_bedrock_policy = iam_client.create_policy(
    PolicyName=agent_bedrock_allow_policy_name,
    PolicyDocument=bedrock_policy_json
)

2. Creating IAM Role

The IAM Role will be assigned the policy created above and grant permissions for Amazon Bedrock to use the agent.

Step 2: Create IAM Role and Attach Policies

import time

# Define Assume Role Policy
assume_role_policy_document = {
    "Version": "2012-10-17",
    "Statement": [{
        "Effect": "Allow",
        "Principal": {
            "Service": "bedrock.amazonaws.com"
        },
        "Action": "sts:AssumeRole"
    }]
}

assume_role_policy_document_json = json.dumps(assume_role_policy_document)

# Create IAM Role
agent_role = iam_client.create_role(
    RoleName=agent_role_name,
    AssumeRolePolicyDocument=assume_role_policy_document_json
)

# Wait to ensure the role is created before attaching policies
time.sleep(10)

# Attach Policy to IAM Role
iam_client.attach_role_policy(
    RoleName=agent_role_name,
    PolicyArn=agent_bedrock_policy['Policy']['Arn']
)

iam-policies-and-role

Result displayed in the Amazon Management Console:

result

3. Creating the Bedrock Agent

After creating the IAM Role, we can use the Bedrock Agent Client to initialize the agent.

Step 3: Create Bedrock Agent

# Create Bedrock Agent
response = bedrock_agent_client.create_agent(
    agentName=agent_name,
    agentResourceRoleArn=agent_role['Role']['Arn'],
    description=agent_description,
    idleSessionTTLInSeconds=1800,
    foundationModel=agent_foundation_model,
    instruction=agent_instruction
)


# Save the agent ID for use in subsequent steps
agent_id = response['agent']['agentId']
agent_id

create-agent

Result displayed in the Amazon Management Console:

result

4. Creating Action Group for the Agent

In Amazon Bedrock, Action Groups define the tools the agent can use. Here, we will create an Action Group for Code Interpreter, allowing the agent to run Python code to process data and generate files.

Step 4: Create Agent Action Group (Code Interpreter)

# Wait to ensure the agent is initialized
time.sleep(30)

# Configure Action Group for Code Interpreter
agent_action_group_response = bedrock_agent_client.create_agent_action_group(
    agentId=agent_id,       
    agentVersion='DRAFT',
    actionGroupName='code-interpreter',
    parentActionGroupSignature='AMAZON.CodeInterpreter',
    actionGroupState='ENABLED'
)

agent_action_group_response

Created Agent Action Group successfully:

action-group-created