AWS Bedrock

Use AWS Bedrock models with PraisonAI Agents

Models

Access foundation models via AWS Bedrock. Format: bedrock/model-id

• Claude: bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0
• Claude Haiku: bedrock/anthropic.claude-3-5-haiku-20241022-v1:0
• Llama: bedrock/meta.llama3-70b-instruct-v1:0
• Titan: bedrock/amazon.titan-text-express-v1
• Cohere: bedrock/cohere.command-r-plus-v1:0

Python

# export AWS_ACCESS_KEY_ID=your-access-key
# export AWS_SECRET_ACCESS_KEY=your-secret-key
# export AWS_REGION=us-east-1
from praisonaiagents import Agent

agent = Agent(
    instructions="You are a helpful assistant",
    llm="bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0"
)
agent.start("Explain cloud computing")

With Tools

# export AWS_ACCESS_KEY_ID=your-access-key
# export AWS_SECRET_ACCESS_KEY=your-secret-key
# export AWS_REGION=us-east-1
from praisonaiagents import Agent

def get_s3_buckets() -> str:
    """List S3 buckets."""
    return "bucket-1, bucket-2, bucket-3"

agent = Agent(
    instructions="You are an AWS assistant",
    llm="bedrock/anthropic.claude-3-5-haiku-20241022-v1:0",
    tools=[get_s3_buckets]
)
agent.start("List my S3 buckets")

Multi-Agent

# export AWS_ACCESS_KEY_ID=your-access-key
# export AWS_SECRET_ACCESS_KEY=your-secret-key
# export AWS_REGION=us-east-1
from praisonaiagents import Agent, Task, AgentTeam

architect = Agent(
    instructions="You design cloud architectures",
    llm="bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0"
)
reviewer = Agent(
    instructions="You review and optimize architectures",
    llm="bedrock/anthropic.claude-3-5-haiku-20241022-v1:0"
)

task1 = Task(description="Design a serverless architecture", agent=architect)
task2 = Task(description="Review and optimize the design", agent=reviewer)

agents = AgentTeam(agents=[architect, reviewer], tasks=[task1, task2])
agents.start()

CLI

export AWS_ACCESS_KEY_ID=your-access-key
export AWS_SECRET_ACCESS_KEY=your-secret-key
export AWS_REGION=us-east-1

# Basic prompt
python -m praisonai "Explain AI" --llm bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0

# Run agents.yaml
python -m praisonai

YAML

framework: praisonai
topic: AWS architecture design
agents:
  architect:
    role: Cloud Architect
    goal: Design scalable cloud solutions
    instructions: You are an expert AWS architect
    llm:
      model: bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0
    tasks:
      design_task:
        description: Design a highly available web application architecture
        expected_output: Detailed architecture diagram and explanation

  reviewer:
    role: Architecture Reviewer
    goal: Ensure best practices
    instructions: You review architectures for security and cost
    llm:
      model: bedrock/anthropic.claude-3-5-haiku-20241022-v1:0
    tasks:
      review_task:
        description: Review the architecture for improvements
        expected_output: Review report with recommendations