☁️ 🔧 AWS Agents with Phidata Framework

• AI-powered agents to manage AWS services (EC2, S3, Lambda) using natural language commands.
• Built with phidata and boto3.

I used Groq (llama-3.3-70b-versatile) which provides a Free API Key.

🚀 Features

• EC2 Management: List, start and stop EC2 instances.
• S3 Management: List buckets, upload/download files
• Lambda Management: List and invoke Lambda functions.
• AI Integration: Execute tasks using natural language commands.

📋 Prerequisites

1. Permissions Required:

   • AmazonEC2FullAccess (for EC2)
   • AmazonS3FullAccess (for S3)
   • AWSLambda_FullAccess (for Lambda)

2. Libraries:

   pip install boto3 phidata

🔐 Configuration

AWS Credentials Setup

Option 1: Environment Variables (dotenv)

# Linux/Mac
export AWS_ACCESS_KEY_ID="AKIAXXXXXXXXXXXXXXXX"
export AWS_SECRET_ACCESS_KEY="XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX"
export AWS_REGION= "YOUR REGION"
export GROQ_API_KEY = "YOUR_API_KEY"

# Windows (PowerShell)
$env:AWS_ACCESS_KEY_ID="AKIAXXXXXXXXXXXXXXXX"
$env:AWS_SECRET_ACCESS_KEY="XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX"
$env:AWS_REGION= "YOUR REGION"
$env:GROQ_API_KEY= "YOUR_API_KEY"

Option 2: AWS Credentials File

Create or edit ~/.aws/credentials (Linux/Mac) or %USERPROFILE%\.aws\credentials (Windows):

[default]
aws_access_key_id = AKIAXXXXXXXXXXXXXXXX
aws_secret_access_key = XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
region = "YOUR_REGION"

Imports

from phi.agent import Agent
from phi.tools.aws_lambda import AWSLambdaTool (comes with phidata)
from ec2tool import EC2Tool 
from s3tool import S3Tool 
from phi.model.Groq import Groq
from dotenv import load_dotenv

load_dotenv()

🛠️ Usage (check official doc for more)

ec2_agent = Agent(
    tools=[EC2Tool(region_name=YOUR_REGION)],
    model=Groq(id="llama-3.3-70b-versatile"),
    name="AWS EC2 Agent",
    instructions="...",  # (I will be posting great Instructions in this repo)
    show_tool_calls=True,
)

s3_agent = Agent(
    tools=[S3Tool(region_name=YOUR_REGION)],
    name="AWS S3 Agent",
    model=Groq(id="llama-3.3-70b-versatile"),
    instructions="...",  # (I will be posting great Instructions in this repo)
    show_tool_calls=True,
)

lambda_agent = Agent(
    tools=[AWSLambdaTool(region_name=YOUR_REGION)],
    model=Groq(id="llama-3.3-70b-versatile"),
    name="AWS Lambda Agent",
    instructions="...",  # (I will be posting great Instructions in this repo)
    show_tool_calls=True,
)

💡 Examples

EC2 Agent

# List all EC2 instances
ec2_agent.print_response("List all EC2 instances.")

# Start an EC2 instance
ec2_agent.print_response("Start the EC2 instance with ID i-1234567890abcdef0.")

S3 Agent

# List S3 buckets
s3_agent.print_response("List all S3 buckets.")

# Upload a file
s3_agent.print_response("Upload example.txt to the bucket named my-bucket.")

Lambda Agent

# List Lambda functions
lambda_agent.print_response("List all Lambda functions.")

# Invoke a Lambda function
lambda_agent.print_response("Invoke the Lambda function named my-function.")

---

I AM NOT SURE :(

1. I am unsure about the AmazonEC2FullAccess, AmazonS3FullAccess, and AmazonLambdaFullAccess permissions.

2. Full Access to some AWS services may not be a great idea.

3. Because in more complex & indirect cases like:

   "Download this (if needed tools are given) and load the summary to the related named S3 bucket"

   can cause unwanted/meant operations decided by Agent.

4. I have concerns about granting an agent full access to AWS service usage, as it might lead to unexpected costs.

5. Idk when I take some advanced courses and keep learning, Some of these can be fixed.

• Phidata Documentation: https://phidata.com
• boto3 Documentation: boto3 Docs

NOTE: A freshman creates this repository. I believe some great opinions about this repo would lead me to learn & develop more.