Skip to main content

​
Overview

A minimal agent that evaluates math expressions. Great for learning Orpheus.

​
Files

agent.yaml 
name: calculator
runtime: python3
module: agent.py
entrypoint: handler

memory: 512
timeout: 60

scaling:
  min_workers: 1
  max_workers: 5
agent.py 
def handler(input_data):
    expression = input_data.get('expression', '2+2')

    try:
        # Safe math evaluation
        result = eval(expression, {"__builtins__": {}}, {})
        return {
            'result': result,
            'expression': expression,
            'status': 'success'
        }
    except Exception as e:
        return {
            'error': str(e),
            'status': 'failed'
        }

​
Deploy & Run

# Deploy
orpheus deploy .

# Run
orpheus run calculator '{"expression": "10 * 5 + 2"}'
Output: 
{
  "result": 52,
  "expression": "10 * 5 + 2",
  "status": "success"
}

​
Add State (Counter)

Make it stateful using workspace: 
import os

def handler(input_data):
    counter_file = '/workspace/count.txt'

    # Load count
    count = 0
    if os.path.exists(counter_file):
        with open(counter_file) as f:
            count = int(f.read())

    # Process
    expression = input_data.get('expression', '0')
    result = eval(expression)
    count += 1

    # Save count
    with open(counter_file, 'w') as f:
        f.write(str(count))

    return {
        'result': result,
        'request_number': count
    }
Now each request shows how many times the agent has been called.

​
Test Autoscaling

# Send 10 concurrent requests
for i in {1..10}; do
  orpheus run calculator '{"expression": "'$i' * 2"}' &
done
wait

# Check scaling
orpheus stats calculator
Watch workers scale up to handle the load.

Next Example

RAG Search Agent →