AWS Lambda ChatGPT Limerick Clock

Why?

My good friend Chris Williams needed a speaker for the April 2023 AWS Portsmouth User's Group (AWSPug) meeting. My latest mad-science-project-in-progress was a ChatGPT limerick clock. I was insipred by Whitney Merrill's ChatGPT poem clock. Since this was an AWS-themed event, I decided to write the code in AWS Lambda Serverless and teach the AWSPug how I did it. You can find a YouTube recording of that event here. There is some discussion of AI & ML followed by a discussion of Python and Lambda.

What?

Inventory for this science experiment:

An AWS Account
An OpenAI Account
Some knowledge of Python

Optional

A battery-operated Rasperry Pi Zero W with e-ink display. You can find the details for that build in the Installation section of this pwnagotchi site.

How?

Add a Billing Account to OpenAI

Choose Payment Methods on the left and Add payment method

Do the deed and click Submit :)

Create an OpenAI API Key

Now click API keys on the left and click Create new secret key

Give your new key a named and then click Create secret key

Copy this key and PASTE IT SOMEWHERE SAFE. Once you leave this screen you will never see the key again.

Create & Test a Local Python 3.7 Limerick Script

OK, let's write some code! Over in the OpenAPI Examples section, there are some Python examples. This particular sample code may not be there, but this is code I liked. Short and easy to understand.

import os
import openai

openai.api_key = 'paste-your-key-here'

response = openai.Completion.create(model="text-davinci-003", prompt="What are the three laws of robotics", temperature=0.6, max_tokens=150)
print(response.choices[0].text)

Before running the code, make sure you install the openai Python library with "pip install openai". Once we have done that, running the code results in this output:

1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
2. A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

I then made some edits to this simple OpenAI Python script to ask for a limerick that includes the current time. The time part was the hardest for me 🙂 Here is that code:

import openai
import time
from datetime import datetime

def askGPT(text):
    openai.api_key = "paste-your-key-here"
    response = openai.Completion.create(
        engine = "text-davinci-003",
        prompt = text,
        temperature = 0.6,
        max_tokens = 150
    )
    return print(response.choices[0].text)
  
def main():
    now = datetime.now()
    myRequest = "Please tell me a limerick about the time " + now.strftime('%I') + ":" + now.strftime('%M') + " " + now.strftime('%p')
    print("myRequest: ", myRequest)
    askGPT(myRequest)
    print('\n')

main()

Running this python script gives us output like this:

myRequest:  Please tell me a limerick about the time 06:53 PM


At half past six in the evening
The sun was still out and gleaming
But soon it would set
And the sky would be wet
And the stars would start their light-beaming

Create a Python 3.7 Lambda Function

Okey dokey. We now have Python that does what we want it to do - tell a time-based limerick. Let's create an AWS Lambda Python function with that code. First step: Create a shell Lambda Python function.

Click "Create function"

Choose the Hello World Python 3.7 Blueprint, name your funtion, and click "Create funtion" at the bottom of the page.

You will now have a working function that you can test if you like.

Now, let's modify this blueprint function by inserting our working ChatGPT code

import json
import datetime
import openai


print('Loading function')


def lambda_handler(event, context):
#    print("Received event: " + json.dumps(event, indent=2))
#    print("value1 = " + event['key1'])
#    print("value2 = " + event['key2'])
#    print("value3 = " + event['key3'])
#    return event['key1']  # Echo back the first key value
#    #raise Exception('Something went wrong')
    openai.api_key = "paste-your-api-key-here";
    now_var = datetime.datetime.now();
    then_var = now_var - datetime.timedelta(hours=4)
    hour_var = then_var.strftime("%I");
    minute_var = then_var.strftime("%M");
    ampm_var = then_var.strftime("%p")
    response = openai.Completion.create(
        engine = "text-davinci-003",
        prompt = "Please tell me a limerick based on the time " + hour_var + ":" + minute_var + " " + ampm_var,
        temperature = 0.6,
        max_tokens = 150
    )
    return (response.choices[0].text)

If we Deploy then Test this function, we will get an error as Lambda does not have the openai python library installed by default. Here is that error including "No module named 'openai'":

Test Event Name
BlogEvent

Response
{
  "errorMessage": "Unable to import module 'lambda_function': No module named 'openai'",
  "errorType": "Runtime.ImportModuleError",
  "stackTrace": []
}

Function Logs
START RequestId: 9adbc452-ddaa-4630-b3ac-1e7490b0bb61 Version: $LATEST
[ERROR] Runtime.ImportModuleError: Unable to import module 'lambda_function': No module named 'openai'
Traceback (most recent call last):
END RequestId: 9adbc452-ddaa-4630-b3ac-1e7490b0bb61
REPORT RequestId: 9adbc452-ddaa-4630-b3ac-1e7490b0bb61	Duration: 1.80 ms	Billed Duration: 2 ms	Memory Size: 128 MB	Max Memory Used: 36 MB	Init Duration: 112.52 ms

Request ID
9adbc452-ddaa-4630-b3ac-1e7490b0bb61

Add a Lambda Python Layer for OpenAI

As usual for me, I found 30 ways that did _not_ work for adding a python layer to Lambda (Upload ZIP from my Mac|Ubuntu|Amazon Linux, aws-cli from Ubuntu|Amazon Linux, others), but one method _did_ work. That method uses the AWS Cloud 9 IDE to add the Python layer and is detailed in this tutorial.

Create a Cloud 9 IDE by searching for and selecting Cloud 9 from any AWS page. From the main Cloud 9 page, click "Create environment"

Here is how I created my Cloud 9 IDE environment (I chose ssh as I could not get Systems Manager to work for some reason)

I only have one VPC, but chose my public subnet so that I could access the Cloud 9 instance

I then clicked "Create". The creation process takes a few minutes. Once the instance is created fully, click "Open in Cloud9"

And here we are!

Alrighty, lets create an openai python layer in Cloud9 and upload that layer into Lambda for use by our limerick function.

ec2-user:~/environment $ python -V
Python 3.7.16
ec2-user:~/environment $ mkdir python
ec2-user:~/environment $ pip install openai -t python
ec2-user:~/environment $ zip -r layer.zip python
ec2-user:~/environment $ aws lambda publish-layer-version --layer-name blog-layer --zip-file fileb://layer.zip --compatible-runtimes python3.7 --region us-east-1

{
    "LayerVersionArn": "arn:aws:lambda:us-east-1:21[SNIP]7716:layer:blog-layer:1", 
    "Description": "", 
    "CreatedDate": "2023-04-19T14:39:34.555+0000", 
    "LayerArn": "arn:aws:lambda:us-east-1:21[SNIP]7716:layer:blog-layer", 
    "Content": {
        "CodeSize": 3294785, 
        "CodeSha256": "fH4qojoKP7Ab+TUq3OV/QdVHsX2z[SNIP]j6uRThDU=", 
        "Location": "https://prod-04-2014-layers.s3.us-east-1.amazonaws.com/snapshots/21[SNIP]7716/blog-layer-60e4ab4[SNIP]495?versionId=qHhUe5v[SNIP]eNFSb&X-Amz-Security-Token=IQoJb3[SNIP]0b583b42a"
    }, 
    "Version": 1, 
    "CompatibleRuntimes": [
        "python3.7"
    ]
}

Let's jump to Lambda Layers and look for our new layer. Ah, yes, our new "blog-layer" is there.

Let's go add that later to our OpenAI Lambda function and see if it works now

Jump back to Lambda Functions, choose the function and click "Layers" then "Add a layer"

Choose "Custom layers", choose your uploaded layer from the drop-down, choose the version from the drop-down, and click "Add"

OK, let's test that function again and see if the openai library error is gone

Woo hoo! It worked!

Now let's add a URL for our function so that our clock can access the function. Click the "Configuration" tab, the "Function URL" sidebar and then click "Create function URL"

For the purpose of this blog I will choose "NONE" for authentication and then click "Save"

I now have a function URL that I can test and use to build my clock

Let's see what happens if I curl that URL from my Cloud 9 instance

ec2-user:~/environment $ curl -s  https://q4t[SNIP]hddzl.lambda-url.us-east-1.on.aws/ 


At 11:34, I'm feeling quite free,
To go out and explore, you see.
The sun is so bright,
It's a beautiful sight,
Let's go out and make some memories!

It works! This is so exciting 🙂

Display Limericks on Your Raspberry Pi Zero W Limerick Clock (WIP)

I'll just lay out the software pieces. All the hardware pieces (Raspberry Pi Zero W, e-ink display, battery) are described in the Pwnagotchi repository. LMK if you have any issues.

E-Ink Code

My e-ink display company (Waveshare) provided a bunch of sample code in their GitHub repo. I used those samples to build a Python script that showed the time and weather. I modified that code to show Lambda limerics. Here is that code that loops every 30 seconds:

#!/usr/bin/python
# -*- coding:utf-8 -*-
import sys
import os
import subprocess as sp
from time import localtime, strftime
import glob

# OpenAI Adds
import openai
import time
from datetime import datetime

picdir = os.path.join(os.path.dirname(os.path.dirname(os.path.realpath(__file__))), 'pic')
libdir = os.path.join(os.path.dirname(os.path.dirname(os.path.realpath(__file__))), 'lib')
if os.path.exists(libdir):
    sys.path.append(libdir)

import logging
from waveshare_epd import epd2in13_V2
import time
from PIL import Image,ImageDraw,ImageFont
import traceback

logging.basicConfig(level=logging.DEBUG)

try:
    logging.info("epd2in13_V2 Demo")
    
    epd = epd2in13_V2.EPD()
    logging.info("init and Clear")
    epd.init(epd.FULL_UPDATE)
    epd.Clear(0xFF)
    
    while True:
# Define fonts
        font15 = ImageFont.truetype(os.path.join(picdir, 'Font.ttc'), 15)
        font24 = ImageFont.truetype(os.path.join(picdir, 'Font.ttc'), 24)
        font48 = ImageFont.truetype(os.path.join(picdir, 'Font.ttc'), 48)
        image = Image.new('1', (epd.height, epd.width), 255)  # 255: clear the frame    
# Get limerick from Lambda

        curl_var = sp.getoutput("curl -s https://rx3[snip]ywp.lambda-url.us-east-1.on.aws/")
        draw = ImageDraw.Draw(image)
# Draw outside border
        draw.rectangle([(0,0),(249,120)],outline = 0)
        draw.rectangle([(1,1),(248,119)],outline = 0)
        draw.rectangle([(2,2),(247,118)],outline = 0)
        draw.rectangle([(3,3),(246,117)],outline = 0)
        draw.rectangle([(4,4),(245,116)],outline = 0)
# Draw limerick
        draw.text((10, -30), curl_var, font = font15, fill = 0)
        epd.display(epd.getbuffer(image))
        time.sleep(30)

    
except IOError as e:
    logging.info(e)
    
except KeyboardInterrupt:    
    logging.info("ctrl + c:")
    epd2in13_V2.epdconfig.module_exit()
    exit()

Auto-Startup

I run Raspbian Linux on my Raspberry Pi Zero W. Any code you want to run at startup goes in /etc/rc.local. I added this line to /etc/rc.local to run the python limerick script on startup:

python3 /home/pi/e-Paper/Pi/python/examples/limerick.py > /dev/null 2>&1 &

Thank you

Thanks for taking the time to read this post I hope you found it educational. I look forward to youyr feedback.

IT in Context

Search This Blog