The Developer Space

5 0

OpenAI provides APIs to programmatically access the functionalities provided by ChatGPT, which allows you to integrate Artificial Intelligence capabilities into your applications and services. This article provides a step-by-step guide to setting up the environment and invoke OpenAI APIs from your AWS Lambda functions. Examples are provided in Python 3.9, but you should be able to use this using any supported language. You can find a list of all supported libraries here.

Sign-up for an OpenAI Account

Go to OpenAI API portal and click on the Log In or Sign Up buttons to login to the OpenAI API portal. Sign-up or Login with your credentials. New users receive a $18 credit which is valid for 3 months. After the free credit limit is exhausted or 3 months, you’ll have to add a payment method to continue using the APIs.

Create an API Secret Key

Click on the Profile image at the top-right of the page and select ‘View API Keys’ menu. Click on the Create new secret key button. You’ll see a popup with the newly generated key. Make a note of this key since you will not be able to get the full key later on. You’ll need the full key to invoke OpenAI APIs.

Once you’ve successfully created the key, you should see something similar to this:

You can delete existing keys and create new ones as many times as you like. You generally delete a key when you think that a key has been compromised and someone else might be using it.

Create a Lambda Layer for OpenAI Python library

There are two ways to invoke OpenAI APIs from your Lambda functions:

1. Use the requests library and invoke the REST endpoints. This approach requires a multitude of manual configurations and is very cumbersome. This article will not discuss this approach.

Install OpenAI Python package locally

2. Use the openai Python library. The library abstracts all complexities into a simple to use python library. To use this library from within Lambda functions, you need to create a Lambda Layer. Lambda Layers make it easy to manage dependencies in Lambda functions. Layers are created using zip files that are extracted and deployed along with Lambda functions as if the files were copied locally. To create the Layer, you need to first install the library locally and create a zip file out of it. Use the following command in VS Code Terminal or Python command line to install the library to a custom path.

pip install openai --target <<local path>>

After the installation is complete, create a folder called ‘Python’ and move all files under it, and then zip the folder. The zip should look like this when extracted:

Create the Lambda Layer

1. Open AWS Console
2. Go to Lambda Console
3. Select Layer from the left menu
4. Click the Create Layer button on to top right.
5. Provide a name for the Layer. For example, openai_python_layer
6. Select Upload a .zip file option and click the Upload button. Select the zip file you created in the previous step.
7. Select compatible architectures. We’ll use x86_64 alone in this example.
8. Select Python 3.9 under Compatible runtimes

When the final screen should look like this, click the Create button.

Create a reusable Lambda function for getting the OpenAI secret key

You need a way to store the OpenAI secret key safely and reuse it across multiple Lambda functions. We’ll do this by creating a Lambda specifically for storing the secret key in an Environment Variable and returning it. Lambdas that actually invoke the OpenAI APIs will first invoke this Lambda and get the key.

Go back to the Lambda console and click Functions on the left menu. Click Create function button and enter the following details:

Function name: openai_get_api_key

Runtime: Python 3.9

Architecture: x86_64

Under Permissions, select Create a new role with basic Lambda permissions

Click Create function button

Once to the function is created, go to the Configuration tab and click Environment variables section. Click Edit and add the following item:

Key: openai_api_key

Value: <<the api key created at the beginning of this article>>

You screen should look like this:

Go back to Code section and paste the code below:

import json
import os

def lambda_handler(event, context):
    return {
        'statusCode': 200,
        'body': {
            'api_key' : os.environ['openai_api_key']
        }
    }

Test your function to make sure it returns the correct key. Using Environment variables like this makes it easy to update the key in future without changing the code also protects the key from being hard-coded in the code.

Create a Lambda function for accessing the OpenAI API

To access OpenAI API, create a new Lambda that actually invokes the APIs and receives the processes the results. Go back to the Lambda console and click Functions on the left menu. Click Create function button and enter the following details:

Function name: openai_invoke_api

Runtime: Python 3.9

Architecture: x86_64

Under Permissions, select Create a new role with basic Lambda permissions

Click Create function button.

Add Role permissions to invoke the Lambda to get API Keys.

The function that you just created needs permission to invoke the Lambda that returns the API key. Go to the Configuration tab and click Permissions section. Under Execution role section, open the Role in a new browser tab.

Click on Add Permissions button and select Create inline policy menu option.

Service: Lambda

Actions: InvokeAsync, InvokeFunction

Resources: <<Enter the ARN of the openai_get_api_key Lambda function>>

In the next screen, give the policy a name and create the policy

Increase Lambda Timeout

The default timeout for Lambda functions is 3 seconds. OpenAI APIs may take more than 3 seconds to execute and return a response based on the API invoked, the model used, and the input parameters provided. Increase timeout to 15 seconds or more to ensure that the Lambda function executes successfully.

Go to the Configuration tab and click General configuration section. Click the Edit button and increase the timeout in the popup and save it.

Note that Lambda functions are charged based on the execution time and higher execution time may result in higher bills.

Add Lambda Layer

To actually start using the openai library in your Lambda functions, you need to add the OpenAI Layer your created in the previous section to your Lambda function.

In the function screen, scroll to the bottom of the page. Under the Layers section, click the Add a layer button. Click on the Custom layers option and select the Layer you created in the previous section. Select version 1 and click Add.

Invoke OpenAI API

There are many API that you can try out. Commonly used APIs are Text Completion, Code Completion and Image APIs. We will see an example of a text completion API in this article using the text-davinci-003 model.

Paste the code below into the code editor and click the Deploy button. Create a Test event and execute the code. Do not forget to change the function ARN in line # 32.

import json
import openai
import boto3

def lambda_handler(event, context):
    
    model_to_use = "text-davinci-003"
    input_prompt="Write an email to Elon Musk asking him why he bought Twitter for such a huge amount"
    
    openai.api_key = get_api_key()
    response = openai.Completion.create(
      model=model_to_use,
      prompt=input_prompt,
      temperature=0,
      max_tokens=100,
      top_p=1,
      frequency_penalty=0.0,
      presence_penalty=0.0
    )
    #print(response)
    text_response = response['choices'][0]['text'].strip()
    return {
        'statusCode':200,
        'body': {
            'response' : text_response
        }
    }
    
def get_api_key():
    lambda_client = boto3.client('lambda')
    response = lambda_client.invoke(
            FunctionName = 'arn:aws:lambda:us-east-1:000000000000:function:openai_get_api_key',
            InvocationType = 'RequestResponse'
        )

    openai_api_key = json.load(response['Payload'])['body']['api_key']
    return openai_api_key

You should see a response similar to this.

{
  "statusCode": 200,
  "body": {
    "response": "Subject: Question about Twitter Acquisition\n\nDear Elon Musk,\n\nI am writing to you to ask why you decided to purchase Twitter for such a large amount. As a tech enthusiast, I am curious to know what motivated you to make such a large investment.\n\nI am sure that you have a good reason for doing so, and I would love to hear your thoughts on the matter.\n\nThank you for your time and I look forward to hearing from you."
  }
}

Note that OpenAI models are generative models and they usually generate different response to the same prompt. If you receive a completely different response, you know why.

If you wish to see the complete response from the API, uncomment line # 20.

Conclusion

Using OpenAI APIs, you can embed Artificial Intelligence into your applications and services. This is a powerful feature that expands the capabilities of your applications and services beyond imagination. Try out with different APIs and different models and parameters to see how this can help you augment the capabilities of your applications and services.

Enjoyed the article and found this useful? Leave a comment below with your feedback and suggestions.

References

1. OpenAI API Documentation
2. Lambda Layer – AWS Documentation