AWS SageMaker sentiment analysis

Introduction

This repo contains the terraform files needed to expose an endpoint for sentiment analysis using AWS SageMaker.

The model used is a pre-trained model from HuggingFace. Here is the link, all credits to the author. I am not responsible for any costs incurred by using this repo, please read the terraform files and understand what resources are being created before running them.

This is a demonstration of how to:

• use the API Gateway to expose a REST API protected by an API key
• use Lambda functions to process the requests and send them to SageMaker
• use SageMaker to deploy a model as an inference endpoint
• use DynamoDB to store the results of the requests
• use S3 to store the Lambda functions code
• use GitHub Actions to implement a CI/CD pipeline to deploy the infrastructure (see the Additional steps section)

Infrastructure diagram

The diagram is clearly simplified, the actual infrastructure is more complex (23 resources in total). IAM is also omitted for brevity.

The diagram is generated using Diagram, see under code/diagrams for the code.

Project structure:

• terraform/: contains the terraform files to provision the resources
• code/lambda/: contains the code for the Lambda function

Final result

Authorized request:

Unauthorized request:

Authorized request (bad request):

Installation

Prerequisites

You have the permissions in AWS to create the following resources:

• IAM roles
• S3 buckets
• SageMaker endpoint
• API Gateway
• Lambda functions
• DynamoDB tables

You (obviously) have terraform installed in your machine.

Clone the repo and provision the resources

git clone https://github.com/fedeztk/aws-sagemaker-sentiment-analysis
cd aws-sagemaker-sentiment-analysis/terraform
# adjust the variables in the variables.tf file
terraform init
terraform apply # check the output and type yes when prompted

The terraform script will output the following:

• The API Gateway endpoint
• The API Gateway path to use
• The id of the key to use in the request header

How to use the endpoint

Get the API key from the AWS console, under API Gateway -> API Keys (check that the id of the key is the same as the one outputted by terraform). Use the key in the request header as follows:

curl -X POST -H "x-api-key: $YOUR_API_KEY" $YOUR_API_GATEWAY_ENDPOINT/$YOUR_API_GATEWAY_PATH/sentiment/?data=$YOUR_TEXT

Or use whatever tool you want to make the request (Postman, Hoppscotch, etc.).

Additional steps

If you want to implement a CI/CD pipeline using GitHub Actions that deploys the infrastructure automatically when you push to the master branch, follow these steps:

• create a .github/workflows folder in the root of the repo
• create the 2 AWS secrets AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY in the repo settings, under Secrets
• create a terraform.yml file in the .github/workflows folder with the following content

name: 'Terraform CI/CD'

on:
  push:
    branches:
    - master
  pull_request:

jobs:
  terraform:
    name: 'Terraform'
    runs-on: ubuntu-latest
    
    env:
      AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
      AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY}}
    
    steps:
    - name: Checkout
      uses: actions/checkout@v2

    - name: Setup Terraform
      uses: hashicorp/setup-terraform@v1
      with:
        terraform_version: 0.11.13

    - name: Terraform Init
      run: terraform init

    - name: Terraform Format
      run: terraform fmt -check

    - name: Terraform Plan
      run: terraform plan

    - name: Terraform Apply
      if: github.ref == 'refs/heads/master' && github.event_name == 'push'
      run: terraform apply -auto-approve