Expense email workflow

This repo contains an email-based workflow for turning expense emails into rows in a spreadsheet. This repo is based on my aws-email-workflow-starter template.

🧐 How it works

graph LR
  A[Email] --> B[SES]
  B --Stores--> C[S3]
  C --Triggers--> D[Lambda]
  D --Add row--> E[Sheet]

Loading

1. An expense email is forwarded to a specific address.
2. SES receives the email and stores it in an S3 bucket.
3. S3 triggers a Lambda function.
4. The Lambda function reads the email from S3 and sends a prompt along with the email's text to GPT.
5. The GPT response includes details like the amount and who the expense was for.
6. The Lambda function adds a row to a Google Sheet with the details.

📂 Project structure

├── .github             # GitHub Actions workflows
├── bin
│   └── aws-app.ts      # AWS CDK entry point
├── lib
│   ├── reader.ts       # Lambda function for reading emails
│   ├── stacks          # AWS CDK stacks
│   └── constructs      # Groupings of AWS resources
└── tests

🧰 Setup instructions

To setup your initial AWS environment, follow these instructions.

🚀 Deploying updates

After you've ran through the setup instructions, you can deploy updates to the Lambda function and AWS resources by running:

npm run deploy

💅 Tips

• AWS CDK library docs are here

• By default, SES is in "Sandbox" mode. This means that you can only send emails to verified email addresses. This might be fine for personal projects. You can take this action to move out of sandbox mode..

• Resource not showing up in AWS? Make sure the region you're viewing in the console matches the region you're deploying to.

• If initial deploy fails and gets stuck in a failed state, you can delete the stack and try again. To preserve the SES domain identity, and destroy all other AWS resources (S3 bucket, SES rules, and Lambda function):

  • Empty the S3 bucket

  • Run the following, where {RESOURCE_PREFIX} is the environment value from your .env

    npx cdk destroy --exclusively {RESOURCE_PREFIX}Reader

    Alternatively you can run npx cdk ls to view the stack names that you can pass to the --exclusively flag.