
Weekend Annoying Task Challenge: EmailSwift AI
EmailSwift AI is a fully serverless web application built for the Weekend Annoying Task Challenge. Powered by AWS Lambda and Amazon Bedrock, it acts as your personal communications assistant to instantly draft professional email replies and stop you from overthinking your inbox.
Vision & What the app does
Like many professionals in the fast-paced tech industry, one of the most annoying, tedious parts of my week is managing my overflowing email inbox. Reading an email takes only a few seconds, but overthinking the reply—ensuring it sounds professional, polite, and concise—often takes five to ten minutes per message. Over an entire week, this mundane task consumes hours of valuable time that could be spent actually building and coding.
To solve this problem for the weekend challenge, I decided to build EmailSwift AI. EmailSwift AI is a streamlined, web-based application that acts as your personal communications assistant. By simply pasting an incoming email into the application interface, it automatically drafts a highly professional, context-aware reply in a matter of seconds. I was inspired by fellow community builder Deepak Raj JS, who built a fantastic tool called ReplySwift AI, and I wanted to remix that concept using my own lightweight serverless architecture.

How I Built It
I wanted to build this solution rapidly over the weekend while staying entirely within the AWS Free Tier. My development process started with writing a single Python script that could simultaneously handle both the frontend user interface and the backend processing.
A pivotal decision in my design was to utilize an AWS Lambda Function URL instead of provisioning a full Amazon API Gateway. By configuring the Lambda function to check the incoming HTTP method, I could serve a clean HTML and JavaScript interface directly to the user when the function receives a standard
GET request. Then, I configured the frontend JavaScript to trigger the AI text generation via a POST request to the exact same URL.
AWS Services Used / Architecture Overview
My architecture is highly streamlined and leverages two primary AWS services, keeping the footprint light and cost-effective:
- AWS Lambda: This acts as the core compute engine and web server. It is exposed to the internet via a Function URL with no authentication required (
Auth type: NONE), making it incredibly simple to use and demonstrate.
- Amazon Bedrock: This powers the generative AI capabilities. I used the
amazon.nova-lite-v1:0model because it is exceptionally fast, highly cost-effective, and perfectly suited for generating short-form text like email replies and summaries.
Architecture Flow: The user accesses the Lambda Function URL in their browser, which serves the frontend. When they paste an email and click "Draft My Reply," a JavaScript
fetch call sends a POST request back to the Function URL. The Lambda function constructs a strict prompt and uses the boto3 SDK to invoke Amazon Bedrock's Converse API. The Nova Lite model drafts the perfect reply, and Lambda returns it directly to the user's screen.What I Learned This challenge provided a massive learning opportunity, particularly when it came to debugging cloud integrations and API responses. While building the app, I encountered a highly frustrating error when trying to process the response from Bedrock. My code kept throwing this error:
Error: list indices must be integers or slices, not str .Initially, I couldn't understand why the application was failing. Through relentless debugging during this challenge, I learned a crucial detail about the
boto3 SDK: Amazon Bedrock's Converse API returns the generated message content as a list of blocks, rather than a simple dictionary string. I had to update my Python code to explicitly target the very first index of that array by adding `` to the extraction path ( response['output']['message']['content']['text'] ). Finding and fixing that stubborn bug gave me a much deeper understanding of how the AWS SDK interacts with generative AI models, and the immense satisfaction of finally seeing the app work perfectly was well worth the effort.Amazon Bedrock's Nova Lite model
Amazon Nova Lite (amazon.nova-lite-v1:0) is a generative AI model accessed through Amazon Bedrock. In your EmailSwift AI architecture, it acts as the core "AI Engine" that receives raw input text and intelligently transforms it into structured, professional output.
Based on our setup and the success of other builders in the challenge, here is an overview of its key features and why it is the perfect "sweet spot" model for this project:
- Exceptional at Text Extraction and Formatting: Nova Lite excels at taking messy, unstructured context—like chaotic meeting notes or hasty emails—and extracting the core intent to generate polished summaries, formatted action items, and ready-to-send replies
- Highly Cost-Effective (Free Tier Eligible): It is a lightweight model that qualifies for use within the AWS Free Tier (which provides up to $200 in credits for new accounts). This makes it ideal for running fully serverless weekend projects without accruing unexpected costs.
- High-Speed Generation: Because it is a "Lite" model, it is incredibly fast. It is capable of reading context and generating professional, context-aware email replies or weekly digests in just a matter of seconds.
- Seamless Serverless Integration: It integrates natively with AWS Lambda using the
boto3SDK's Converse API. As we learned during debugging, it returns its generated content in a highly structured list array format, making it easy to extract the exact text blocks you need for your web frontend.

Link to App You can try out EmailSwift AI live right here:
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article