This is a submission for the Nylas Challenge: AI Expedition.
What I Built and Why
For this challenge, I built SmartInbox, an AI-powered email assistant designed to integrate AI with the Nylas API, demonstrating how these technologies can work together. The core idea was to show how AI can be trained with a user's email data to provide accurate, context-aware summaries and smart replies using Retrieval-Augmented Generation (RAG).
Hosted Auth by Nylas
To securely connect users' emails and obtain the necessary permissions, I used the hosted authentication provided by Nylas. This allows users to safely link their email accounts to SmartInbox.
AI Training with Custom Data
SmartInbox allows users to train their AI model by uploading any email dataset of their choice. This training enhances the AI's ability to generate more relevant and accurate responses based on the user's unique email data.
Specialized Summaries and Smart Replies
Once the AI is trained, users can view their emails and receive specialized summaries and smart replies tailored to the context of each email. This feature highlights the potential of combining AI with the Nylas API to streamline email management.
Please note that since this is a demo project, the functionality to send emails has not been included.
Demo
Check out this short video where I show the main features of SmartInbox and how it helps with email management. Training the AI Model with a large dataset might take a while when watching the demo.
You can also view the project live here to explore its features firsthand.
Code
You can find the source code on GitHub. This project is open source and available under the MIT license.
SmartInbox is an AI-powered email assistant that helps you efficiently manage your emails. It provides intelligent email summaries and smart reply suggestions, allowing you to handle your inbox with ease.
Features
AI-Powered Summaries: Automatically generates concise summaries of opened emails, so you can quickly grasp the main points without reading through the entire message.
Smart Reply Suggestions: Offers context-aware reply suggestions using Retrieval-Augmented Generation (RAG) based on your email data. This saves time by allowing you to respond to emails with just a click.
Training with Personal Email Data: The AI assistant is trained using your own email data, ensuring that the summaries and suggestions are tailored to your specific communication style and preferences.
Creating SmartInbox was a great learning experience. I began by setting up a full-stack app using the NEON OSS Next.js Starter, which gave me a strong base. Then, I built an easy-to-use email dashboard with shadcn/ui and connected Nylas for secure email authentication.
Nylas was key in securely connecting to users' email accounts. It allowed me to fetch email data, which I then processed with an AI model.
To make the AI work better, I trained it using public email datasets so it could give more accurate summaries and replies. I also used Claude AI to make the tool even smarter.
During this project, I learned how AI can simplify handling emails and how to combine different APIs and machine learning tools. I'm proud of how it all came together to create a tool that helps people save time. This project has inspired me to keep exploring how AI can improve everyday tasks.
Spruce Emmanuel
