Phil

Inspiration

We wanted to solve the problem of having unorganized emails while also making them easier to read. Allowing users to catch up on emails, no matter their size. We also saw a Reddit post about the same issue, and this was our turning point to focus on this project.

What it does

After installing the extension, click Scan Inbox to activate it—no need to open Gmail. Using Google’s Gmail API and AI agents, the extension analyzes your most recent emails and automatically categorizes each one with a label based on its content. Emails that weren’t present during your last scan are also tagged as New. Results appear in a side panel that opens automatically when you click the extension. From there, you can click any email card to open the original message in Gmail.

How we built it

We started without a template, so we needed to do everything from scratch. First, we split into 2 teams: one focused on building the web page for the project, and the other focused on getting the API keys and users set up. After that, we worked concurrently on getting the Agent to label the emails properly and updating the UI to look sleek and modern. We then worked through overloaded APIs and duplicate tag issues.

Challenges we ran into

We ran into a lot of integration issues with the Gmail API and Gemini Flash 3.0. Through the process, our original email was flagged by Google, so we had to fix our environments for each member. We also encountered a lot of challenges with the speed of the AI agent processing the emails, which we solved with batching and adding a local storage feature, so emails do not get categorized if they have already been.

Accomplishments that we're proud of

Developing Inbox AI and building a functioning/working prototype within 24 hours is an accomplishment we’re really proud of. It was exciting to see our idea come to life so quickly, especially under a tight deadline, and we're excited to keep on developing it!

What we learned

We learned a lot about working with the Google API, RESTful API's to build a Chrome Extension, and all of the setup with authentication that it requires. It was really interesting to learn how to implement a Chrome extension from scratch and a valuable experience. Additionally, we learned about programming in JS, CSS, and HTML, working together in a single codebase.

What's next for Phil

Plans for this project are:

• Change the agent used to a fully local agent, specifically Google's Vertex Agent.
  • This keeps the user's privacy secured on their own device without any open-to-the-public Agent/ AI.
• Updates to the UI for a better and smoother experience.
• Add features:
  • Time stamps on the email stubs
  • Grouping email threads, so they look nicer and do not take up as much space

Built With

• chrome
• css
• gemini
• gmailapi
• html
• javascript