SmartCookie

CMU F24 67443 Mobile Application Design and Development
by Alanna Cao, Emma Tong, Evelynn Chen, and Vicky Chen

SmartCookie is an iOS app that streamlines students' studying by offering tools for organizing class materials of different formats and interacting with AI-powered features such as a context-aware chatbot and dynamically generated quizzes with error tracking. Its mobile-first design enables learning on the go, providing students with a centralized, interactive, and intelligent platform for note management and learning enhancement.

Promo video

Pitch presentation

Tech stack

• Swift
• SwiftUI
• Firebase
• Firebase Storage
• OpenAI API
• UIKit
• PDFKit

Installation and Setup

Important: This project relies on Firebase and the OpenAI API, but our subscriptions for these services are no longer active. These instructions apply to anyone who wishes to personally install our app, as well as those who want to test or develop it further. To set up the app locally, follow the steps below:

1. Clone the repository

    git clone https://github.com/evelynnchen-cmu/SmartCookie.git
   

2. Add Firebase Configuration File

   • Go to the Firebase Console, create a new Firebase project (if you haven't already), and configure Firestore and Storage.
   • Once your Firebase project is set up, download the GoogleService-Info.plist file.
   • Add the downloaded GoogleService-Info.plist file to the base of the project directory. This file must not be included in the repository, so make sure it’s added to your local project and not pushed to any public repository.
   • Ensure that the Info.plist file in the app is updated with your Firebase configuration.

3. Replace the OpenAI API Key in Secrets.plist

   • Obtain your own OpenAI API key by following the OpenAI API setup guide.
   • Replace the OpenAIKey value in the Secrets.plist file with your own API key.
   • The Secrets.plist file is designed to hold your secret keys (like the OpenAI API key), and it should be added to your local project. Make sure to not include it in the repository for security purposes.

4. Open the project in Xcode. Build and run the app on an iPhone (not simulator) with iOS version 18. If necessary, can downgrade our app to run on lower iOS versions although some icons are not available in lower iOS versions and will not display (icons for notes and add new note icon (located in the top right) in CourseView/FolderView and icons for uploading new image/pdf in NoteView (located in the top right)). The UI will also be slightly different.

   • Can downgrade to iOS version >= 17.6. If the minimum iOS version is 18.0, can change it by going to Targets -> Team10Firebase -> General -> Minimum Deployments. You must also change the Info.plist.
   • If the bundleID does not work or Xcode says it is taken, change it in Signing & Capabilities to [YourAppleID].Team10Firebase. You must also make this change in the Info.plist bundle ID field (update to match signing team, [YourAppleID].Team10Firebase).

Design Decisions

SmartCookie's design prioritizes user-centric features, guided by iterative feedback and testing. The following are some of our key design decisions we want to highlight:

• Organized Note Uploads: PDF uploads are restricted to the course view, while images can be uploaded throughout the app. This distinction was made to centralize note organization and streamline workflows based on user feedback emphasizing the importance of structure in course-based contexts.

• Streamlined Settings Management: A centralized settings page reduces navigation complexity. Users can configure AI features, choosing whether to restrict functionality (e.g., chat responses, quiz generation) to their notes or include external sources. This design balances flexibility with simplicity, catering to diverse study preferences.

• Chat Prompt Recommendations: Pre-filled prompt suggestions (e.g., "Summarize the key points of...") enable faster AI chatbot interactions, reducing typing effort and enhancing usability during high-stress study sessions.

• Focused PDF Text Extraction: Designed the PDF upload feature to focus specifically on text-based PDFs, as these are more compatible with the app’s current parsing technologies. While the app does not currently extract images embedded within PDFs, this limitation reflects the trade-offs in development prioritization, with the potential for future updates to address it. This feature was prioritized because most users expressed a desire to migrate existing materials, such as their professor’s lecture slides, into the app. By enabling students to integrate these resources seamlessly, the app enhances their ability to apply its features—like summarization, organization, and quiz generation—directly to their learning.

• Streak-Based Incentives: Streak-based incentives replaced a traditional points system after feedback indicated greater alignment with students' intrinsic motivators such as maintaining daily or weekly study goals. This feature encourages consistent usage without unnecessary gamification.

Tech Decisions

• OCR (Parsing Images): OpenAI API vs. AWS Textract, Google Cloud Vision
  • Chose OpenAI's gpt4o-mini model because it was best suited for our use case and was the most cost effective. AWS Textract broke down the given image into individual words and gave more information (like its bounding box and confidence score) that we didn't need. OpenAI's gpt-vision model performed about the same as 4o-mini, but took significantly longer to parse than 4o-mini.
• PDFKit vs. OpenAI API parse pdf
  • Chose PDFKit over OpenAI for parsing to optimize for local processing and cost efficiency.
  • While using the OpenAI API for parsing a PDF similarly to how images are parsed would've allowed us to maintain consistency within our app, PDFKit offered all of the basic functionality for supporting PDF upload and parse that our users wanted prioritized the most.
• Firebase Firestore vs AWS S3
  • Chose Firebase Firestore because there was more classroom help with this service and it is built to integrate with Firebase.

Future Extensions

• Expand app to support accessibility features like audio transcriptions
• Offer better support for parsing and displaying advanced content such as code, math symbols, etc.
• Support PDF image extraction
• Allow users to friend other users to keep each other accountable or start friend streaks similar to Duolingo.