SpendMood

Input page
User dashboard

Inspiration: All three of us have had experience with compulsive buying triggered by our emotions, when feeling down, stressed, or overly happy. We realized how connected emotions are to our financial habits, yet we feel as though this intersection is rarely addressed. So we decided to design a software that would explore how deep this overlap really is, helping people visualize how their emotions correlate with their moods, and hopefully aiding users in improving their spending.

What is SpendMood: SpendMood allows users to track their financial and mental health in the same software. When users first login in, they are directed straight to a form asking them about their most recent purchase, providing ease in tracking. They then have the ability to access their dashboard, which includes a calendar of their past purchases and AI-powered insights from Gemini. Though AI is rapidly improving, we don’t trust it to give medical advice just yet, so if SpendMood determines that users are making particularly harmful purchases with negative moods, the software will suggest users to seek assistance. Additionally, our minimalist UI is designed to make tracking as easy as possible, avoiding overwhelming users who may be struggling mentally.

How we built our project: We built SpendMood using React, JavaScript, HTML, CSS, and Vite to create a local web environment. We used Anthropic's Claude and ChatGPT to ensure that our styling and functionality aligned with our design vision. We started off with a few key design points: a calendar for visualization, simple interface, and emotional tracking with spending summaries, and continued to build from there as we attended workshops, learning about the tools at our disposal. Our backend is written in python. It pulls user data from the input form, passes it to Gemini for parsing and pattern analysis, returns a JSON response to add to the calendar, and runs an intervention protocol to detect excessive spending patterns.

Integrations: Auth0: allowed us to create secure user accounts with email verification and individual data storage Nessie (Capital One API): created mock data to simulate real bank account linking Gemini API: Powers our analysis engine for pattern recognition and parsing

Challenges: This was most of our team’s first time building a full-stack application with a Python and React frontend, so connecting the two was a challenge. We also were trying multiple new softwares, so it was a challenge to learn and integrate each one for our project. Another issue we ran into was configuring Gemini to provide accurate analysis. Getting personal responses from the model required extensive prompt engineering. Finally, we had difficulty deploying, the frontend worked, however the backend required a service we had never used before. Auth0 needed permissions in order to operate, which we could not implement in the given time.

What we learned: Through this project we learned GitHub collaboration across different parts of the stack, integrating Python backends with React frontends, working with multiple APIs, prompt engineering for AI models to gain insight, and implementing secure authentication. We also learned how to work together as a team having just met at the start of DivHacks.

Next steps: Implement Opik to improve consistency in our responses Develop a mobile app to increase ease of use Implement direct connects to license therapists to improve insights given