Our landing page!

IMPORTANT

To test use credentials:

User: tester@journai.app Password: test123

Inspiration

It is common recommendation from therapists to their clients to journal. This practice is meant to help clients reflect on their feelings and work through them. We thought that this practice could be made even more fruitful by using AI to break down their day with them.

What it does

Our project is a web based, AI powered journal. A user creates an account on the site using their email and logs into the app. From there, the user can look at some of their previous journal entries, or compose a new one. When the user writes a new entry, it is sent to GPT-3 which performs a sentiment analysis. The sentiment is ranked on a scale 0-10. GPT-3 returns its analysis as well as any notable keywords in your entry. The app then shows the user the sentiment analysis and asks the user to write a follow-up entry based on the keyword(s). If the analysis concludes the user may be in a particularly dangerous mental state, we have triggers to recommend hotlines and resources to them.

How we built it

We used firebase to host the website, to handle user accounts and authentication, and act as our database. We used Google Cloud Platform to handle API requests to OpenAi's services. We used HTML, CSS, and JavaScript as the glue.

Challenges we ran into

We have never used Firebase before, so we had to read a lot of documentation and watch some tutorial videos to get spun up. We also used Firebase's Firestore database is nonrelational, which none of use have ever used before. We encountered race conditions when trying to write and then retrieve journal entries. We have to familiarize ourselves with JavaScript's async format. It took us a while to get it working, trying to fix our queries for saving and retrieving user entries. We have never used Google Cloud Platform before, so it took us a lot of time to getting our cloud functions that handled our API requests to work within Firebase.