Delve

Inspiration

We were inspired to create Delve because of our own experiences journaling. We've all gone through the ups and downs of navigating life's challenges, and journaling has played a critical role in helping us work through them. However, when we recommended journaling to our friends, many of them expressed difficulty creating insightful journal entries. This made the journaling experience less rewarding for them, making it difficult for them to find motivation to write. We created Delve to help kickstart their journaling habit by providing personalized prompts based on their own writing and presenting statistical insights into their mood.

What it does

Delve is an AI-assisted journaling website that uses a Large Language Model to process the content in the user’s journal entry and generate questions that prompt the user to delve deeper into their thoughts. The model finds various topics that the user mentions in their entry and generates questions for each topic. Once the user responds to one of the questions generated by the model, the model once again processes the content in the user’s journal entry and generates follow-up questions that prompt the user to delve deeper into their thoughts. This process continues until the user is satisfied with their journaling experience, resulting in a hierarchical structure that resembles a tree structure. Furthermore, using data from the user’s various journal entries, the website creates a mood board that provides a high level analysis of the user’s mood trends and recurring themes in their journal entries.

How we built it

We built an interactive prototype of Delve using Figma, creating graphics with DALL-E 3. For the actual website, we used the React framework and coding languages such as HTML, CSS, and JavaScript to develop the front-end. To generate the entry-based journaling questions and prompts, we used Perplexity’s API to access a Large Language Model. This model followed a tree structure, allowing the user to delve deeper into their thoughts.

Challenges we ran into

One of the challenges we faced was learning how to use GitHub to collaborate with each other when building the website. We made the mistake of working independently on the same branch, which led to merge conflicts when we made changes to the same lines of code or the same files.

Additionally, coming into DubHacks, none of us had ever used Perplexity's API before. Learning how to integrate the API into our front-end and ensuring that the LLM provided questions specific to the journal’s content took lots of quick learning and flexibility.

Another challenge we experienced was incorporating interactive text entry fields in the login and sign-up pages of the Figma prototype. Since Figma did not include a native text entry field, we resorted to third-party components and plug-ins to find a solution. However, many resources we found did not support multiple text entry fields in the same frame. We ultimately resolved the issue after extensive research by finding a third-party component designed specifically for login pages.

Through these challenges and many more, we learned the value of clear and timely communication between different parts of the team to solve problems quickly and create a cohesive final product.

Accomplishments that we're proud of

We are proud of how quickly we were able to learn and implement the Perplexity API given our lack of experience with it. The API integrated seamlessly with the rest of our project and played a critical role in realizing our overall vision. We are also proud of how we were able to create a cohesive set of cartoon images with DALL-E 3 for our Figma, despite the model’s tendency to ignore negative commands and generate inconsistent images. The final product was both aesthetically pleasing and representative of our brand and product.

What we learned

Some things that we learned were the importance of version control workflows, thinking outside of the box and finding alternative solutions to problems, and clear and timely communication between team members. These lessons were essential for us to create a successful project efficiently.

What's next for Delve

In the future, we plan to flesh out the Figma prototype and implement it in React. We also plan to incorporate a text-to-speech feature in the journaling section, develop a method to save journal entries, and enhance our model to create more content-specific and unique questions.

Built With

• css
• figma
• html
• javascript
• perplexity
• react