MindConstellation

Graph View
Calendar View
Journal Entry View

Inspiration

Imagine this: It's 1:47 AM. You can’t sleep. Your mind is replaying conversations about something someone said earlier today. You open your notes app and type a few sentences. A week later, you forget what you wrote. A month later, you don’t remember how you felt. Your emotions exist in fragments, scattered, disconnected, and hard to understand in hindsight. Therapy is powerful, but not always accessible. Mental health apps often reduce reflection to mood sliders and streaks.

What It Does

MindConstellation takes your journal entries, analyzes the emotions within them, and maps them visually. Each entry becomes a node in a dynamic constellation, with related entries connected by shared themes or emotional patterns. Users can explore these connections to uncover recurring feelings, identify triggers, and track personal growth over time.

How We Built It

When a user makes a journal entry, it is first transcribed using Whisper. Our LLaMA 3.2 model then asks follow-up questions to better gauge the emotions within the user’s story. The user’s responses are analyzed by LLaMA 3.2 to generate insights about their feelings and experiences. These insights are passed to an All-MiniLM model to create embeddings for each entry, which are used to identify connections between entries. All data is stored in the database and sent back to the frontend, where the dynamic constellation visualization updates in real time.

Challenges We Ran Into

Accurately classifying the wide range of emotions expressed in journal entries, since subtle differences could change the meaning of a reflection.
Designing the dynamic constellation map so that each entry becomes a node and connections based on shared themes remain clear as the number of entries grows.
Integrating multiple AI models on the backend to process entries and respond in real time while maintaining performance, coordination, and user privacy.

Accomplishments We're Proud Of

We successfully created a functional, interactive AI journal that reveals meaningful emotional patterns. The constellation visualization makes abstract reflection tangible, and our AI accurately detects recurring themes across entries. Users can now gain insight into their emotional history in ways that traditional journaling cannot offer.

What We Learned

We learned how to integrate multiple technologies, speech transcription, LLMs, embeddings, and interactive visualizations, into a cohesive, functional system.
We also discovered the potential of AI to enhance journaling by helping users explore and understand their emotions in a more structured and insightful way.