MindNest – AI Micro-Journaling for Student Wellness

Inspiration

College students often face emotional highs and lows but lack the time, motivation, or privacy to maintain consistent wellness habits. Traditional journaling apps are often too time-consuming or generic. We wanted to create a tool that empowers students to check in with themselves quickly and meaningfully—while using AI to do the heavy lifting.

What it does

MindNest is a lightweight AI-powered web app that enables students to:

• Write or speak short daily reflections.
• Automatically detect emotions using text and speech-based AI.
• Visualize mood trends with an intuitive dashboard.
• Receive personalized wellness suggestions including music, quotes, and breathing routines.
• Export a weekly emotional wellness summary as a PDF.
• Store all data locally for complete privacy.

How we built it

We built the frontend using React, TypeScript, Vite, and Tailwind CSS. For emotion detection:

• We used RoBERTa, fine-tuned on the GoEmotions dataset, for text-based emotion classification.
• openSMILE was used to extract audio features for speech emotion recognition.
• A fusion engine combined the text and voice emotion predictions. Agentic responses and care suggestions were generated using Ollama (GPT4All/Mistral) via prompt templates. We used Recharts for mood graphs, jsPDF for PDF generation, and stored data in localStorage for privacy.

Challenges we ran into

• Merging multi-modal emotion predictions (text + voice) in a reliable way.
• Ensuring suggestions felt meaningful, not generic.
• Managing accurate sentiment detection on very short entries.
• Making the interface soothing, simple, and emotionally intuitive.
• Generating PDF reports with dynamic layouts for emotion summaries.

Accomplishments that we're proud of

• Built a fully working MVP in just a few days.
• Created a smooth journaling experience that respects user privacy.
• Developed a lightweight emotion detection pipeline that works offline.
• Delivered a user-friendly dashboard and care suggestions that feel genuinely helpful.

What we learned

• Prompt engineering is essential to guide meaningful AI outputs in wellness contexts.
• Multi-modal emotion detection can significantly enhance user understanding.
• Students respond better to tools that feel gentle and optional—not clinical or overwhelming.
• Offline-first apps (via localStorage and compact models) can still provide impactful AI experiences.

What's next for MindNest – AI Micro-Journaling for Student Wellness

• Integrate anonymous peer journaling exchanges for support.
• Expand to multilingual emotion recognition.
• Add a streak system and gamified wellness encouragement.
• Collaborate with university counseling centers to offer MindNest as a self-help tool.
• Deploy as a PWA for offline mobile use.

Built With

• date-fns
• gpt4all
• jspdf
• localstorage
• lucide-react
• ollama
• opensmile
• react
• recharts
• roberta-(huggingface-transformers)
• tailwind-css
• typescript
• vite