This is a submission for the Built with Google Gemini: Writing Challenge
What I Built with Google Gemini
I built Serenity, a comprehensive, privacy-first mental health and wellbeing application designed to act as your personal caretaker, health coach, and daily companion.
Mental health tools are often fragmented—you use one app for habit tracking, another for journaling, and yet another for meditation. Serenity solves this by combining all these tools into one cohesive, beautifully designed platform. It includes:
- Mood Check-ins & Insights: Track daily emotional trends.
- Daily Journal: A safe space for reflection.
- Habit & Goal Tracker: Build positive routines and set long-term targets.
- Wellness Hub: Interactive exercises including Box Breathing, 5-4-3-2-1 Grounding, Body Scan meditations, and Eye Strain Relief.
The Role of Google Gemini:
Google Gemini (gemini-3-flash-preview) acts as the core "brain" and empathetic heart of the app. Instead of just a generic chatbot, Serenity uses the @google/genai SDK to read your recent moods, incomplete habits, journal entries, and long-term goals as context.
Gemini provides highly tailored advice, health tips, and emotional support. For example, if you set a goal to "Run a 5K", you can click "Ask Serenity for a Plan", and Gemini will instantly generate a personalized diet, exercise, and lifestyle plan based on your current habits and mood history. It also proactively generates warm, encouraging reminders if you forget to complete your daily habits!
Demo
You can try out the live version of Serenity here:
🔗 Live Demo
🔗 Github Repo
(Note: The app is designed with a "Warm Organic" aesthetic using Tailwind CSS and Framer Motion to create a calming user experience.)
What I Learned
Building Serenity was an incredible learning experience:
- Contextual AI Engineering: I learned how to effectively pass structured JSON data (user moods, habits, journals) into Gemini's system prompt to create an AI that truly "remembers" and understands the user's current state.
- UI/UX for Mental Health: I learned the importance of color psychology and micro-interactions. Using glassmorphism, soft sage/earth tones, and smooth Framer Motion animations helped create a UI that actually lowers stress rather than adding to it.
- Interactive Component Design: Building the Wellness Hub taught me how to manage complex state and timers in React, especially for the Body Scan and Eye Strain Relief exercises.
Google Gemini Feedback
What worked well:
The new @google/genai SDK is incredibly intuitive. The speed of the gemini-3-flash-preview model is phenomenal—it makes the conversational assistant feel truly real-time. I was also deeply impressed by how well Gemini understood the structured context I passed to it. It seamlessly connected the dots between a user feeling "anxious" in their mood log and suggesting the built-in grounding exercises.
Where I ran into friction:
The main challenge was prompt engineering to get the tone exactly right. Initially, the AI sounded a bit too clinical or robotic. I had to refine the system instructions heavily to ensure it consistently spoke with the warmth, empathy, and brevity of a "personal caretaker." Additionally, handling the multimodal inputs (allowing users to upload images to the chat) required some careful base64 encoding logic to pass to the API correctly, but once set up, it worked flawlessly.
Built with ❤️ for the Hack for Mental Health 2026 & DEV Challenge.
Top comments (2)
Beautiful UI and concept, Sanjay! It looks really good. Getting the AI tone to sound like a warm caretaker instead of a clinical robot sounds kind of tricky, but feeding it structured JSON context like daily moods is a very smart way to solve that
Thanks a lot! I really appreciate that.
Yes, getting the AI to feel more like a supportive companion rather than a clinical assistant was one of the biggest challenges. The structured JSON context (like daily moods and activities) helped a lot in guiding the tone and responses.