ManoRaga

• 

  Recommendation for User

Inspiration

Mental health challenges such as stress, anxiety, and emotional fatigue are increasingly common, especially among students and young professionals. While exploring ways to address this, I was inspired by the role of Indian classical music in emotional healing. In traditional musicology, ragas are designed to evoke and regulate specific emotional states and have been used for centuries as a form of mental and emotional therapy.

The idea behind ManoRaga was to combine this ancient knowledge with modern AI reasoning, creating a system that can recommend suitable ragas based on a person’s emotional state. The goal was to bridge neuromusicology and generative AI in a meaningful and accessible way.

What it does

ManoRaga is an AI-powered recommendation system that: Takes a user’s emotional input in natural language Identifies the dominant emotion Recommends the most suitable Indian classical raga Explains how that raga supports mental and emotional well-being The system ensures that all recommendations are grounded in a structured dataset of ragas and their emotional effects.

How we built it

Python for core logic Gemini API (Flash model) for emotion reasoning and explanation generation Pandas to manage a CSV-based raga dataset Jupyter Notebook for development and demonstration The workflow involves: Loading a CSV dataset of ragas and their emotional benefits Passing the dataset and user emotion as context to Gemini Constraining the model to select only from the provided ragas Generating a calm, supportive explanation tailored to the user This approach ensures accurate, explainable, and culturally grounded recommendations.

Challenges we ran into

API quota limitations when using larger Gemini models Model compatibility and deprecated SDK issues during setup Ensuring the AI did not hallucinate ragas outside the dataset Designing prompts that balance emotional sensitivity with structured reasoning These challenges were resolved by switching to Gemini Flash, optimizing API usage, and enforcing dataset-driven reasoning.

Accomplishments that we're proud of

Successfully integrated Gemini AI for real-time emotional reasoning Built a functional, end-to-end AI system under time constraints Combined traditional Indian music theory with modern AI Designed a solution that is lightweight, explainable, and scalable Adapted to technical limitations without compromising project goals

What we learned

How to work with Gemini models and quotas effectively The importance of adaptability when facing real-world constraints Prompt engineering for reasoning over structured data How AI can complement domain expertise instead of replacing it Practical problem-solving during a fast-paced hackathon environment

What's next for ManoRaga

Facial emotion detection using computer vision to automatically identify a user’s emotional state through facial expressions, reducing the need for manual input Integration of actual raga-based audio tracks so users can directly listen to recommended music within the application Real-time emotion tracking using camera input combined with AI-based sentiment analysis Personalization using user history, preferred ragas, and time-of-day awareness Integration with wearable devices (heart rate, stress indicators) for biofeedback-driven recommendations Deployment as a full web or mobile application with an intuitive user interface Collaboration with mental health professionals to validate and enhance therapeutic effectiveness

Built With

• api
• flash-latest
• gemini
• jupyter
• pandas
• python