Inspiration

Post-operative care is often confusing, isolating, and emotionally overwhelming — especially when patients are home without easy access to medical staff. We wanted to build a friendly, AI-powered companion that feels like a supportive guide through recovery — answering questions, easing anxiety, and making the experience more comfortable. That’s how HealBuddy was born.

What it does

HealBuddy is a multi-agent AI web platform that helps patients recover confidently after surgery. It includes: A symptom checker (Gemini agent) that provides safe, short, non-diagnostic suggestions A recovery chatbot (GPT-4) that gives encouragement, motivation, and self-care tips A profile form that stores patient recovery context in MongoDB A calm companion cat that reacts when clicked or fed — designed to reduce stress and lift mood Everything is available in one place, without needing to download any app.

How we built it

Frontend: HTML, CSS, JavaScript (with separate pages for each agent and the calming cat) Backend: Python Flask server with route-based agent dispatcher (/chat) AI Agents: Gemini (for safety-focused responses) and GPT-4 (for motivational support) Database: MongoDB Atlas for storing patient profile context Deployment: Backend hosted on Render; frontend deployed on GoDaddy

Challenges we ran into

Limiting the LLM output to short, helpful messages — avoiding overly clinical or verbose answers Coordinating two LLM platforms (Gemini + GPT) with a shared router Handling environment variables securely for deployment Ensuring UI worked across multiple pages without breaking interactions Syncing frontend integration with backend responses in real-time during the hackathon

Accomplishments that we're proud of

Created a working multi-agent platform in <24 hours using two LLM APIs Designed a calming and helpful UX with chatbot, charts, and interactive elements Personalized all LLM responses based on a user’s recovery profile Built a full-stack solution from scratch with database integration and deployed backend Built an adorable “talking tom” style cat that purrs,eats when clicked 💛

What we learned

How to coordinate multiple LLMs in one project How to craft effective prompts that stay concise, calm, and safe How to set up and query MongoDB in a Flask-based project Prompt design tricks to control hallucination and enforce tone Full deployment workflow across backend and frontend stacks

What's next for HealBuddy

📚 RAG integration: Build a knowledge base using vector embeddings from post-op PDFs and medical documents 🧠 Smarter agent routing based on condition, recovery day, and emotion 🌍 Multilingual support so non-English speakers feel equally supported 📈 Progress tracker that gives encouraging feedback based on usage patterns 🤝 Conversation history + doctor mode for real clinical escalation and handoff

Share this project:

Updates