Inspiration
SonarCare is an intelligent first point of contact that could provide reliable medical information, help users understand their symptoms, and guide them toward appropriate care when needed.
What it does
An AI-powered medical advice chatbot that provides intelligent healthcare guidance through natural conversation.
How I built it
Backend:
- LangGraph Multi-Agent System: Implemented sophisticated agent orchestration for specialized medical tasks
- Intent Detection Pipeline: Built intelligent routing system to determine user needs and select appropriate agents
- Firebase Integration: Used for session management, chat history, and user data persistence
- Socket.IO Streaming: Real-time response delivery for enhanced user experience
- Perplexity Sonar API: Integrated for accessing current medical research and information
- FastAPI Framework: RESTful API design with proper CORS configuration
Frontend:
- React with TypeScript: Type-safe component architecture
- Real-time Communication: Socket.IO client for streaming responses
- Responsive Design: Mobile-first approach with modern UI/UX principles
- Dynamic Title System: Intelligent conversation labeling based on content analysis
Challenges I ran into:
Performance Optimization: Initially, our intent detection system was taking 20+ seconds per response, making the chatbot feel sluggish compared to modern AI assistants. We had to analyze and optimize the agent selection pipeline.
Title Generation Accuracy: The original system generated generic titles like "New Chat" instead of meaningful descriptions. We completely rewrote the title generation algorithm to analyze actual conversation content dynamically.
Agent Orchestration Complexity: Balancing the sophistication of our multi-agent system with response speed required iterative refinement of the LangGraph implementation.
Medical Accuracy vs. Responsiveness: Ensuring medical information accuracy while maintaining conversational flow and appropriate disclaimers about professional medical care.
What's next for SonarCare
Implementing agentic workflows to optimize monotonous tasks in healthcare industry.
Built With
- fast-api
- firebase
- langgraph
- next.js
- python
- shadcn/ui
- socket.io
- sonar-api
- tailwind
- typescript

Log in or sign up for Devpost to join the conversation.