NodeTree

What it does

Leveraging AI and decision tree to answer questions in a tree-like structure. Instead of just giving a single reply like ChatGPT, it creates expandable answer branches with autoscaling, so users can dive deeper into specific parts of the response. This way, users can explore different directions and get a more complete understanding of the original question or topic and lead to a more complete answer with help of RAG.

How we built it

Data Integration:
- We connected to APIs and used web scraping to aggregate high-quality healthcare data.
Retrieval-Augmented Generation (RAG) System:
- We implemented a retrieval system to fetch relevant documents from our indexed database.
- The frontend provides a clean, user-friendly interface for inputting symptoms and displaying results through nodetree using Next.js and React Flow to build mindmap-like responses.

Challenges we ran into

Data Processing Issues: Parsing large datasets efficiently required optimization.
Latency Optimization: Improving response times while running complex retrieval and generation models.

Accomplishments that we're proud of

Building a scalable and efficient RAG model.
Overcoming challenges in real-time retrieval and response generation while maintaining performance.

What we learned

Construct AI-response in Tree Structure using React Flow

What's next for NodeTree

User Feedback & Validation: Work with domain professionals to validate model outputs and fine-tuning with QLoRA.
Real-time Personalization: Allow users to input history for personalized insights.
Local file repository: Transfer users' Obsidian markdown files to NodeTree.