Main introduction
book mind maps

My Journey at the Hackathon: Building BookMind 📚🧠

What Inspired Me

As a fan of intricate book series like Harry Potter or Game of Thrones, I’ve always struggled to keep track of complex character relationships and storylines. This sparked the idea for BookMind, an AI-powered tool designed to simplify narratives by visualizing character connections and answering relationship-based questions interactively.

The goal was to merge technology with storytelling, making it easier for readers to immerse themselves in even the most intricate plots.

What I Learned

This hackathon was an incredible learning opportunity where I gained new technical and collaborative skills:

RAG Systems: Integrated Retrieval-Augmented Generation (RAG) pipelines with FAISS to perform high-speed vector similarity searches.
Llama-stack Contributions: Extended the Llama3.2-3B-instruct model by implementing SambaNova Systems' inference API within a custom Docker container.
Team Collaboration: Working closely with Wonha Lee taught me how to balance creativity and execution under tight deadlines.

How We Built BookMind

Architecture Overview:
- Frontend: Developed a user-friendly interface for querying character relationships and visualizing mind maps.
- Backend: Built on a RAG pipeline with Llama3.2-3B and FAISS for fast, accurate results.
- Graphical Mind Maps: Designed a multi-hop reasoning system to extract relationships, format them in JSON, and render dynamic graphs.
Core Features:
- AI chatbot for deep-dive questions about character relationships.
- Auto-generated, graphized mind maps for visual storytelling.
- A framework for community contributions to enhance insights collaboratively.

Challenges We Faced

Critical Bug Under Pressure: We encountered a significant error in the relationship extraction logic just before the judging round. Debugging on the spot highlighted the importance of composure and teamwork.
Data Structuring for Visualizations: Formatting mind map data in JSON to handle complex relationships required a meticulous, modular approach.

Final Thoughts

This hackathon was a rewarding experience, allowing us to push the boundaries of what AI can achieve in storytelling. BookMind represents the intersection of creativity and cutting-edge technology, and I’m excited to see how it can evolve further.

If you’re curious about our work, check out BookMind on GitHub!