Booky

Read with friends and share ideas.
Every person is a star, fueled by the inner wisdom found in books.
The Star Archive
Share your inner light while keeping the secrets of the story.
Navigating the depths of your inner star with K2 intelligence.
Explore deeper meanings with AI questions, safely orbiting around spoilers.
Keep track of every inhabitant in this universe.
Seek wisdom without crossing the event horizon.
AI-Powered Visual Recaps.

Inspiration

We've all had that moment — reading a book and wishing someone else had read the same passage so we could talk about it. Reading is deeply personal, yet the best conversations happen when we discover someone saw the same sentence completely differently.

We wanted to build a platform where reading becomes a shared experience across time and space — where you can highlight a line from The Great Gatsby at 2am and wake up to find your friend highlighted the same line, but for a completely different reason.

What it does

Booky is a social reading platform that turns solitary reading into a connected experience:

Highlight & Annotate — Mark passages that resonate with you and add personal notes
AI-Powered Reading Notes — K2 Think V2 analyzes your highlights alongside your friends' and generates structured comparisons: shared passages with different perspectives, unique observations only you caught, and what your friends noticed that you missed
Reading Style Analysis — AI identifies each reader's "lens" (philosophical, emotional, critical) and shows how different people approach the same text
Interactive Story Choices — Detroit: Become Human-style moral dilemmas at key moments, letting you see how your choices compare with other readers
Author Chat — Talk with AI personas of classic authors about their work
3D Planet Visualization — Each reader has a planet in a solar system, with orbital distance reflecting reading similarity between friends
Spoiler Detection — AI automatically flags content beyond your reading progress

How we built it

Frontend: Next.js 16 + React 19 + TypeScript, with Three.js for 3D planet visualization, D3 for constellation maps, and Framer Motion for animations
Backend: FastAPI (Python) deployed on GCP Cloud Run
Database: MongoDB Atlas for user data, highlights, reading progress, and generated notes
AI Engine: K2 Think V2 — powers reading notes generation, question generation, author chat, interactive choices, character descriptions, spoiler detection, and comic recap scene prompts
Image Generation: Google Vertex AI Imagen 3.0 for comic-style story recaps
Hosting: Vercel (frontend) + GCP Cloud Run (backend), custom domain at booky.ink

Challenges we ran into

Semantic highlight matching — Getting K2 to identify thematically similar (not identical) passages required careful prompt engineering and temperature tuning
3D performance — Rendering an interactive solar system with orbital mechanics while keeping the app responsive on mobile
Real-time social reading — Designing highlight visibility so friends' annotations enrich your experience without spoiling it
Deployment coordination — Wiring up frontend (Vercel), backend (Cloud Run), and database (MongoDB Atlas) across team members simultaneously under hackathon time pressure

Accomplishments that we're proud of

K2's reasoning capabilities turned out to be a perfect fit for literary analysis — the model genuinely "thinks through" why two readers highlighted differently, producing insights that surprised even us
The 3D planet solar system creates real emotional connection: watching a friend's planet orbit closer as your reading tastes align feels meaningful, not gimmicky
The full reading experience — from highlighting a passage, to seeing a friend's annotation on the same line, to generating a comparison note that surfaces the
tension between your perspectives — works end to end
We shipped a complete, deployed product at booky.ink with 8 classic books fully loaded and every AI feature functional

What we learned

Reasoning models like K2 Think V2 excel at literary analysis because they can reason through nuanced comparisons rather than just pattern-match
The social layer of reading — seeing how differently someone interprets the same text — is far more engaging than reading alone
Building with multiple AI capabilities (text analysis, image generation, persona chat) in one app requires thoughtful UX so it feels cohesive, not like a feature demo

What's next for Booky

Exposing K2's <think> reasoning process as a visible "AI thought journey" in reading notes — letting readers see how the AI connected their perspectives
Real-time collaborative reading sessions with live cursors and shared annotations
Book clubs with group reading notes and discussion threads
Expanding beyond classic literature to contemporary titles