ScholAr

Inspiration

I often ran into frustrations with existing chatbots:

• Limited file uploads hidden behind paywalls
  
• Long wait times before being able to upload again
  
• Answers that sometimes came from outside sources instead of my own documents
  

That made me ask: why not build my own chatbot—offline, private, and tailored for learning?

That’s how ScholAr was born: a study assistant that lives on your desktop, works offline, and goes beyond just Q&A by generating quizzes and flashcards.

What it does

ScholAr is an offline, privacy-first AI study assistant that helps students transform their documents into interactive learning tools.

• Q&A Chat: Ask questions about your uploaded files and get answers powered by RAG (Retrieval-Augmented Generation).
  
• Quiz Generation: Instantly turn study materials into quizzes.
  
• Flashcards: Create flashcards to memorize concepts more effectively.
  

All processing is local—your documents never leave your computer.

How I built it

• Backend

  • FastAPI (Python) for the API
    
  • LangChain + Ollama for the RAG pipeline and LLM inference
    
  • ChromaDB for vector storage and similarity search
    

• Frontend

  • Next.js + TypeScript for the UI
    
  • Electron to package into a desktop app
    
  • TailwindCSS for styling
    

Challenges I ran into

• Apple Gatekeeper restrictions: Distributing the app for free on macOS was a challenge because of Apple’s strict app signing requirements.
  
• Frontend learning curve: I had no prior experience with Next.js or Electron, so I had to quickly adapt and learn while building.
  

Accomplishments I am proud of

• Built a fully functional offline RAG system using LangChain + Ollama
  
• Designed and launched my first-ever frontend project with Next.js + Electron
  
• Created a cross-platform desktop app that respects user privacy
  

What I learned

• How to integrate multiple technologies (FastAPI, LangChain, Ollama, ChromaDB, Next.js, Electron) into a cohesive product
  
• How to design and build user-friendly UIs, despite starting with no frontend framework experience

What’s next for ScholAr

• History & Session Tracking: Let users revisit past chats, quizzes, and flashcards
  
• Model selection: Let users choose what model to use as they may not have a strong enough computer to use gpt-oss
• Multi-document support: Cross-reference multiple files in one study session
  
• Spaced repetition: Smarter flashcards for long-term memorization
  
• Mobile app: Companion app for studying on the go
  
• Collaboration features: Share study sessions with friends
  

Built With

• chromadb
• electron
• fastapi
• langchain
• nextjs
• ollama