Turn dense research papers into bite‑size insights in seconds — then ask follow‑up questions powered by a tiny RAG system. Built for hackathon speed, demo‑ready polish, and real usefulness.

Preview (drop your own PDF/TXT!)

• Demo GIF/Screenshot (placeholder): docs/demo.gif

• ⚡ Instant value: upload → Summary, Key Points, ELI5, Action Items
• 🧠 Actually useful: ask questions grounded in your paper with RAG
• 🧭 Friendly UX: copy, download Markdown, quick “Learn More” links
• 🔧 Simple stack: Flask + React + Gemini; easy to run and extend

• Summary: clear 2–3 paragraph overview
• Key Points: concise bullets for skimming
• ELI5: explain like I’m 5, in plain language
• Action Items: concrete, imperative takeaways
• Learn More: YouTube/StackOverflow/GfG/Wikipedia links seeded from results
• Copy/Download: one‑click copy or export to Markdown
• RAG Chat: asks Gemini with top‑k chunks from your doc (+ optional refs)

• One‑time setup
  • Windows: powershell -ExecutionPolicy Bypass -File scripts/setup.ps1
  • macOS/Linux: bash scripts/setup.sh
• Start both
  • Windows: powershell -ExecutionPolicy Bypass -File scripts/start.ps1
  • macOS/Linux: bash scripts/start.sh
• Open the Vite URL → upload a PDF/TXT → read results → switch to Chat → ask questions

• Backend
  • Windows: cd backend && ./.venv/Scripts/Activate.ps1 && python app.py
  • macOS/Linux: cd backend && source .venv/bin/activate && python app.py
• Frontend
  • cd frontend && npm run dev

• Backend: create backend/.env → GEMINI_API_KEY=... (Google AI Studio key)
• Frontend (optional): frontend/.env → VITE_API_BASE_URL=http://127.0.0.1:5000

1. Upload a .pdf or .txt to /api/summarize (PDFs parsed via PyMuPDF)
   • Conservatively skip obvious cover/permission pages on the first 1–2 pages; fall back to all pages if filtering would remove everything
2. Call Gemini to produce strict‑JSON sections: summary, key_points, eli5, action_items
3. Chunk + embed your paper (text-embedding-004) into an in‑memory index; best‑effort fetch reference URLs/DOIs to extend the index
4. /api/ask retrieves the most relevant chunks and asks Gemini to answer using only those sources → answer + simple source tags

• Backend: Flask, flask‑cors, python‑dotenv, PyMuPDF, requests, Gemini (Google Generative Language API)
• Frontend: React + Vite

• Health: GET /api/health

  • curl http://127.0.0.1:5000/api/health

• Summarize: POST /api/summarize (multipart file)

  • curl -F file=@/path/paper.pdf http://127.0.0.1:5000/api/summarize
  • Returns JSON: summary, key_points[], eli5, action_items[], optionally references_used[]

• Ask: POST /api/ask (JSON)

  • curl -H "Content-Type: application/json" -d '{"question":"What is the main contribution?"}' http://127.0.0.1:5000/api/ask
  • Returns: { "answer": string, "sources": string[] }

• Upload → Parse (PyMuPDF/UTF‑8) → Chunk (+overlap) → Embed (Gemini) → In‑memory index
• Ask → Embed Q → Retrieve top‑k → Compose prompt with sources → Gemini → Answer + source tags

• Innovation: lightweight RAG + summarization pipeline with reference augmentation
• Impact: turns hours of reading into minutes; great for literature reviews
• Polish: fast setup scripts, clean UI, copy/download, friendly errors, fallback parsing

• “Could not summarize document”
  • Check the backend console. If a PDF has boilerplate on every page (e.g., some IEEE exports), we now fall back to parsing all pages
• 400 “Unsupported file type”
  • Only .pdf and .txt are supported by /api/summarize
• Missing GEMINI_API_KEY
  • You’ll still get a 200 with placeholder content; add a valid key for real summaries
• CORS or wrong URL
  • Frontend defaults to http://127.0.0.1:5000. Override via frontend/.env

• Snippet‑level citations with hover previews
• Chunk viewer + relevance scores
• Streaming chat responses
• Multi‑document workspaces and persistence

• Built with love for the Roo‑Code Hackathon