Savant is a research assistant platform built during HackCU that helps users explore papers through three connected surfaces:

• A FastAPI backend for PDF ingestion, retrieval, graph generation, and session storage
• A Next.js frontend for document chat, voice interaction, citations, and graph exploration
• A Chrome side panel extension for generating concept trees from research pages

HackCU_12--main/
|- savant-backend/     FastAPI API, PDF ingestion, retrieval, graph endpoints
|- savant-frontend/    Next.js research cockpit UI
|- savant-extension/   Chrome side panel extension for paper context trees
|- .gitignore

• Upload PDF papers and chunk them page by page
• Store document chunks in MongoDB Atlas with vector search support
• Query papers with Gemini-backed answers plus lexical fallback
• Generate concept graphs and paper use cases from document context
• Save chat sessions and share them across users
• Optional Solana payment gating before query execution
• Voice input plus ElevenLabs or browser TTS playback
• Chrome extension support for ResearchGate, arXiv, and Semantic Scholar

• Frontend: Next.js 16, React 19, TypeScript, Tailwind CSS
• Backend: FastAPI, Motor, PyPDF2, Gemini API, ElevenLabs
• Extension: Vite, React, TypeScript, Chrome Extensions Manifest V3
• Data: MongoDB Atlas vector search

• Node.js 20+
• npm
• Python 3.10+
• MongoDB Atlas database
• Gemini API key
• ElevenLabs API key for audio output
• Solana wallet and RPC endpoint if payment gating is enabled

Copy savant-backend/.env.example to savant-backend/.env and fill in the values.

Required backend variables:

• MONGODB_URI
• GEMINI_API_KEY
• ELEVENLABS_API_KEY
• ELEVENLABS_VOICE_ID
• SOLANA_RPC_URL
• MASTER_WALLET
• QUERY_PRICE_SOL
• REQUIRE_SOLANA_PAYMENT

Copy savant-frontend/.env.example to savant-frontend/.env.local.

Required frontend variables:

• NEXT_PUBLIC_API_BASE_URL
• NEXT_PUBLIC_SOLANA_RPC_URL
• NEXT_PUBLIC_MASTER_WALLET
• NEXT_PUBLIC_QUERY_PRICE_SOL
• NEXT_PUBLIC_REQUIRE_SOLANA_PAYMENT

The extension currently calls the backend at http://127.0.0.1:8000. Make sure the backend is running locally before testing the side panel.

Create a vector index on savant.documents named vector_index:

{
  "fields": [
    {
      "numDimensions": 768,
      "path": "embedding",
      "similarity": "cosine",
      "type": "vector"
    }
  ]
}

cd savant-backend
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt
uvicorn main:app --host 127.0.0.1 --port 8000 --reload

cd savant-frontend
npm install
npm run dev -- --hostname 127.0.0.1 --port 3000

cd savant-extension
npm install
npm run build

Then open chrome://extensions, enable Developer Mode, choose Load unpacked, and select the generated savant-extension/dist folder.

1. Start the backend
2. Start the frontend
3. Build and load the Chrome extension
4. Upload a PDF in the web app or open a supported paper page for the extension

• Root documentation added
• Frontend environment template added
• Frontend and extension documentation added
• Existing .gitignore already covers common Python, Node, build, and secret files

git init
git add .
git commit -m "Initial commit"
git branch -M main
git remote add origin <your-github-repo-url>
git push -u origin main

If this folder is already inside another Git repository, create the GitHub remote from that existing repo root instead of running git init again.

• savant-frontend/README.md now contains app-specific frontend setup details
• savant-extension/README.md documents how to build and load the browser extension
• No license file was added because license choice is a product decision and should match how you want others to use the code