⬡ DocuSphere AI

RAG-powered documentation assistant — ingest any PDF or URL, then chat with your docs using AI.

Built with FastAPI + LangChain + ChromaDB on the backend and React on the frontend.

---

🖥️ Demo

Sidebar — Ingest Docs	Chat — Ask Questions
Upload PDFs or paste URLs	Get grounded, sourced answers

---

🏗️ Architecture

docusphere/
├── backend/          # FastAPI + LangChain RAG pipeline
│   ├── main.py       # API endpoints + RAG logic
│   ├── run_server.py # Uvicorn server entry point
│   ├── requirements.txt
│   └── .env          # API keys (not committed)
│
└── frontend/         # React chat interface
    ├── src/
    │   ├── App.js    # Main UI component
    │   └── index.css # Tailwind CSS
    └── package.json

---

⚙️ How It Works

1. Ingest — User uploads a PDF or pastes a URL
2. Chunk — Text is split dynamically based on document size (150–1000 chars per chunk)
3. Embed — Chunks are embedded using all-mpnet-base-v2 (HuggingFace)
4. Store — Embeddings stored in ChromaDB (local persistent vector store)
5. Query — User asks a question → query is rewritten into keywords → similarity search → LLM answers using retrieved context
6. Respond — Answer returned with source excerpts and confidence badge

---

🚀 Getting Started

Prerequisites

• Python 3.11+
• Node.js 18+
• Groq API Key (free)

---

Backend Setup

# 1. Go to backend folder
cd backend

# 2. Create and activate virtual environment
python -m venv venv
venv\Scripts\activate        # Windows
source venv/bin/activate     # Mac/Linux

# 3. Install dependencies
pip install -r requirements.txt

# 4. Create .env file
GROQ_API_KEY=your_groq_api_key_here

# 5. Start server
python run_server.py

Backend runs at http://127.0.0.1:8000

---

Frontend Setup

# 1. Go to frontend folder
cd frontend

# 2. Install dependencies
npm install

# 3. Start dev server
npm start

Frontend runs at http://localhost:3000

---

🔌 API Endpoints

Method	Endpoint	Description
GET	/health	Server health check
POST	/ingest/pdf	Upload and ingest a PDF file
POST	/ingest/url	Ingest content from a URL
POST	/query	Ask a question against ingested docs
GET	/documents	List all ingested documents

---

🧠 Tech Stack

Backend

Tool	Purpose
FastAPI	REST API framework
LangChain	RAG pipeline orchestration
ChromaDB	Local vector database
HuggingFace (all-mpnet-base-v2)	Text embeddings
Groq (llama-3.1-8b-instant)	LLM for answer generation
pdfplumber	PDF text extraction
httpx + BeautifulSoup	URL scraping

Frontend

Tool	Purpose
React 18	UI framework
Axios	API calls
react-markdown	Render AI responses with formatting
react-dropzone	Drag-and-drop PDF upload
Tailwind CSS	Styling

---

✨ Features

• 📄 PDF Ingestion — Upload any text-based PDF
• 🔗 URL Ingestion — Scrape and ingest any public webpage
• 💬 Chat Interface — ChatGPT-style conversation UI
• 🔍 Query Rewriting — Questions auto-converted to search keywords for better retrieval
• 📎 Source Citations — Every answer shows which document it came from
• ✅ Confidence Badge — "Grounded" or "No match" shown per response
• ⚡ Dynamic Chunking — Chunk size auto-adjusts based on document length

---

🔒 Environment Variables

Create a .env file in the backend/ folder:

GROQ_API_KEY=your_groq_api_key_here

📌 Notes

• ChromaDB stores data locally in backend/my_local_db/ — this folder is excluded from git
• The embedding model (all-mpnet-base-v2) downloads automatically on first run (~420MB)
• Scanned/image PDFs are not supported — only text-based PDFs work

---

👨‍💻 Author

Adarsh Beriwala
GitHub