ResearchMate

Chat with your research papers, compare them side by side, and surface the gaps nobody's addressed yet.

ResearchMate is an AI-powered research workspace built on retrieval-augmented generation (RAG). Upload a stack of PDFs and it indexes them into a searchable knowledge base you can query, compare, and interrogate — with every answer grounded in your actual sources.

Table of Contents

• Features
• How It Works
• Tech Stack
• Project Structure
• Quickstart
• Environment Variables
• API Endpoints

Features

• Chat with your papers — ask natural-language questions, get answers grounded in your uploaded collection with cited sources
• Paper Comparison — generate a structured side-by-side comparison (summary, methodology, strengths) across multiple papers
• Research Gap Discovery — surface unexplored directions, open problems, and future work suggestions from your collection
• Guardrails — similarity-threshold validation keeps responses grounded in the uploaded material instead of hallucinating

How It Works

Each uploaded PDF moves through a RAG pipeline before it's queryable:

Upload → Extract → Chunk → Embed → Store (ChromaDB)
                                        ↓
                    Query → Retrieve → Rerank → Generate (Gemini) → Cite

1. Extract — text is pulled from the PDF (PyMuPDF)
2. Chunk — text is split into overlapping segments for retrieval
3. Embed — chunks are vectorized (sentence-transformers) and stored in ChromaDB
4. Retrieve & Rerank — relevant chunks are pulled for a query and reranked for relevance
5. Generate — Gemini produces an answer grounded in the retrieved chunks, with citations back to source papers

Tech Stack

Frontend	React · TypeScript · Vite · Tailwind CSS · TanStack Query · React Router
Backend	FastAPI · ChromaDB · sentence-transformers · PyMuPDF · Google Gemini
Infra	Docker · Docker Compose

Project Structure

ResearchMate/
├── frontend/                   # React + TypeScript + Vite app
│   └── src/
│       ├── api/                 # API client functions
│       ├── features/            # Feature modules (chat, comparison-table, pdf-upload, research-gap)
│       ├── pages/                # Route-level pages
│       └── components/           # Shared UI and layout components
├── backend/                    # FastAPI app
│   └── app/
│       ├── api/v1/endpoints/     # Route handlers
│       ├── modules/               # RAG, comparison, research-gap, citation, guardrails logic
│       ├── infrastructure/        # LLM provider integration
│       └── core/                   # Settings and logging
├── docker-compose.yml
├── Dockerfile.backend
└── Dockerfile.frontend

Quickstart

Prerequisites: Node.js 20+, Python 3.11+, a Google Gemini API key

git clone https://github.com/<your-username>/ResearchMate.git
cd ResearchMate

# Backend
cd backend
python -m venv venv && source venv/bin/activate   # Windows: venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env    # then add your GEMINI_API_KEY
uvicorn app.main:app --reload

# Frontend (in a new terminal)
cd frontend
npm install
cp .env.example .env
npm run dev

• Backend: http://localhost:8000 (interactive docs at /docs)
• Frontend: http://localhost:5173

Or with Docker:

docker-compose up --build

Environment Variables

backend/.env

Variable	Description	Default
ENVIRONMENT	Application environment mode	development
LOG_LEVEL	Logging verbosity	INFO
CORS_ORIGINS	Allowed frontend origins	["http://localhost:5173"]
GEMINI_API_KEY	Google Gemini API key (required)	—
GEMINI_MODEL	Gemini model identifier	gemini-2.5-flash-lite

frontend/.env

Variable	Description	Default
VITE_API_BASE_URL	Base URL for the backend API	http://localhost:8000/api/v1

API Endpoints

All endpoints are prefixed with /api/v1. Full interactive docs (Swagger UI) are available at /docs when the backend is running.

Method	Endpoint	Description
GET	/health	Health check
POST	/upload	Upload PDF files
POST	/extract	Extract text from uploaded PDFs
POST	/chunk	Chunk extracted text
POST	/embed	Generate embeddings for chunks
POST	/retrieve	Retrieve relevant chunks for a query
POST	/rerank	Rerank retrieved chunks
POST	/chat	Chat with the active paper collection
POST	/citations	Generate citations for a response
POST	/comparison	Compare papers in the active collection
POST	/research-gap	Identify research gaps and future work
POST	/guardrails	Validate a query against guardrail thresholds