RAGLab

Interactive RAG experimentation lab — compare retrieval strategies and generation modes side-by-side on real documents.

Demo / portfolio project. Run it locally to explore how different RAG pipelines behave.

---

What it does

Pick a document, choose how to chunk/embed/retrieve, pick a RAG mode, and ask questions. Streaming answers plus retrieval insights show how each configuration changes quality.

Area	Options
RAG modes	Basic · Self-RAG · Agentic RAG · Graph RAG
Retrieval	Vector (FAISS) · Hybrid (BM25 + RRF) · Multi-query · Cross-encoder rerank
Chunking	Recursive · Fixed · Semantic · Structure-aware
Embeddings	OpenAI text-embedding-3-small · Local SentenceTransformer

Also included: knowledge-graph extraction, contextual compression, guided UI walkthrough, and SSE token streaming.

---

Tech stack

Layer	Stack
Backend	Python, FastAPI, LangChain, FAISS, NetworkX, BM25
Frontend	Next.js, React, TypeScript, Tailwind CSS
Models	Gemini (generation) · OpenAI / local (embeddings) · MiniLM (rerank)

---

Project structure

RAGLab/
├── .github/workflows/ci.yml   # lint, typecheck, build, security
├── Makefile                   # setup / run / test
├── raglab-backend/
│   ├── main.py                # FastAPI app
│   ├── config.py              # env config
│   ├── document_store.py      # load + index documents
│   ├── api/                   # REST + SSE routes
│   ├── chunking/              # text + structure chunkers
│   ├── embeddings/            # embedding providers
│   ├── graph/                 # knowledge graph + graph retrieval
│   ├── ingestion/             # PDF parsing
│   ├── middleware/            # error logging
│   ├── retrieval/             # vector / hybrid / multi-query / rerank
│   ├── services/              # RAG generation + compression
│   ├── documents/             # sample PDFs
│   └── verify_system.py       # import / config smoke check
└── raglab-frontend/
    ├── app/                   # Next.js App Router
    ├── components/            # UI
    └── lib/                   # API client

---

Quick start

Prerequisites

• Python 3.11+
• Node.js 18+
• OpenAI API key (embeddings)
• Google Gemini API key (generation)

Setup

git clone https://github.com/guptaom31619-prog/RagLab.git
cd RagLab

make setup

Edit raglab-backend/.env:

OPENAI_API_KEY=sk-...
GEMINI_API_KEY=...

Run

make dev

Service	URL
Frontend	http://localhost:3000
Backend	http://localhost:8000
API docs	http://localhost:8000/docs

Verify

make test    # import + config smoke check
make lint    # frontend lint
make build   # frontend production build

---

Make commands

Command	Description
make setup	Install deps + create .env
make dev	Run backend + frontend
make stop	Stop local processes
make test	Backend smoke check
make lint / make build	Frontend quality checks
make clean	Remove __pycache__ / .next
make clean-cache	Wipe document + KG caches

---

API (overview)

Method	Endpoint	Description
GET	/health	Health check
GET	/documents	List sample documents
POST	/load-document	Index a document with chosen config
POST	/ask	Full answer
POST	/ask-stream	SSE streaming answer

Example /ask body:

{
  "question": "What was Microsoft's total revenue?",
  "search_method": "hybrid",
  "use_compression": false,
  "rag_mode": "basic"
}

Interactive docs: /docs when the backend is running.

---

Environment

Variable	Required	Description
OPENAI_API_KEY	Yes	Embeddings
GEMINI_API_KEY	Yes	Answer generation
NEXT_PUBLIC_API_URL	No	Frontend → backend URL (default http://localhost:8000)

Never commit .env. Use .env.example as the template.

---

CI

GitHub Actions runs on every push / PR to main:

• Backend — Python syntax, structure, requirements check
• Frontend — tsc, ESLint, production build
• Security — no committed .env / hardcoded API keys

---

License

MIT — see LICENSE.