🧠 RAG Knowledge Assistant

A Retrieval-Augmented Generation (RAG) system that transforms your documents into an intelligent, queryable knowledge base — enabling accurate, context-aware answers grounded in your own data.

---

📌 Overview

The RAG Knowledge Assistant combines the power of large language models with semantic document retrieval to answer questions based on your custom knowledge sources. Unlike standard LLMs, this system grounds its responses in your actual documents — eliminating hallucinations and ensuring answers are always relevant and traceable.

Why RAG?

Standard LLM	RAG Knowledge Assistant
Answers from training data only	Answers from your documents
May hallucinate facts	Grounded in retrieved context
Static knowledge cutoff	Always up-to-date with your data
No source attribution	Cites source documents

---

🏗️ Architecture

                        ┌─────────────────────────────────────┐
                        │           RAG Pipeline              │
                        │                                     │
  Your Documents ──────►│  INGESTION                          │
  (PDF, TXT, etc.)      │  ├── Load & parse documents         │
                        │  ├── Chunk text                     │
                        │  └── Embed & store in vector DB     │
                        │            │                        │
  User Query ──────────►│  RETRIEVAL                          │
                        │  ├── Embed query                    │
                        │  ├── Similarity search              │
                        │  └── Fetch top-k relevant chunks    │
                        │            │                        │
                        │  API / LLM                          │
                        │  ├── Augment prompt with context    │
                        │  └── Generate grounded answer       │
                        └─────────────────────────────────────┘
                                      │
                               Answer + Sources

---

📁 Project Structure

rag-knowledge-assistant/
│
├── api/                    # REST API layer
│   ├── app.py              #   FastAPI / Flask application
│   ├── routes.py           #   Endpoint definitions
│   └── schemas.py          #   Request / response models
│
├── ingestion/              # Document ingestion pipeline
│   ├── loader.py           #   Document loaders (PDF, TXT, DOCX, web)
│   ├── chunker.py          #   Text chunking strategies
│   └── embedder.py         #   Embedding generation & vector storage
│
├── retrieval/              # Semantic retrieval engine
│   ├── retriever.py        #   Vector similarity search
│   ├── reranker.py         #   Result reranking
│   └── generator.py        #   LLM response generation with context
│
├── Dockerfile              # Docker image definition
├── docker-compose.yml      # Multi-service orchestration
├── requirements.txt        # Python dependencies
└── README.md

---

🛠️ Tech Stack

Layer	Technology
API	FastAPI / Flask
Embeddings	OpenAI text-embedding-ada-002 / HuggingFace
Vector Store	FAISS / ChromaDB / Pinecone
LLM	OpenAI GPT-4 / GPT-3.5
Document Parsing	LangChain / LlamaIndex
Containerization	Docker + Docker Compose
Language	Python 3.9+

---

🚀 Getting Started

Prerequisites

• Python 3.9+
• Docker & Docker Compose
• OpenAI API key (or compatible LLM provider)

1. Clone the Repository

git clone https://github.com/workgarimaswami/rag-knowledge-assistant.git
cd rag-knowledge-assistant

2. Set Up Environment Variables

cp .env.example .env

Edit .env:

OPENAI_API_KEY=your_openai_api_key_here
VECTOR_STORE=faiss          # faiss | chroma | pinecone
EMBEDDING_MODEL=text-embedding-ada-002
LLM_MODEL=gpt-4
CHUNK_SIZE=500
CHUNK_OVERLAP=50

3. Install Dependencies

pip install -r requirements.txt

4. Ingest Your Documents

python ingestion/loader.py --source ./documents/

5. Start the API

python api/app.py

API available at http://localhost:8000.

---

🐳 Running with Docker

# Build and start all services
docker-compose up --build

# Run in detached mode
docker-compose up -d

# Stop services
docker-compose down

Services started:

• api — REST API on port 8000
• vector-db — Vector store on port 6333 (if using ChromaDB)

---

🔌 API Reference

Ingest Documents

POST /ingest

{
  "source": "path/to/document.pdf",
  "metadata": { "category": "finance", "author": "Jane Doe" }
}

Ask a Question

POST /query

{
  "question": "What are the key findings from the Q3 report?",
  "top_k": 5
}

Response:

{
  "answer": "According to the Q3 report, revenue grew by 23% year-over-year...",
  "sources": [
    { "document": "q3_report.pdf", "page": 4, "relevance_score": 0.94 }
  ]
}

Health Check

GET /health → { "status": "ok", "documents_indexed": 142 }

---

🔧 Configuration

Parameter	Default	Description
CHUNK_SIZE	500	Token size per document chunk
CHUNK_OVERLAP	50	Overlap between consecutive chunks
TOP_K	5	Number of retrieved chunks per query
VECTOR_STORE	faiss	Backend vector store
LLM_MODEL	gpt-4	Language model for generation

---

📂 Supported Document Types

• 📄 PDF (.pdf)
• 📝 Plain text (.txt)
• 📋 Word documents (.docx)
• 🌐 Web URLs
• 📊 Markdown (.md)
• 🗂️ CSV / JSON

---

🧪 Running Tests

pytest tests/ -v

---

🗺️ Roadmap

• Multi-modal support (images, tables)
• Streaming responses
• Conversation memory / chat history
• Web UI / chat interface
• Support for multiple vector store backends
• Document versioning and update tracking

---

🤝 Contributing

1. Fork the repository
2. Create your feature branch (git checkout -b feature/your-feature)
3. Commit your changes (git commit -m 'Add your feature')
4. Push to the branch (git push origin feature/your-feature)
5. Open a Pull Request

---

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

---

👩‍💻 Author

Garima Swami · @workgarimaswami

---

⭐ If this project helped you build something awesome, give it a star!