Basic RAG (Retrieval-Augmented Generation)

Overview

This project is a simple Python implementation of Retrieval-Augmented Generation (RAG).
It demonstrates how to:

• Load and split .txt documents
• Generate embeddings using OpenAI
• Store and query embeddings in ChromaDB
• Generate concise answers with context retrieved from documents

It was created to practice document retrieval, vector databases, and LLM-based question answering.

---

Tools & Libraries

The project uses the following Python libraries:

• openai → to generate embeddings and chat responses
• chromadb → as the vector database for storing and retrieving document chunks
• python-dotenv → to manage API keys securely in a .env file

---

Files in this Repository

• main.py → Entry point to run the pipeline and ask questions interactively
• config.py → Handles environment variables and configuration
• data_loader.py → Loads documents and splits them into chunks
• embeddings.py → Generates embeddings using OpenAI
• vector_store.py → Sets up and manages ChromaDB collection
• rag.py → RAG pipeline logic (retrieval + response generation)
• articles/ → Folder containing .txt files used as knowledge base
• requirements.txt → List of dependencies

---

Workflow

1. Load documents → split into chunks
2. Create embeddings for each chunk
3. Store embeddings in ChromaDB
4. User asks a question
5. Retrieve top matching chunks
6. LLM generates an answer using context