RAG-unittesting

This repository demonstrates a basic Retrieval-Augmented Generation (RAG) pipeline using Hugging Face models and the LlamaIndex library, along with comprehensive unit tests for all major components.

Features

• Loads documents from a Hugging Face dataset (e.g., SQuAD)
• Embeds documents and builds a vector index
• Retrieves relevant documents for a query
• Generates answers using a language model
• Includes robust unit and integration tests for the RAG workflow

Structure

• main.py — Example script to run the RAG pipeline
• rag_system.py — Core RAG implementation
• test_ragsystem.py — Unit and integration tests (using unittest and unittest.mock)
• requirements.txt — Python dependencies

Quick Start

1. Install dependencies:

pip install -r requirements.txt

2. Set your OpenAI API key in a .env file:

OPENAI_API_KEY=your_openai_api_key

3. Run the example:

python main.py

4. Run all tests:

python test_ragsystem.py

Notes

• The RAG system uses Hugging Face models for embeddings and LLMs, and OpenAI for LLM inference.
• All major RAG steps are covered by unit tests, including error handling.

---

For learning and experimentation.