This uses a basic TF-IDF vectorizer for retrieval (implemented from scratch using NumPy, as we don't have scikit-learn). The "generation" part is mocked by concatenating the retrieved context into a simple response string—for a real-world system, you'd integrate an LLM (e.g., via API) to generate more natural answers based on the retrieved context.
To use this project:

Download the rag_agent_project.zip file.
Unzip it to a directory.
Install the required package: pip install numpy.
Run python main.py and follow the prompts to ask questions.