🧠 Simple RAG System

📘 Overview

This repository contains a minimal Retrieval-Augmented Generation (RAG) system built with Hugging Face Transformers and a curated dataset of 61 factual statements about Berlin. The system enables users to ask natural language questions and receive contextually relevant answers grounded in the dataset.

🔍 What is RAG?

Retrieval-Augmented Generation (RAG) is a hybrid approach that combines:

• Retrieval: fetching relevant documents from a knowledge base (in this case, a list of Berlin facts).
• Generation: using a language model to generate answers based on the retrieved context.

This notebook demonstrates a lightweight RAG pipeline using Hugging Face's pipeline and rag-token-base model to answer questions about Berlin.

---

📂 Contents

• documents: A Python list of 61 factual strings about Berlin.
• RAG pipeline setup using Hugging Face Transformers.
• Example queries and generated answers.
• Optional: similarity scoring and document ranking (if implemented).

---

⚙️ How It Works

1. Dataset: A list of 61 factual sentences about Berlin is stored in the documents variable.
2. Indexing: The dataset is passed to a retriever (e.g., DPRQuestionEncoder) to build a vector index.
3. Querying: A user question is encoded and matched against the document embeddings to retrieve top-k relevant facts.
4. Answer Generation: The retrieved facts are passed to a generative model (facebook/rag-token-base) to produce a final answer.

---

🚀 Example Usage

print(rag("What happens in Munich?"))

Ach, so you vant to know about Munich, ja? Hmm… zis document… it speaks only of… Berlin!

It tells me Berlin has a Tempodrom, a… how you say… event hall. Und a film festival, ze Berlinale! A very important one, yes. Zere is also… nightlife, many bars und clubs. Und theaters, like ze Berliner Ensemble und ze Volksbühne.

But about Munich… I do not know! Zis document is all about Berlin.

---

📊 Dataset Summary

Attribute	Value
Total Facts	61
Format	Python list of strings
Language	English
Domain	Berlin, Germany

---

💡 Use Cases

• Educational chatbots about Berlin
• Lightweight RAG experimentation
• Semantic search demos
• NLP fine-tuning with domain-specific corpora

---

📄 License

This project is provided for educational and research purposes. If you use or modify this code or dataset, please provide appropriate attribution.

---