📚 Simple RAG Chatbot

A beginner-friendly Retrieval-Augmented Generation (RAG) chatbot that allows users to ask questions about information stored in a PDF document.

This project uses Gemini 3 Flash as the language model.

---

🚀 Technologies Used

• Python: Main programming language
• Streamlit: Creates the web interface
• PyPDF: Extracts text from PDF files
• Sentence Transformers: Creates text embeddings
• ChromaDB: Stores and searches embeddings
• Gemini 3 Flash: Generates answers
• python-dotenv: Loads the Gemini API key

---

🧠 What is RAG?

RAG (Retrieval-Augmented Generation) combines document retrieval with an AI language model.

Instead of asking Gemini to answer a question only from its general knowledge, this project first searches the user's document for relevant information.

---

🔍 How the System Works

1. Loading the Gemini API Key

• The Gemini API key is stored in an environment file rather than directly inside the Python source code.

• This helps protect sensitive credentials and prevents accidentally exposing the API key on GitHub.

2. Gemini Model

• The project uses Gemini 3 Flash Preview as the Large Language Model.

• Gemini is responsible for generating the final answer using the relevant information retrieved from the PDF.

3. Reading the PDF

• The PDF is processed using PyPDF.

• The system reads each page and extracts the available text.

• The process is:

    PDF → Pages → Extracted Text

• The extracted text is then passed to the next stage.

4. Splitting Text into Chunks

• Large documents are divided into smaller sections called chunks.

• The process is:

    Large Document → Full Text → Multiple Smaller Chunks

• Chunking makes it easier to search for specific information instead of processing the entire document for every question.

5. Creating Embeddings

• Each text chunk is converted into a numerical representation called an embedding using Sentence Transformers.

• Embeddings represent the semantic meaning of text.

• This allows the system to identify text that is similar in meaning, even when the exact words are different.

6. Storing Information in ChromaDB

• The text chunks and their embeddings are stored in ChromaDB, which acts as the vector database.

• ChromaDB allows the system to efficiently search for chunks that are semantically similar to the user's question.

• The stored information can be thought of as:

    Text Chunk → Embedding → Vector Database

7. Retrieving Relevant Information

• When the user asks a question, the question is converted into an embedding.

• The embedding is compared with the embeddings stored in ChromaDB.

• The system retrieves the most relevant chunks from the PDF.

• This is the Retrieval part of RAG.

8. Generating the Answer with Gemini

• The retrieved chunks are provided to Gemini together with the user's question.

• Gemini uses this context to generate the final response.

• The system instructs Gemini to answer using the retrieved document information.

• If the required information cannot be found in the retrieved context, the chatbot can indicate that the answer is not available in the document.

• This is the Generation part of RAG.

---

📚 Concepts Practiced

• Retrieval-Augmented Generation (RAG)
• Large Language Models (LLMs)
• Gemini API
• PDF text extraction
• Text chunking
• Embeddings
• Vector databases
• Semantic search
• Similarity search
• Context retrieval
• Prompt construction
• Streamlit
• Environment variables
• Python project structure

---

⭐ Project Goal

The goal of this project is to understand the fundamental concepts behind Retrieval-Augmented Generation by building a simple RAG application from scratch using Python, ChromaDB, Sentence Transformers, Streamlit, and Gemini.