RAG-Based Data Retrieval with Ollama

This project demonstrates a Retrieval-Augmented Generation (RAG) system using the Ollama model for document-based question answering. The goal of the project is to integrate documents, retrieve relevant information, and generate concise answers using the Ollama Gemma 2B model.

Features

• Loads and processes news articles from a specified directory (news_articles).
• Splits documents into manageable chunks for efficient retrieval.
• Uses ChromaDB for vector storage and retrieval of document embeddings.
• Generates embeddings for documents using the Ollama Gemma 2B model.
• Queries the document database and generates answers using Ollama's conversational model.

Setup

1. Clone the repository:

   git clone https://github.com/your-username/RAG-Based-Data-Retrieval.git
   cd RAG-Based-Data-Retrieval

2. Create a virtual environment:

   python -m venv venv

3. Activate the virtual environment:

   • On Windows:

     venv\Scripts\activate

   • On macOS/Linux:

     source venv/bin/activate

4. Install required packages:

   pip install -r requirements.txt

5. Set up environment variables by creating a .env file in the root directory:

   CHROMA_PERSISTENT_STORAGE_PATH="chroma_persistent_storage"

6. Make sure you have access to Ollama for the embeddings and model queries.

Usage

1. Prepare a folder news_articles/ containing .txt files of articles you wish to process.

2. Run the application:

   python app.py

3. Query the system by providing a question. The system will fetch relevant document chunks, generate embeddings, and generate a response based on the retrieved context.

Code Explanation

1. Loading Documents: The load_documents_from_directory function loads all .txt files from a specified directory and stores their content as documents.
2. Text Chunking: The split_text function splits long documents into smaller, manageable chunks to allow for more accurate retrieval.
3. Embedding Generation: The get_ollama_embedding function generates embeddings for document chunks using the Gemma 2B model from Ollama.
4. Storage in ChromaDB: The embeddings are stored in ChromaDB for fast retrieval during queries.
5. Query Processing: The query_documents function generates an embedding for the user’s query and retrieves the most relevant document chunks. The system uses the retrieved chunks to generate an answer.
6. Answer Generation: The generate_response function combines the relevant context and generates concise answers using Ollama's conversational model.

Example Query

You can use this example to query the system:

question = "What is human life expectancy in the US and Bangladesh?"
relevant_chunks = query_documents(question)
answer = generate_response(question, relevant_chunks)
print(answer)