RAG method on Python

Description

Retrieval-Augmented Generation (RAG) is a framework developed to enhance the accuracy and currency of large language models (LLMs)

In this project, I implemented the RAG method using Python and a local LLM running with LM Studio.

Installation

1. Clone the repository:

   git clone https://github.com/Ivan-Inby/python-rag.git
   cd python-rag

2. Create a virtual environment and activate it:

    python -m venv venv
    source venv/bin/activate  # For Windows use `venv\Scripts\activate`

3. Install dependencies:

    pip install -r requirements.txt

To run the application, you need to deploy a local server with LLM using LM Studio. You can adapt the code for use with another LLM, for example ChatGPT.

Preparing data

1. Place your PDF files in the data folder inside the project.

2. Run the script to create the database:

   python create_database_pdf.py

   This script will extract text from PDF files, break it into chunks, and save it into a ChromaDB database along with metadata.

Searching and generating answers

1. Run a script to search and generate answers based on the question entered:

   python ask.py “Your question is here”

   For example:

   python ask.py “What is Wandering Monster?”

   The script will search the database, create a context based on the found data and send it to the language model to generate a response.

Project structure

• create_database_pdf.py: Script for creating a database from PDF files.
• ask.py: Script to search the database and generate answers.
• data/: A folder to store PDF files.
• knowledge_base/: A folder to store the ChromaDB database.
• requirements.txt: File with project dependencies.
• README.md: Project description.

Example of use

1. Place the PDF files in the data folder.
2. Run create_database_pdf.py to create a database.
3. Use ask.py to search and generate answers.

Dependencies

• sentence-transformers
• chromadb
• openai
• langchain-community
• pypdf

License

This project is licensed under the terms of the MIT License.

Contacts

For questions and suggestions, please contact TG: @spider_lolo.