RAG: Retrieval-Augmented Generation with LLMs

Introduction

RAG is a project aimed at developing a Retrieval-Augmented Generation system utilizing Large Language Models (LLMs) and data ingested from documents.

Tech Stack

• Python 3.11: Programming language used for development.
• Poetry: Dependency management and virtual environment configuration.
• FastAPI: Framework for building REST APIs.
• Uvicorn: ASGI server powering FastAPI for serving HTTP requests.
• LangChain: Chain orchestration tool.
• Pydantic: Data validation and settings management using Python type annotations.

Installation

Local Installation

This project uses Poetry for dependency management and virtual environment configuration. You can use your preferred Python environment setup easily, like conda or venv, and poetry will use your current environment. If no environmet is found, poetry will create a .venv folder in this project's root folder.

Prerequisites

• Latest version of Poetry: pipx install poetry
• Ensure Python 3.11 is available in your $PATH
• Configure environment variables using an .env file (refer to example.env).
  • Missing fields may result in validation errors

Steps

1. Clone the repository and navigate to the project directory.
2. Run poetry install to install project dependencies.
3. Set up your environment variables in the .env file.
4. Start the application with make run, which runs poetry run python -m app.

Docker Installation

Run docker compose up -d

• Note: This may take several minutes, as the current build uses local embedding models that require a lot of dependencies.

While running container, you can upload a Markdown file at the data volume.

About

This project has been strongly influenced and supported by other amazing projects like PrivateGPT and LangChain.