Drive-RAG

Google Drive Connected Retrieval-Augmented Generation (RAG) System.

Drive-RAG is a modular RAG application that connects directly to a shared Google Drive folder or file, processes documents, stores embeddings in a vector database, and enables conversational question answering over your own documents.

The goal of this project is to provide a simple yet extensible RAG framework that allows users to:

• Ingest documents directly from Google Drive
• Automatically chunk and embed documents
• Store embeddings in a vector database
• Perform retrieval-augmented question answering with streaming responses
• Inspect retrieved context alongside generated answers

The system is designed to be modular and extensible, making it easy to experiment with different databases, embedding models, and LLM providers.

---

🚀 Installation

This project uses uv for dependency and environment management.

1️⃣ Install uv (if not installed)

curl -Ls https://astral.sh/uv/install.sh | sh

2️⃣ Create the virtual environment and install dependencies

uv sync

This will:

• Create a virtual environment
• Install all project dependencies

---

▶️ Running the Project

Start the application with:

python src/main.py

The Gradio UI will open in your browser.

---

🧠 How It Works

1. Provide a shared Google Drive link.

2. The system:

   • Downloads documents
   • Loads and parses them
   • Splits them into chunks
   • Generates embeddings
   • Stores them in a vector database (ChromaDB by default)

3. Ask questions in the chat interface.

4. The system retrieves relevant chunks and streams LLM-generated answers.

5. Retrieved context is displayed alongside the response.

---

🏗 Architecture Overview

• Downloader → Google Drive integration
• Loader → Document parsing
• Chunker → Text segmentation
• Embedder → Embedding model abstraction
• Vector DB → ChromaDB
• LLM Interface → OpenAI SDK
• UI → Gradio

The architecture is modular and designed for easy extension.

---

📌 TODO

• Add support for additional vector databases (e.g., Neo4j)
• Add support for multiple LLM providers/models
• Add a test UI that:
  • Accepts a test.json file
  • Runs RAG queries
  • Calculates retrieval and generation metrics