AGENTIC AI RAG FOR GOOGLE DRIVE STORAGE

---

Google Drive RAG Chatbot

📌 Overview

This application enables secure, read-only Retrieval-Augmented Generation (RAG) over a user’s Google Drive documents using open-source LLMs and vector search. It extracts, chunks, embeds, and indexes documents into ChromaDB, then answers user queries with grounded, context-aware responses — without modifying or storing user files.

✅ Only read access is requested (drive.readonly). 🔒 No file creation, modification, or deletion. 🌍 Designed for French-speaking users with LLM providers like Antropic or Mistral support.

---

🔧 Key Components

Layer	Technology
Auth & Data Access	Google OAuth2 + drive.readonly
Document Loader	LangChain Google Drive Loader
Embedding Model	Sentence Transformers / Mistral Embeddings
Vector DB	Chroma (local or cloud)
LLM	Claude Sonnet 3.7 and Mistral 7B (quantized, via OpenLLM or local inference)
Orchestration	LangChain + LangGraph
Frontend	Gradio
Observability	LangSmith (this option is not a free service)

---

⚙️ Setup & Configuration

1. Google Cloud Platform Setup

1. Create a Google Cloud Project.
2. Enable the Google Drive API.
3. Configure the OAuth Consent Screen:
   • User Type: External (for testing) or Internal (Workspace only)
   • Add your email as a Test User
   • Declare the OAuth scope: https://www.googleapis.com/auth/drive.readonly
4. Create OAuth 2.0 Client ID → download credentials.json. 📝 Important: For public deployment, Google verification is required for sensitive scopes.

2. Local Installation

git clone https://github.com/donat-konan33/google-drive-agentic-rag.git
cd google-drive-agentic-rag
poetry install

Place your credentials.json in the root or src/ directory.

---

3. Run the Pipeline

# Step 1: Ingest & index Google Drive documents
poetry run python -m src.googledriveagenticiarag.rag_pipeline

# Step 2: Launch the chat interface
poetry run python -m src.googledriveagenticiarag.main

The app will:

• Prompt for Google OAuth authentication.
• Fetch and chunk your Drive files.
• Generate embeddings and store them in ChromaDB.
• Launch a Gradio/Streamlit UI for querying.

🔐 Security & Compliance

• Minimal scope: Only drive.readonly is used.
• No data retention: Files are processed in-memory; only embeddings & metadata are stored locally.
• User consent: Explicit OAuth approval required.
• Private by default: Runs locally unless deployed to GCP.

Deployment Options

Environment	Notes
Local (Dev)	Uses local GPU/CPU via OpenLLM or LM Studio
GCP (Prod)	Deploy with Vertex AI, Cloud Run, or Compute Engine
ChromaDB Cloud	Optional managed vector DB (SOC 2 compliant)

---

---

Credits: Inspired by @khuyentran1401's RAG pipeline tutorial (CodeCut AI).