RAG Chatbot with Google Docs Integration: Retrieval-Augmented Generation-based Intelligent Document Q&A System

This is a retrieval-augmented generation (RAG) chatbot that integrates Google Docs and local files (PDF/DOCX/TXT) to build a knowledge base. Key features include hybrid AI engines (Google Gemini for RAG-based document answers, GPT-4o-mini for general knowledge queries), WayneTech-style modern dark UI, and real-time document sync. It solves the disconnect between traditional RAG systems and cloud document services, providing a unified knowledge query solution for individuals and teams.

The project's core innovation lies in deep integration of RAG architecture with cloud document services. It supports:

• Local files: Text-based PDF, DOCX, TXT parsing and vectorization.
• Google Docs: Direct sync from Google Drive, real-time updates after document modification, and collaboration support via Google Docs' editing features. This combines local file flexibility with cloud collaboration capabilities.

Hybrid AI Engine:

• RAG mode (document-related queries): Uses Google Gemini to generate answers based on retrieved document fragments, reducing hallucinations.
• General mode (open knowledge queries): Uses GPT-4o-mini to leverage internal model knowledge.

Retrieval System:

• Embedding model: Google Gemini Embeddings.
• Vector database: FAISS (Facebook AI Similarity Search).
• Flow: User query → vectorization → FAISS search → top-K relevant fragments → context injection → answer generation.

Project Structure:

• App: Services (auth/pipeline/docs), routes, application factory.
• Static: Frontend resources (CSS/JS).
• Templates: HTML files.
• Docs: Documentation and release notes.
• Uploads: Local file storage.

Tech Stack:

Layer	Technology
Backend	Flask (Python)
Frontend	HTML+CSS+JS
Vector DB	FAISS
Embeddings	Google Gemini Embeddings
LLMs	Google Gemini (RAG), GPT-4o-mini (general)
Auth	Google OAuth2.0

Data Flows:

• Document intake: Upload → parse → chunk → vectorize → FAISS storage.
• Query processing: Query → vectorize → FAISS search → build prompt → generate answer → return with references.

Prerequisites: Python3.9+, OpenAI API key, Google Gemini API key, GCP credentials (for Google Docs integration). Setup:

1. Create .env file with keys (FLASK_SECRET_KEY, OPENAI_API_KEY, GEMINI_API_KEY).
2. Configure GCP: Enable Drive/Docs APIs, create OAuth2 credentials, download credentials.json. Quick Start:

• Install dependencies: pip install -r requirements.txt.
• Start service: python run.py (access http://localhost:8000).

Use Cases: Personal knowledge management, team collaboration knowledge base, customer support automation, research literature assistant, contract/policy query.

Technical Highlights:

• Hybrid model selection (auto/manual switch between RAG and general modes).
• Source traceability (answers include document references).
• Streaming response (Server-Sent Events for real-time output).
• Voice input (Web Speech API support).

Expansion:

• Document types: Markdown, HTML, EPUB, OCR for scanned PDFs, table parsing.
• Retrieval optimizations: Hybrid keyword+vector search, reranking, multi-language support.
• Features: Conversation history persistence, multi-user access control, analytics.

Comparison with Other RAG Systems:

Feature	This Project	Open-source RAG	Commercial RAG
Google Docs Integration	✅ Native	❌ Need custom dev	Partial
Hybrid AI Engine	✅ Dual models	Usually single	Usually single
Self-hosted	✅ Full control	✅ Full control	❌ Cloud-only

Summary: This is a complete, well-designed document QA system. It integrates RAG with Google Docs, uses hybrid AI to balance accuracy and versatility, and offers a modern UI. Ideal for personal/team knowledge management and a reference for RAG development.

Project Link: https://github.com/ankitrout07/RAG-Chatbot-with-Google-Docs-Integration (MIT License)