DOCUMIND

Deep Learning Final Project

IBM Docling • BARTpho & ViT5 • PhoBERT & XLM-RoBERTa • Sentence Transformers

---

DocuMind is an AI-powered personal notebook assistant designed to help students and researchers manage and summarize their documents efficiently using state-of-the-art Deep Learning models.

Get Started · AI Stack · Features · Docs

🌍 Vietnamese Version

---

Product Demo

Onboarding	Login	Register
Home Screen	Notebook List	Notebook Detail
AI Chat	Summary	Profile

And more features like Settings, Notifications...

---

Student information

Student ID	Full name	GitHub	Email
2001230791	Doan Tan Minh Tan	TanDoan1234	doanminhtan.dev@gmail.com

---

Purpose

DocuMind is an AI-powered personal notebook assistant designed to help students and researchers manage and summarize their documents efficiently. By leveraging advanced Deep Learning techniques, DocuMind allows users to upload documents (PDF, Docx) and receive high-quality summaries and context-aware insights, ensuring accurate information processing.

---

🧠 AI & Deep Learning Stack

The project applies state-of-the-art Deep Learning techniques to optimize Vietnamese document processing:

• Document Processing: Uses IBM Docling for advanced layout analysis and high-quality Markdown extraction from complex documents (PDF, Docx, Pptx).
• Large Language Models (LLM):
  • Summarization: Evaluated using both BARTpho and ViT5 for optimal performance.
  • Question Answering: Comparative implementation of PhoBERT and XLM-RoBERTa for context-aware extraction.

---

📂 Directory layout

DocuMind/
├── mobile/                      ← Flutter mobile application
├── backend/                     
│   ├── app/                     ← Application logic (API, Core, Models)
│   ├── processor/               ← AI Pipeline (Docling, Embedding, Summarizer)
│   ├── main.py                  ← FastAPI entry point
│   └── Dockerfile               ← Backend container definition
├── docs/                        ← Setup guides and documentation
├── tests/                       ← Backend & AI testing scripts
├── ai/                          ← Pre-trained models and training results
├── assets/                      ← Project assets (logos, demo screenshots)
├── docker-compose.yml           ← Service orchestration (Backend & DB)
├── pyproject.toml               ← Dependency management (uv)
└── .env                         ← Environment variables (DB, Keys)

---

🛠️ Development Setup

We provide two ways to set up the DocuMind development environment. Choose the one that best fits your workflow:

• 🚀 Docker Setup Guide (Recommended): Get up and running in minutes with a fully containerized environment (Backend + Database).
• 🔧 Local Setup Guide: Manual installation for those who want to run services directly on their machine.

Prerequisites

• Python 3.12+
• uv
• Flutter SDK (for mobile)

Backend Setup & Testing

1. Install dependencies and sync the environment:

   uv sync

2. Run AI validation tests (Models will be downloaded on first run):

   • Document Processing: uv run python tests/test_docling_processor.py
   • Semantic Similarity: uv run python tests/test_embedding_service.py
   • Summarization: uv run python tests/test_summarization.py
   • AI Question Answering: uv run python tests/test_qa.py

3. Start the main server:

   uv run python backend/main.py

Mobile Setup

1. Run the application:

   cd mobile
   flutter run

---

📚 Documentation

Getting Started

• 📖 Introduction - Learn what DocuMind offers.
• ⚡ Quick Start - Get up and running in 5 minutes.
• 🔧 Installation - Comprehensive setup guide.

User Guide

• 📱 Interface Overview - Understanding the layout.
• 📚 Notebooks - Organizing your research.
• ✍️ Summarization - Document summary features.
• 💬 AI Chat - AI conversations with your files.

Advanced Topics

• ⚙️ Document Processor - Technical deep-dive into document processing.
• 🤖 AI Models - AI model configuration and details.
• 📂 AI Directory - Understanding the AI research folder.
• 🚀 Deployment - Production deployment guides.

---