BetterDoc - AI-Powered Word Document Enhancement Tool

BetterDoc is a Microsoft Office Word add-in that leverages AI technology to intelligently reimagine and improve your document content while preserving formatting, hyperlinks, and document structure.

Features

Core Functionality

• Paragraph-by-Paragraph Processing: Navigate through your document paragraph by paragraph with intelligent skipping of short paragraphs, questions, and empty content
• AI Content Enhancement: Use Azure AI Foundry or local Ollama models to rewrite content with customizable prompts
• Formatting Preservation: Automatically detects and preserves text formatting (bold, italic, underline, font sizes)
• Hyperlink Protection: Detects hyperlinks and ensures they remain intact during content processing
• Flexible Model Support: Choose between cloud-based Azure models or local Ollama models

Smart Processing

• Selective Processing: Only processes paragraphs with 7+ words, skips questions and empty paragraphs
• Word Count Control: Prevents AI from increasing word count beyond original content
• Retry Logic: Automatically retries if hyperlink text is lost during processing
• Batch Operations: Select and process multiple paragraphs simultaneously

User Interface

• Visual Navigation: Up/Down arrow buttons to move through paragraphs
• Real-time Preview: View original and modified text side-by-side
• Document Highlighting: Current paragraph is selected in the document for easy reference
• Model Selection: Choose from Azure AI Foundry or local Ollama models

Architecture

The project consists of two main components:

1. Office Add-in (betterdoc-office-addin/)

• Frontend: HTML/CSS/JavaScript task pane interface
• Backend Integration: Communicates with AI services via proxy server
• Office.js Integration: Uses Microsoft Office JavaScript API for document manipulation

2. Multi-Model Proxy Server (ollama-service/)

• Azure AI Foundry Integration: Secure connection to Azure AI Foundry using Python MSAL authentication
• Ollama Integration: Optional local model support via Ollama API
• Smart Routing: Automatically routes requests to Azure or Ollama based on model selection
• CORS Handling: Enables cross-origin requests from the Office add-in

Quick Start

Prerequisites

• Microsoft Word (Office 365 or Office 2019+)
• Node.js (18+)
• Python 3.7+ with msal and requests packages
• Azure AI Foundry access (for cloud models)
• Ollama (optional, for local models)

1. Clone and Setup

git clone <repository-url>
cd betterdoc2
cp confidential_sample.py confidential.py

2. Configure Azure AI Foundry

Edit confidential.py with your Azure details:

CLIENT_ID = "your-azure-ad-client-id"
TENANT_ID = "your-azure-ad-tenant-id"
API_BASE = "https://your-endpoint.azure-api.net/api/projects/your-project/models"
API_VERSION = "2024-05-01-preview"
DEFAULT_MODEL = "gpt-4.1"

3. Install Dependencies

# Python dependencies
pip install msal requests

# Node.js dependencies
cd ollama-service
npm install

4. Start the Server

cd ollama-service
node server.js

5. Install Office Add-in

cd betterdoc-office-addin/BetterdocAddin
npm install
npm run build

Install Development Certificates (Required for HTTPS):

# Navigate to the betterdoc directory
cd betterdoc2

# Install the Office Add-in development certificates
npx office-addin-dev-certs install

This will:

• Generate SSL certificates in %USERPROFILE%\.office-addin-dev-certs\
• Install the CA certificate for trusted access to https://localhost
• Create the required localhost.crt and localhost.key files

Then sideload in Word:

• Insert > My Add-ins > Upload My Add-in
• Select manifest.xml

Configuration Details

Azure AI Foundry Setup

1. Azure AD Application Registration:

   • Create an app registration in Azure AD
   • Note the CLIENT_ID and TENANT_ID
   • Configure API permissions for Azure AI Foundry

2. Azure AI Foundry Project:

   • Create an Azure AI Foundry project
   • Get the API endpoint URL
   • Ensure your account has access to the project

3. Authentication:

   • Uses Python MSAL for interactive authentication
   • Tokens are cached automatically for future use
   • One-time browser authentication setup

Local Models (Optional)

1. Install Ollama:

   # Visit https://ollama.ai for installation instructions

2. Pull Models:

   ollama pull gemma3:12b
   ollama pull qwq:latest
   ollama pull deepseek-r1:14b

3. Start Ollama:

   ollama serve

Available Models

Azure Models

• gpt-4.1 (Azure) - Powered by Azure AI Foundry (no local setup required)

Local Models (Requires Ollama)

• gemma3:12b - Google's Gemma model (12B parameters)
• gemma3:27b - Google's Gemma model (27B parameters)
• qwq:latest - QwQ reasoning model
• deepseek-r1:14b - DeepSeek R1 model (14B parameters)
• deepseek-r1:32b - DeepSeek R1 model (32B parameters)

Usage Guide

Basic Workflow

1. Open the Add-in: In Word, find "Betterdoc" in your add-ins and open the task pane

2. First-Time Setup:

   • Select an Azure model for immediate use
   • For local models, ensure Ollama is installed and running

3. Navigate Document: Use the ↑/↓ buttons to move through paragraphs

   • Automatically skips short paragraphs (<7 words), questions, and empty content
   • Current paragraph is highlighted in the document

4. Select Content: Check the "Reimagine" checkbox for paragraphs you want to process

   • Use "Select All" to mark all valid paragraphs
   • Use "Unselect All" to clear selections

5. Configure Settings:

   • Model Selection: Choose between Azure AI Foundry or local Ollama models
   • Prompt Selection: Choose from predefined prompts or create custom ones
     • "Make More Professional": Enhances formal tone
     • "Reduce Word Count": Makes content more concise
     • "Custom Prompt": Write your own instructions

6. Process Content: Click "Reimagine Now" to send selected paragraphs to the AI

   • Azure models work immediately (after one-time authentication)
   • Local models require Ollama to be running

7. Review and Apply:

   • Review AI-generated content in the "Modified/LLM Text" area
   • Edit the content if needed
   • Click "Apply" for current paragraph or "Apply All" for batch processing

Error Handling

The system provides helpful error messages:

• Ollama not installed: "Sorry, you need to install and run Ollama to use local models..."
• Model not available: "Sorry, the model 'gemma3:12b' is not installed in Ollama..."
• Azure authentication needed: Automatic browser-based authentication flow

Testing

Test Azure AI Foundry Connection

python test_azure_ai.py

Test Token Authentication

cd ollama-service
python token_fetch.py --scope "your-client-id/access_api"

Test Server

cd ollama-service
node server.js

Security

• Confidential Configuration: All sensitive data is in confidential.py (gitignored)
• Token Caching: Azure tokens are cached securely for reuse
• No Hardcoded Secrets: Use confidential_sample.py as a template
• Automatic Authentication: Browser-based OAuth flow for Azure

Troubleshooting

Azure AI Foundry Issues

• Ensure confidential.py is configured correctly
• Check Azure AD app permissions
• Verify Azure AI Foundry project access
• Complete browser authentication when prompted

Local Model Issues

• Install Ollama from https://ollama.ai
• Run ollama serve to start the service
• Pull required models with ollama pull <model-name>
• Check that Ollama is running on port 11434

Office Add-in Issues

• Ensure manifest.xml is properly sideloaded
• Check browser console for JavaScript errors
• Verify server is running on https://localhost:8000
• Use "Refresh Source" button to reload document content

Development

Project Structure

betterdoc2/
├── betterdoc-office-addin/          # Office add-in
│   └── BetterdocAddin/
│       ├── manifest.xml             # Office add-in manifest
│       └── src/taskpane/
│           ├── taskpane.html        # Main UI
│           ├── taskpane.js          # Core logic
│           └── styles.css           # Styling
├── ollama-service/                  # Multi-model proxy server
│   ├── server.js                    # Main server (Node.js)
│   └── token_fetch.py               # Azure authentication (Python)
├── confidential.py                  # Your Azure configuration (gitignored)
├── confidential_sample.py           # Configuration template
├── test_azure_ai.py                 # Test Azure connection
└── README.md                        # This file

Adding New Models

1. Azure Models: Update isAzureModel() function in server.js
2. Local Models: Add to window.availableModels array in taskpane.html
3. Custom Prompts: Modify the prompts dropdown in taskpane.html

License

[Add your license information here]