AI Email Assistant for Outlook 365 (Logistics Industry)

Overview

This AI Email Assistant is designed to enhance email productivity for logistics staff using Microsoft Outlook 365 Classic Edition. The assistant helps users by:

• Fetching Outlook emails securely and efficiently.
• Summarizing emails in multiple languages using the DeepSeek API.
• Generating AI-powered email replies with customizable language, tone, and format.
• Identifying spam emails and ensuring data security by operating locally to prevent privacy breaches.
• Reading and extracting text from PDF attachments using pdf.js and pdf2image (via Electron + Python integration).

This project is designed to improve communication efficiency, reduce email processing time, and support non-native English speakers in writing professional emails.

---

Key Features

✅ Fetches emails directly from Outlook 365 Classic Edition. ✅ Summarizes lengthy email conversations in clear and concise language. ✅ Generates suggested replies that users can review and customize. ✅ Supports multiple languages for improved client communication. ✅ Detects and flags potential spam emails. ✅ Extracts text from PDF attachments by converting them to images and processing them via Llama 3.2's image capabilities. ✅ Prioritizes security by running operations locally to prevent data leaks.

---

Setup Instructions

1. Prerequisites

Ensure you have the following installed:

• Node.js (v18 or higher)
• Electron (latest version)
• Python 3.10+ (for PDF-to-image conversion)
• Outlook 365 Classic Edition (with API access)

---

2. Install Dependencies

Run the following commands in your terminal:

npm install
npm install pdfjs-dist
pip install pdf2image flask

---

3. Electron Main Process Setup

Ensure your main.ts (or electron.js) includes the Python process starter:

const { app, BrowserWindow, ipcMain } = require("electron");
const { spawn } = require("child_process");
const path = require("path");
const axios = require("axios");

app.whenReady().then(() => {
    const pythonProcess = spawn("python", [path.join(__dirname, "pdf_converter.py")]);

    pythonProcess.stdout.on("data", (data) => console.log(`Python: ${data}`));
    pythonProcess.stderr.on("data", (data) => console.error(`Python Error: ${data}`));

    app.on("window-all-closed", () => {
        pythonProcess.kill();
        app.quit();
    });
});

---

4. Python PDF-to-Image Converter

Create a new file called pdf_converter.py for PDF-to-image conversion.

from flask import Flask, request, jsonify
from pdf2image import convert_from_path
import os

app = Flask(__name__)
OUTPUT_FOLDER = "output_images"
os.makedirs(OUTPUT_FOLDER, exist_ok=True)

@app.route("/convert-pdf", methods=["POST"])
def convert_pdf():
    if "file" not in request.files:
        return jsonify({"error": "No file provided"}), 400

    file = request.files["file"]
    pdf_path = os.path.join(OUTPUT_FOLDER, file.filename)
    file.save(pdf_path)

    images = convert_from_path(pdf_path, dpi=300)
    image_paths = []

    for i, img in enumerate(images):
        img_path = os.path.join(OUTPUT_FOLDER, f"{file.filename}_page_{i + 1}.png")
        img.save(img_path, "PNG")
        image_paths.append(img_path)

    return jsonify({"images": image_paths})

if __name__ == "__main__":
    app.run(port=5001)

---

5. Running the Application

To start the app:

npm run start

---

6. Usage

• Fetch Emails: Automatically pulls selected Outlook emails.
• Summarize Emails: Click the "Summarize Email" button to generate concise email summaries in your preferred language.
• Generate Replies: Click the "Write a Reply" button to create suggested responses.
• Spam Detection: Click "A Spam Email?" to check if an email is potentially spam.
• PDF Extraction: Upload PDF attachments — the assistant will extract text automatically for improved understanding.

---

Project Structure

📂 src
 ┣ 📂controllers
 ┃ ┣ 📜ConfigController.ts
 ┃ ┣ 📜OllamaController.ts
 ┣ 📂models
 ┃ ┣ 📜Files.ts
 ┃ ┣ 📜OutlookEmailItem.ts
 ┣ 📂renderer
 ┃ ┣ 📜app.tsx
 ┃ ┣ 📜renderer.ts
 ┣ 📜main.ts
 ┣ 📜pdf_converter.py
 ┣ 📜package.json
 ┗ 📜README.txt

---

Known Issues & Troubleshooting

1. PowerShell Execution Policy Error:
   ➡️ Run this in PowerShell (Admin) to bypass temporarily:

   Set-ExecutionPolicy Bypass -Scope Process

2. pdf.worker.entry Error:
   ➡️ Ensure you have the correct import statement in app.tsx:

   import * as pdfjsLib from "pdfjs-dist";

import pdfWorker from "pdfjs-dist/build/pdf.worker.js"; pdfjsLib.GlobalWorkerOptions.workerSrc = pdfWorker;

3. **PDF Content Not Extracting:**  
➡️ Ensure your `handleFileChange()` function includes PDF text extraction logic.  
➡️ Check the console for extracted text for debugging.

---

## Future Improvements
🔹 Enhance prompt engineering for more natural and professional responses.  
🔹 Implement **confidence scoring** to flag uncertain AI-generated content.  
🔹 Improve **error handling** for failed PDF extractions and API limits.  

---

## Credits
Developed by *Wynter Lin** and **Yang Yue (Danny)** for the **IMT 574 Project** at the **University of Washington**.

---

## License
This project is licensed under the **MIT License**. Feel free to contribute and improve the codebase! 😊