XingC Intelligent Email Assistant: Directly Call Multi-Model AI for Email Processing in Outlook

XingC Intelligent Email Assistant is an open-source Microsoft Outlook Web add-in that integrates multiple large models including DeepSeek, Tongyi Qianwen, and Gemini. It supports full features like intelligent reply, email summarization, draft optimization, and privacy protection, and can be directly embedded into the Outlook workflow to double the efficiency of email processing.

• Original author/maintainer: XingC233
• Source platform: GitHub
• Release time: June 2026 This project is an AI-enhanced Web add-in designed specifically for Microsoft Outlook. It adopts a front-end and back-end separation architecture (back-end: Python FastAPI; front-end: JavaScript + Office.js) and uses HTTPS communication to meet security requirements. Users can complete the full AI-assisted email operations without switching windows.

System Architecture

Component	Tech Stack	Service Address
Backend API	Python·FastAPI·uvicorn	https://localhost:8000
Outlook Add-in	JavaScript·Office.js·Webpack	https://localhost:3000

Environment Requirements

Requires installation of Python 3. 10+, Node.js 18+, OpenSSL, and Outlook Web App, and at least one LLM service API key (DeepSeek/Alibaba Cloud Bailian/Google AI Studio) must be prepared.

Deployment Process

1. Clone the repository and install dependencies (the back-end needs to download the spaCy English NLP model);
2. Configure the .env file and fill in the API key;
3. Generate a self-signed SSL certificate;
4. Import manifest.xml in Outlook Web App to add the add-in.

• Intelligent Reply Generation: One-click generation of bilingual replies in multiple styles/lengths;
• Email Summary Extraction: Automatically extract core information from long emails;
• Intelligent Draft Optimization: Polish drafts from dimensions like clarity and grammar;
• Text Rewriting and Translation: Support style conversion and language switching;
• RAG Knowledge Base Integration: Generate enterprise-compliant replies by referring to private documents;
• Privacy Protection: Automatically detect and anonymize PII sensitive information;
• Multi-Model Switching: Support free selection of models like DeepSeek V3, Tongyi Qianwen Plus, Gemini 2.5 Flash Lite.

Applicable groups:

• Business professionals: Quickly generate professional replies;
• Multinational teams: Flexible switching between Chinese and English;
• Enterprise users: Generate compliant replies based on internal knowledge bases;
• Privacy-sensitive users: Protect sensitive information. Value: Seamlessly embed into the Outlook workflow, avoid context interruptions, and significantly improve email processing efficiency.

Back-end configurations are centralized in the config.yaml file, which can be adjusted:

• Default LLM model (deepseek-v3/qwen-plus/gemini-flash-lite);
• RAG parameters (chunk size, similarity threshold, embedding model);
• CORS allowed origin addresses. The configuration-based design facilitates customization by individuals or teams according to actual scenarios.