SmartBrief

KO | EN

gitlite — search

#python #rust #flutter #config #data #github-config #rust-lang #terraform #nodejs #typescript #javascript #website

★ 30 Open GitHub ↗

A low-code open source solution for document summarization using Microsoft Power Platform and Azure OpenAI

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# SmartBrief - a low-code AI Document Summarizer

> AI-Powered Document Briefing Platform

[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://raw.githubusercontent.com/rkneela0912/SmartBrief/main/LICENSE)
[![GitHub stars](https://img.shields.io/github/stars/rkneela0912/SmartBrief.svg)](https://github.com/rkneela0912/SmartBrief/stargazers)
<a href="https://doi.org/10.5281/zenodo.17329304"> <img alt="Download on Zenodo" src="https://img.shields.io/badge/Download%20on-Zenodo-1f77b4?style=flat&logo=zenodo&logoColor=white&labelColor=555555" height="20"> </a>

[📖 Documentation](docs/project_report.md) | [📄 Research Paper](docs/Research_paper.md) | [🚀 Implementation Guide](docs/IMPLEMENTATION_GUIDE.md) 
| [🎥 SmartBrief Walkthrough](https://drive.google.com/file/d/1TZVVxG4UXP6XgIW3CHzfkYntldLvyOyK/view?usp=sharing) | [🧑‍💻👩‍💻 Flow Walkthrough - For Devs](https://drive.google.com/file/d/10-JCX-tjGN7zKQ9KSGmalv_fK5QTW11M/view?usp=sharing)

---

## 🧠 Overview

**SmartBrief**, a low-code AI Document Summarizer is an enterprise-grade, low-code framework for AI-powered document summarization. I developed this platform to demonstrate how Microsoft Power Platform and Azure OpenAI (GPT-4o) can be seamlessly integrated to create production-ready AI solutions without extensive custom development.

This project addresses the critical business challenge of extracting key insights from large documents efficiently, enabling organizations to process hundreds of documents daily with minimal manual effort.

---

## 🎯 Objectives

Through this research and implementation, I aimed to:

- Evaluate the effectiveness of Retrieval-Augmented Generation (RAG) based summarization in a low-code environment
- Demonstrate seamless integration of Power Platform components (Power Apps, Power Automate, Dataverse) with Azure AI services
- Create a reusable and scalable architecture for AI-powered document processing within enterprises
- Achieve production-grade performance with 92% accuracy and sub-3-second processing latency

---

## 🧩 Architecture

![Architecture Diagram](https://raw.githubusercontent.com/rkneela0912/SmartBrief/main/images/architecture_diagram.png)

**Key Components:**

- **Presentation Layer:** Power Apps canvas application for document upload and summary viewing
- **Orchestration Layer:** Power Automate cloud flows managing the processing pipeline
- **AI Processing Layer:** Azure OpenAI (GPT-4o) with 128K context window
- **Data Layer:** Microsoft Dataverse for secure storage and audit trails

---

## 🛠️ Tech Stack

- **Power Apps:** User interface and interaction layer
- **Power Automate:** Workflow orchestration and business logic
- **Azure OpenAI (GPT-4o):** Core summarization using the latest multimodal language model with 128K context window
- **Microsoft Dataverse:** Enterprise-grade data storage with built-in security
- **AI Builder:** Document processing and text extraction
- **Azure Active Directory:** Authentication and authorization

---

## ⚙️ Implementation Steps

### Quick Start

1. **Dataverse Schema:** Create tables in Dataverse to store uploaded documents and their summaries
2. **Power App UI:** Design a user-friendly Power App with file upload control and summary gallery
3. **Power Automate Flow:** Build a flow that triggers on file upload, converts documents to text, calls Azure OpenAI API, and saves results
4. **Azure OpenAI Integration:** Configure GPT-4o deployment and connect to Power Automate
5. **Deployment and Testing:** Deploy the solution and test with various document types

### Detailed Guide

For complete step-by-step implementation instructions, see [Implementation Guide](docs/IMPLEMENTATION_GUIDE.md).

---

## 📊 Performance Evaluation

| Metric | Value | Description |
|--------|-------|-------------|
| **Summarization Accuracy** | 92% | ROUGE-L score against human-generated summaries |
| **Processing Latency** | 2.5s | Average time from document upload to summary display |
| **User Satisfaction** | 4.8/5 | Based on user survey of 20 participants over 6 weeks |
| **Cost per Document** | $0.05 | Average Azure OpenAI API cost per document |
| **Supported Formats** | PDF, DOCX | With extensibility for additional formats |
| **Context Window** | 128K tokens | GPT-4o capability for long documents |

---

## 🔬 Research & Publications

### Academic Papers

- **[Research Paper](docs/Research_paper.md)** - Complete research paper with methodology, evaluation, and ethics considerations
- **[Project Report](docs/project_report.md)** - Detailed technical report with implementation insights

### Technical Documentation

- **[Implementation Guide](docs/IMPLEMENTATION_GUIDE.md)** - Step-by-step deployment instructions
- **[Model Notes](docs/MODEL_NOTES.md)** - GPT-4o configuration and optimization guide
- **[API Reference](docs/references.bib)** - BibTeX citations and references

---

## 💡 Key Features

✅ **GPT-4o Powered** - Latest Azure OpenAI model with 128K context window  
✅ **Intelligent Chunking** - Automatic handling of long documents with parallel processing  
✅ **Low-Code Implementation** - Built entirely with Power Platform (minimal custom code)  
✅ **Enterprise Security** - Azure AD authentication and Dataverse security model  
✅ **Cost Effective** - ~$0.05 per document with transparent cost tracking  
✅ **Production Ready** - Comprehensive error handling and monitoring  
✅ **Scalable Architecture** - Handles concurrent processing with configurable limits  
✅ **Audit Trail** - Complete tracking of document processing history

---

## 🚀 Use Cases

### Enterprise Applications

- **Contract Review:** Quickly extract key terms and obligations from legal documents
- **Research Analysis:** Summarize academic papers and technical reports
- **Meeting Minutes:** Generate concise summaries from lengthy meeting transcripts
- **Compliance Documentation:** Extract critical information from regulatory documents
- **Customer Feedback:** Synthesize insights from lengthy customer surveys

### Performance Characteristics

- Documents under 5 pages: **1.8s average latency**
- Documents 5-20 pages: **2.5s average latency**
- Documents over 20 pages: **4.2s average latency**
- Monthly cost for 1,000 documents: **~$50**

---

## 📖 Citation

If you use **SmartBrief** in your research or project, please cite:

```bibtex
@software{neela2025smartbrief,
  author = {Neela, Ranjith Kumar},
  title = {SmartBrief: AI-Powered Document Briefing Platform},
  year = {2025},
  publisher = {GitHub},
  url = {https://github.com/rkneela0912/SmartBrief},
  note = {Enterprise-grade low-code framework for AI-powered document summarization}
}
```

For the research paper:

```bibtex
@article{neela2025smartbrief,
  title={SmartBrief: An Enterprise-Grade Low-Code Framework for AI-Powered Document Summarization},
  author={Neela, Ranjith Kumar},
  journal={IEEE Access},
  year={2025},
  note={Under Review}
}
```

---

## 🤝 Contributing

Contributions are welcome! I encourage:

- Bug reports and feature requests via [Issues](https://github.com/rkneela0912/SmartBrief/issues)
- Pull requests for improvements and extensions
- Documentation enhancements
- Use case sharing and feedback

Please see [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.

---

## 📄 License

This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.

---

## 👨‍💻 Author

**Ranjith Kumar Neela**  
Independent AI Researcher & Software Engineer

📧 Email: iamranjithkneela@gmail.com  
🔗 LinkedIn: [linkedin.com/in/ranjithkumarneela]([https://linkedin.com/in/ranjithkumarneela](https://www.linkedin.com/in/ranjith-k-n-846992125/))  
🐙 GitHub: [@rkneela0912](https://github.com/rkneela0912)  
🌐 Website: [www.ranjithneela.com](https://rkneela0912.github.io/portfolio/)

---

## 🙏 Acknowledgments

I would like to acknowledge:

- Microsoft for the Power Platform and Azure OpenAI services
- OpenAI for the GPT-4o model
- The open-source community for inspiration and tools
- Early users and testers who provided valuable feedback

---

## 📊 Project Stats

- **Development Time:** 3 months (research + implementation)
- **Lines of Code:** ~2,500 (Power Automate expressions + Python utilities)
- **Test Documents:** 500+ documents across various domains
- **User Testing:** 20 participants over 6 weeks
- **Documentation:** 15,000+ words across multiple documents

---

## 🔗 Related Projects

- [Azure OpenAI Documentation](https://learn.microsoft.com/en-us/azure/ai-services/openai/)
- [Microsoft Power Platform](https://powerplatform.microsoft.com/)

---

**⭐ If you find SmartBrief useful, please consider starring the repository!**

SmartBrief

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README