DocuSheriff 🤠

Wrangling insights from annual reports with Generative AI and music-driven storytelling.

Overview

DocuSheriff reimagines how we interact with annual reports by combining generative AI, tone-aware music, and a Wild West–inspired interface. Users ask natural language questions and receive accurate, page-referenced answers powered by GPT-4o. Each answer is accompanied by AI-generated music that reflects the emotional tone of the content, making financial exploration more engaging and accessible.

Features

• Natural language Q&A over complex annual reports
• Accurate answers with document page references
• Handles text, tables, and visual elements
• AI-generated music (SOUNDRAW) matched to the tone of the response
• Wild West–inspired themed experience

Tech Stack

• Language Model: GPT-4o via LangChain (RAG-based architecture)
• PDF Processing: PyMuPDF
• Backend: Python + Flask
• Frontend: Unity (for immersive experience)
• Audio Integration: SOUNDRAW (pre-generated AI music), Unity AudioSource
• Tone Classification: Custom keyword-based classifier and prompt-tagging

Team

• Varsha Viswanathan
• Shreya Krishnan
• Asvataman Venkatesan
• Sai Greeshma Nerella

Acknowledgments

• LTIMindtree for the challenge prompt and report dataset
• SOUNDRAW for enabling AI-generated music

---