GPT-10-K

DUKE_multi.py - Annual Report Analyzer

Overview

DUKE_multi.py is a Streamlit-based web application designed to analyze multiple annual reports in JSON or PDF format using OpenAI's GPT-4 model. The application allows users to upload multiple 10-K files and provides detailed answers to specific questions based on the content of the reports.

Features

• Multi-file Upload: Users can upload multiple JSON or PDF files simultaneously for analysis.
• Text Extraction: The app extracts text from the uploaded files. For JSON files, it parses and retrieves the relevant content. For PDF files, it utilizes PyMuPDF to extract text from each page.
• Content Chunking: The extracted content is chunked into manageable parts to ensure efficient processing by the GPT-4 model.
• AI-Powered Analysis: The application leverages Azure's OpenAI service to analyze the content of the reports and provide answers to user queries.
• Detailed Responses: The GPT-4 model provides detailed answers along with source data from the original documents, ensuring transparency and traceability of the analysis.
• Persistent Vector Database: The application uses ChromaDB to store and manage embeddings of the combined document content for efficient querying and analysis.

Usage

1. Upload Files: Users can upload one or more annual reports in JSON or PDF format.
2. Enter Query: Users input specific questions they need answers to, based on the content of the uploaded reports.
3. Analyze: Upon clicking the "Get Answer" button, the application processes the reports, queries the GPT-4 model, and displays the answers along with the source data.

Requirements

• streamlit
• openai
• PyMuPDF
• azure-openai
• chromadb
• json
• os
• base64

How to Run

1. Ensure all dependencies are installed as per the requirements.txt.
2. Place the logo image (logo.png) in the same directory as DUKE_multi.py.
3. Run the application using Streamlit:

   streamlit run DUKE_multi.py