bott-wa/multi-ai-document-analysis

Multi-AI Document Analysis 🤖📄

Compare AWS Textract, Claude AI, and OpenAI GPT-4V for document processing and OCR tasks.

Features

🔥 5 Processing Methods:

• —AWS Textract (MCP) - Professional OCR with form/table detection
• —Claude AI (Direct) - Vision + reasoning with Haiku model
• —OpenAI GPT-4V (Direct) - Advanced vision + reasoning
• —Hybrid (Textract + Claude) - OCR + AI enhancement
• —Hybrid (Textract + OpenAI) - OCR + GPT-4V enhancement

📊 Cost Comparison:

• —Budget: AWS Textract (~$0.0015/page)
• —Balanced: Claude Hybrid (~$0.007/document)
• —Premium: OpenAI Hybrid (~$0.015/document)

✨ Supported Formats: PNG, JPEG, PDF, TIFF (10MB max)

Setup

To run locally or fork this space:

1. 1.Add your API keys in the Space settings (or .env file locally):

   AWS_ACCESS_KEY_ID=your_key
   AWS_SECRET_ACCESS_KEY=your_secret  
   AWS_DEFAULT_REGION=us-east-1
   ANTHROPIC_API_KEY=your_anthropic_key
   OPENAI_API_KEY=your_openai_key

1. 1.Install dependencies:

   pip install -r requirements.txt

1. 1.Run the app:

   python app.py

Usage

1. 1.Test Connections: Verify your API keys work
2. 2.Upload Document: Drop any PDF, image, or form
3. 3.Select Method: Choose your processing approach
4. 4.Compare Results: See the differences in output and cost

Architecture

This app demonstrates a multi-AI comparison framework for document analysis:

• —Direct API integrations (no cloud infrastructure needed)
• —Structured JSON output for easy integration
• —Cost and performance benchmarking
• —Hybrid processing for enhanced accuracy

License

MIT License - feel free to fork and modify!