RAG-Anything Test Setup

This project demonstrates how to use RAG-Anything with MinerU v2.5 for multimodal document processing.

Installation

This project uses uv for dependency management. The project has been initialized with:

uv init --python 3.10
uv add 'raganything[all]'

Files

• raganything_example.py - Main example script with full RAG-Anything functionality
• test_mineru.py - Test script to verify MinerU installation (no API key required)

Usage

1. Test MinerU Installation

First, verify that MinerU v2.5 is properly installed:

# Test installation only
uv run python test_mineru.py --check-only

# Test with a document (parsing test)
uv run python test_mineru.py --file path/to/your/document.pdf

2. Run Full Example

To run the complete RAG-Anything example, you'll need an OpenAI API key:

# Using command line argument
uv run python raganything_example.py path/to/document.pdf --api-key YOUR_API_KEY

# Using environment variable
export OPENAI_API_KEY=your_api_key_here
uv run python raganything_example.py path/to/document.pdf

# With custom options
uv run python raganything_example.py document.pdf \
  --api-key YOUR_API_KEY \
  --parser mineru \
  --parse-method auto \
  --output-dir ./output \
  --working-dir ./rag_storage

3. Available Options

• --parser: Choose between "mineru" (default) or "docling"
• --parse-method: Choose between "auto" (default), "ocr", or "txt"
• --output-dir: Directory for processed documents (default: ./output)
• --working-dir: Directory for RAG storage (default: ./rag_storage)
• --base-url: Custom OpenAI base URL (optional)

Features Demonstrated

1. Document Processing: Complete end-to-end processing of PDFs, Office documents, images
2. Multimodal Support: Processing of text, images, tables, and equations
3. Query Types:
   • Pure text queries
   • VLM-enhanced queries (automatic image analysis)
   • Multimodal queries with specific content
4. MinerU v2.5 Integration: Latest version with improved parsing capabilities

Requirements

• Python 3.10+
• uv package manager
• OpenAI API key (for full functionality)
• LibreOffice (for Office document processing)

Troubleshooting

If you encounter issues:

1. MinerU not found: Run uv run python test_mineru.py --check-only
2. Office documents not processing: Install LibreOffice
3. API errors: Verify your OpenAI API key and base URL
4. Memory issues: Try using --parse-method txt for text-only processing

Example Output

The script will:

1. Process your document using MinerU v2.5
2. Extract text, images, tables, and equations
3. Create a knowledge graph
4. Run example queries to demonstrate capabilities
5. Save processed files to the output directory