XML Analysis Framework

A production-ready XML document analysis and preprocessing framework with 29 specialized handlers designed for AI/ML data pipelines. Transform any XML document into structured, AI-ready data and optimized chunks with 100% success rate across 71 diverse test files.

🔗 Part of Analysis Framework Suite

This framework is part of a unified suite of document analysis tools that share a consistent interface:

• analysis-framework-base - Base interfaces and shared models
• xml-analysis-framework - XML document analysis (this package)
• docling-analysis-framework - PDF/Office documents via Docling
• document-analysis-framework - Text, code, config files
• data-analysis-framework - Structured data analysis

All frameworks implement the same BaseAnalyzer and BaseChunker interfaces from analysis-framework-base, enabling:

• Consistent API across document types
• Easy framework switching with minimal code changes
• Unified result format for downstream processing
• Shared tooling and utilities

🚀 Quick Start

Simple API - Get Started in Seconds

import xml_analysis_framework as xaf

# 🎯 One-line analysis with specialized handlers
result = xaf.analyze("path/to/file.xml")
print(f"Document type: {result['document_type'].type_name}")
print(f"Handler used: {result['handler_used']}")

# 📊 Basic schema analysis  
schema = xaf.analyze_schema("path/to/file.xml")
print(f"Elements: {schema.total_elements}, Depth: {schema.max_depth}")

# ✂️ Smart chunking for AI/ML
chunks = xaf.chunk("path/to/file.xml", strategy="auto")
print(f"Created {len(chunks)} optimized chunks")

# 💾 Save chunks to JSON
import json

# Convert chunks to JSON-serializable format
chunks_data = [
    {
        "chunk_id": chunk.chunk_id,
        "content": chunk.content,
        "element_path": chunk.element_path,
        "start_line": chunk.start_line,
        "end_line": chunk.end_line,
        "elements_included": chunk.elements_included,
        "metadata": chunk.metadata,
        "token_estimate": chunk.token_estimate
    }
    for chunk in chunks
]

# Write to file
with open("chunks_output.json", "w") as f:
    json.dump(chunks_data, f, indent=2)

Advanced Usage

import xml_analysis_framework as xaf

# Enhanced analysis with full results
analysis = xaf.analyze_enhanced("document.xml")

print(f"Type: {analysis.type_name} (confidence: {analysis.confidence:.2f})")
print(f"AI use cases: {len(analysis.ai_use_cases)}")
if analysis.quality_metrics:
    print(f"Quality score: {analysis.quality_metrics.get('completeness_score')}")
else:
    print("Quality metrics: Not available")

# Different chunking strategies
hierarchical_chunks = xaf.chunk("document.xml", strategy="hierarchical")
sliding_chunks = xaf.chunk("document.xml", strategy="sliding_window") 
content_chunks = xaf.chunk("document.xml", strategy="content_aware")

# Process chunks
for chunk in hierarchical_chunks:
    print(f"Chunk {chunk.chunk_id}: {len(chunk.content)} chars")
    print(f"Path: {chunk.element_path}, Elements: {len(chunk.elements_included)}")

# 💾 Save different chunking strategies to separate files
import json

# Helper function to convert chunk to dict
def chunk_to_dict(chunk):
    return {
        "chunk_id": chunk.chunk_id,
        "content": chunk.content,
        "element_path": chunk.element_path,
        "start_line": chunk.start_line,
        "end_line": chunk.end_line,
        "elements_included": chunk.elements_included,
        "metadata": chunk.metadata,
        "token_estimate": chunk.token_estimate
    }

# Save each strategy's results
strategies = {
    "hierarchical": hierarchical_chunks,
    "sliding_window": sliding_chunks,
    "content_aware": content_chunks
}

for strategy_name, chunks in strategies.items():
    chunks_data = [chunk_to_dict(chunk) for chunk in chunks]
    
    with open(f"chunks_{strategy_name}.json", "w") as f:
        json.dump({
            "strategy": strategy_name,
            "total_chunks": len(chunks_data),
            "chunks": chunks_data
        }, f, indent=2)
    
    print(f"Saved {len(chunks_data)} chunks to chunks_{strategy_name}.json")

Expert Usage - Direct Class Access

# For advanced customization, use the classes directly
from xml_analysis_framework import XMLDocumentAnalyzer, ChunkingOrchestrator

analyzer = XMLDocumentAnalyzer(max_file_size_mb=500)
orchestrator = ChunkingOrchestrator(max_file_size_mb=1000)

# Custom analysis
result = analyzer.analyze_document("file.xml")

# Custom chunking with config (result works directly now!)
from xml_analysis_framework.core.chunking import ChunkingConfig
config = ChunkingConfig(
    max_chunk_size=2000,
    min_chunk_size=300,
    overlap_size=150,
    preserve_hierarchy=True
)
chunks = orchestrator.chunk_document("file.xml", result, strategy="auto", config=config)

# 💾 Save with analysis metadata
import json
from datetime import datetime

output_data = {
    "metadata": {
        "file": "file.xml",
        "processed_at": datetime.now().isoformat(),
        "document_type": result.type_name,
        "confidence": result.confidence,
        "handler_used": result.handler_used,
        "chunking_config": {
            "strategy": "auto",
            "max_chunk_size": config.max_chunk_size,
            "min_chunk_size": config.min_chunk_size,
            "overlap_size": config.overlap_size,
            "preserve_hierarchy": config.preserve_hierarchy
        }
    },
    "analysis": {
        "ai_use_cases": result.ai_use_cases,
        "key_findings": result.key_findings,
        "quality_metrics": result.quality_metrics
    },
    "chunks": [
        {
            "chunk_id": chunk.chunk_id,
            "content": chunk.content,
            "element_path": chunk.element_path,
            "start_line": chunk.start_line,
            "end_line": chunk.end_line,
            "elements_included": chunk.elements_included,
            "metadata": chunk.metadata,
            "token_estimate": chunk.token_estimate
        }
        for chunk in chunks
    ]
}

with open("analysis_and_chunks.json", "w") as f:
    json.dump(output_data, f, indent=2)

print(f"Saved complete analysis with {len(chunks)} chunks to analysis_and_chunks.json")

🎯 Key Features

1. 🧠 29 Specialized XML Handlers

Automatically detects and analyzes different XML document types:

• Security & Compliance: SCAP, SAML, SOAP
• DevOps & Build: Maven POM, Ant, Ivy, Spring, Log4j
• Content & Documentation: RSS/Atom, DocBook, XHTML, SVG
• Enterprise Systems: ServiceNow, Hibernate, Struts configurations
• Data & APIs: GPX, KML, GraphML, WADL/WSDL, XML Schemas

2. ⚡ Intelligent Chunking Strategies

• Hierarchical: Preserves document structure and relationships
• Sliding Window: Fixed-size chunks with configurable overlap
• Content-Aware: Groups related content based on semantic meaning
• Auto-Selection: Automatically chooses best strategy based on document type

3. 🤖 AI/ML Ready Output

• Token-Optimized: Chunks sized for LLM context windows
• Rich Metadata: Each chunk includes context, line numbers, and relationships
• JSON Export: Easy integration with vector stores and AI pipelines
• Quality Metrics: Automated assessment of data completeness and structure

4. 🔒 Enterprise Security

• Safe XML Parsing: Uses defusedxml to prevent XXE attacks
• File Size Limits: Configurable limits to prevent resource exhaustion
• Minimal Dependencies: Only defusedxml + Python standard library

📋 Supported Document Types

Category	Document Types	Common Use Cases
Security & Compliance	SCAP, SAML, SOAP	Vulnerability scanning, authentication, web services
Build & Configuration	Maven POM, Ant, Spring, Log4j	Dependency management, build automation, app config
Enterprise Systems	ServiceNow, Hibernate, Struts	IT service management, ORM mapping, web frameworks
Content & Media	RSS/Atom, DocBook, XHTML, SVG	Feeds, documentation, web content, graphics
Geospatial	GPX, KML, GraphML	GPS tracking, maps, network graphs
APIs & Services	WADL, WSDL, OpenAPI	REST APIs, SOAP services, API documentation
Data Exchange	XLIFF, XML Sitemap, Generic XML	Translations, SEO, custom formats

🔒 Security

XML Security Protection

This framework uses defusedxml to protect against common XML security vulnerabilities:

• XXE (XML External Entity) attacks: Prevents reading local files or making network requests
• Billion Laughs attack: Prevents exponential entity expansion DoS attacks
• DTD retrieval: Blocks external DTD fetching to prevent data exfiltration

Security Features

import xml_analysis_framework as xaf

# Safe parsing - malicious XML will be rejected automatically
try:
    result = xaf.analyze("potentially_malicious.xml")
except Exception as e:
    print(f"Security threat detected: {e}")

# The framework automatically protects against:
# - XXE attacks
# - Billion laughs / exponential entity expansion
# - External DTD retrieval

Best Practices

1. Always use the framework's parsers - Never use xml.etree.ElementTree directly
2. Validate file sizes - Set reasonable limits for your use case
3. Sanitize file paths - Ensure input paths are properly validated
4. Monitor for security exceptions - Log and alert on security-blocked parsing attempts

File Size Limits

The framework includes built-in file size limits to prevent memory exhaustion:

import xml_analysis_framework as xaf
from xml_analysis_framework import XMLDocumentAnalyzer, ChunkingOrchestrator

# Default limits are reasonable for most use cases
# But you can customize them:

# Create analyzer with custom 50MB limit
analyzer = XMLDocumentAnalyzer(max_file_size_mb=50.0)
result = analyzer.analyze_document("large_file.xml")

# Create chunking orchestrator with 100MB limit  
orchestrator = ChunkingOrchestrator(max_file_size_mb=100.0)
chunks = orchestrator.chunk_document("large_file.xml", result)

# For simple API, defaults are used automatically
try:
    result = xaf.analyze("very_large_file.xml")
except ValueError as e:
    print(f"File too large: {e}")

🔧 Installation

# Install from PyPI (recommended)
pip install xml-analysis-framework

# Install from source
git clone https://github.com/redhat-ai-americas/xml-analysis-framework.git
cd xml-analysis-framework
pip install -e .

# Or install development dependencies
pip install -e .[dev]

Dependencies

• defusedxml (0.7.1+): For secure XML parsing protection
• Python standard library (3.8+) for all other functionality

🧪 Testing

The framework includes comprehensive tests for all handlers and features:

# Run all tests
python -m pytest tests/

# Run specific test categories
python -m pytest tests/unit/           # Unit tests for handlers
python -m pytest tests/integration/    # Integration tests
python -m pytest tests/comprehensive/  # Full system tests

🤖 AI/ML Integration

AI Processing Pipeline

XML Documents → Analysis Framework → Structured Output → AI/ML Systems

1. Document Analysis (29 specialized handlers)
2. Smart Chunking (token-optimized)
3. JSON Export (with metadata)
4. Integration with:
   - Vector databases (semantic search)
   - LLMs (document Q&A, analysis)
   - Graph databases (relationship mapping)
   - ML pipelines (feature extraction)

Common AI Use Cases

• Security Intelligence: Analyze SCAP reports, detect vulnerabilities, compliance monitoring
• DevOps Automation: Dependency analysis, configuration validation, build optimization
• Enterprise Search: Semantic search across technical documentation and configurations
• Knowledge Extraction: Extract structured data from XML for ML training datasets

🚀 Extending the Framework

Adding New Handlers

from xml_analysis_framework.base import XMLHandler, SpecializedAnalysis, DocumentTypeInfo

class CustomHandler(XMLHandler):
    def can_handle_xml(self, root, namespaces):
        # Check if this handler can process the document
        if root.tag == 'custom-format':
            return True, 1.0  # (can_handle, confidence)
        return False, 0.0
  
    def detect_xml_type(self, root, namespaces):
        return DocumentTypeInfo(
            type_name="Custom Format",
            confidence=1.0,
            version="1.0"
        )
  
    def analyze_xml(self, root, file_path):
        return SpecializedAnalysis(
            type_name="Custom Format",
            confidence=1.0,
            key_findings={"custom_data": "value"},
            ai_use_cases=["Custom AI application"],
            structured_data={"extracted": "data"},
            file_path=file_path,
            handler_used="CustomHandler"
        )
    
    def extract_xml_key_data(self, root):
        # Extract key data specific to your format
        return {"key": "value"}

Custom Chunking Strategies

from xml_analysis_framework.core.chunking import XMLChunkingStrategy, XMLChunk
import xml.etree.ElementTree as ET

class CustomChunking(XMLChunkingStrategy):
    def chunk_document(self, file_path, specialized_analysis=None):
        chunks = []
        tree = ET.parse(file_path)
        root = tree.getroot()
        
        # Custom chunking logic
        for i, element in enumerate(root):
            chunk = XMLChunk(
                chunk_id=f"custom_{i}",
                content=ET.tostring(element, encoding='unicode'),
                element_path=f"/{element.tag}",
                start_line=1,
                end_line=10,
                parent_context=None,
                metadata={"custom": True},
                token_estimate=100,
                elements_included=[element.tag]
            )
            chunks.append(chunk)
        
        return chunks

# Use with the framework
import xml_analysis_framework as xaf
from xml_analysis_framework import ChunkingOrchestrator

orchestrator = ChunkingOrchestrator()
# The orchestrator will use your custom strategy when needed

🤝 Contributing

We welcome contributions! Whether you're adding new XML handlers, improving chunking algorithms, or enhancing AI integrations, your contributions help make XML analysis more accessible and powerful.

Priority contribution areas:

• 🎯 New XML format handlers (ERP, CRM, healthcare, government)
• ⚡ Enhanced chunking algorithms and strategies
• 🚀 Performance optimizations for large files
• 🤖 Advanced AI/ML integration examples
• 📝 Documentation and usage examples

👉 See CONTRIBUTING.md for complete guidelines, development setup, and submission process.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• Designed as part of the AI Building Blocks initiative
• Built for the modern AI/ML ecosystem
• Community-driven XML format support