semantic-copycat-oslili 1.2.9

Semantic Copycat Open Source License Identification Library

semantic-copycat-oslili

A high-performance tool for identifying licenses and copyright information in local source code, producing detailed evidence of where licenses are detected with support for all 700+ SPDX license identifiers.

What It Does

semantic-copycat-oslili analyzes local source code to produce evidence of:

• License detection - Shows which files contain which licenses with confidence scores
• SPDX identifiers - Detects SPDX-License-Identifier tags in ALL readable files
• Package metadata - Extracts licenses from package.json, pyproject.toml, METADATA files
• Copyright statements - Extracts copyright holders and years with intelligent filtering

The tool outputs standardized JSON evidence showing exactly where each license was detected, the detection method used, and confidence scores.

Key Features

• Evidence-based output: Shows exact file paths, confidence scores, and detection methods
• Parallel processing: Multi-threaded scanning with configurable thread count
• Three-tier detection:
  • Dice-Sørensen similarity matching (97% threshold)
  • TLSH fuzzy hashing (optional)
  • Regex pattern matching
• Smart normalization: Handles license variations and common aliases
• No file limits: Processes files of any size with intelligent sampling
• Enhanced metadata support: Detects licenses in package.json, METADATA, pyproject.toml
• False positive filtering: Advanced filtering for code patterns and invalid matches

Installation

pip install semantic-copycat-oslili

Required Dependencies

The package includes all necessary dependencies including python-tlsh for fuzzy hash matching, which is essential for accurate license detection and false positive prevention.

Usage

CLI Usage

# Scan a directory and see evidence
oslili /path/to/project

# Scan with parallel processing (4 threads)
oslili ./my-project --threads 4

# Scan a specific file
oslili /path/to/LICENSE

# Save results to file
oslili ./my-project -o license-evidence.json

# With custom configuration and verbose output
oslili ./src --config config.yaml --verbose

# Debug mode for detailed logging
oslili ./project --debug

Example Output

{
  "scan_results": [{
    "path": "./project",
    "license_evidence": [
      {
        "file": "/path/to/project/LICENSE",
        "detected_license": "Apache-2.0",
        "confidence": 0.988,
        "detection_method": "dice-sorensen",
        "match_type": "text_similarity",
        "description": "Text matches Apache-2.0 license (98.8% similarity)"
      },
      {
        "file": "/path/to/project/package.json",
        "detected_license": "Apache-2.0",
        "confidence": 1.0,
        "detection_method": "tag",
        "match_type": "spdx_identifier",
        "description": "SPDX-License-Identifier: Apache-2.0 found"
      }
    ],
    "copyright_evidence": [
      {
        "file": "/path/to/project/src/main.py",
        "holder": "Example Corp",
        "years": [2023, 2024],
        "statement": "Copyright 2023-2024 Example Corp"
      }
    ]
  }],
  "summary": {
    "total_files_scanned": 42,
    "licenses_found": {
      "Apache-2.0": 2
    },
    "copyrights_found": 1
  }
}

How It Works

Three-Tier License Detection System

The tool uses a sophisticated multi-tier approach for maximum accuracy:

1. Tier 1: Dice-Sørensen Similarity with TLSH Confirmation

   • Compares license text using Dice-Sørensen coefficient (97% threshold)
   • Confirms matches using TLSH fuzzy hashing to prevent false positives
   • Achieves 97-100% accuracy on standard SPDX licenses

2. Tier 2: TLSH Fuzzy Hash Matching

   • Uses Trend Micro Locality Sensitive Hashing for variant detection
   • Catches license variants like MIT-0, BSD-2-Clause vs BSD-3-Clause
   • Pre-computed hashes for all 700+ SPDX licenses

3. Tier 3: Pattern Recognition

   • Regex-based detection for license references and identifiers
   • Extracts from comments, headers, and documentation

Additional Detection Methods

• Package Metadata Scanning: Detects licenses from package.json, composer.json, pyproject.toml, etc.
• Copyright Extraction: Advanced pattern matching with validation and deduplication
• SPDX Identifier Detection: Finds SPDX-License-Identifier tags in source files

Library Usage

from semantic_copycat_oslili import LegalAttributionGenerator

# Initialize generator
generator = LegalAttributionGenerator()

# Process a local directory
result = generator.process_local_path("/path/to/source")

# Process a single file  
result = generator.process_local_path("/path/to/LICENSE")

# Generate evidence output
evidence = generator.generate_evidence([result])
print(evidence)

# Access results
for license in result.licenses:
    print(f"License: {license.spdx_id} ({license.confidence:.0%} confidence)")
for copyright in result.copyrights:
    print(f"Copyright: © {copyright.holder}")

License Detection

The package uses a three-tier license detection system:

1. Tier 1: Dice-Sørensen similarity (97% threshold)
2. Tier 2: TLSH fuzzy hashing (97% threshold)
3. Tier 3: Machine learning or regex pattern matching

Output Format

The tool outputs JSON evidence showing:

• File path: Where the license was found
• Detected license: The SPDX identifier of the license
• Confidence: How confident the detection is (0.0 to 1.0)
• Match type: How the license was detected (license_text, spdx_identifier, license_reference, text_similarity)
• Description: Human-readable description of what was found

Configuration

Create a config.yaml file:

similarity_threshold: 0.97
max_extraction_depth: 10
thread_count: 4
custom_aliases:
  "Apache 2": "Apache-2.0"
  "MIT License": "MIT"

Documentation

• Full Usage Guide - Comprehensive usage examples and configuration
• API Reference - Python API documentation and examples
• Changelog - Version history and changes