Installation
Quick Install
pip install stickler-eval
Requirements
- Python 3.12+
- conda (recommended)
Conda Setup
# Create a dedicated environment
conda create -n stickler python=3.12 -y
conda activate stickler
# Install from PyPI
pip install stickler-eval
Development Install
If you want to contribute or run from source:
git clone https://github.com/awslabs/stickler.git
cd stickler
# Create conda environment
conda create -n stickler python=3.12 -y
conda activate stickler
# Install with dev dependencies
pip install -e ".[dev]"
Optional Dependencies
Stickler's core comparators (Exact, Levenshtein, Numeric, Fuzzy) work out of the box. The AI-powered comparators require additional packages, installable via extras:
SemanticComparator
Uses AWS Bedrock Titan embeddings for cosine similarity.
pip install stickler-eval[semantic]
- AWS credentials configured (
aws configureor environment variables) - Access to Amazon Bedrock with Titan embedding models enabled
BERTComparator
Uses BERTScore for contextual similarity. Runs locally -- no cloud services needed.
pip install stickler-eval[bert]
- GPU recommended for performance, but CPU works
LLMComparator
Uses AWS Bedrock via strands-agents for LLM-powered comparison.
pip install stickler-eval[llm]
- AWS credentials configured
- Access to Amazon Bedrock with your chosen model enabled
All optional dependencies
Install everything at once:
pip install stickler-eval[bert,semantic,llm]
Verify Installation
python -c "import stickler; print('Stickler installed successfully')"
Run the quick start example:
python examples/scripts/quick_start.py
Run the test suite:
pytest tests/
Troubleshooting
See Known Issues for platform-specific problems (e.g., NumPy/GCC compatibility on RHEL).