Installation
Get started with Benchwise by installing the SDK and setting up your environment.
Install via pip
pip install benchwise
Install from Source
git clone https://github.com/Benchwise/benchwise.git
cd benchwise
pip install -e .
Install with Optional Dependencies
Benchwise has several optional dependency groups for different use cases:
# Development tools (includes test + lint)
pip install -e ".[dev]"
# Testing tools
pip install -e ".[test]"
# Linting and formatting tools
pip install -e ".[lint]"
# Evaluation metrics (ROUGE, BLEU, BERT-score, etc.)
pip install -e ".[metrics]"
# LLM API clients (OpenAI, Anthropic, Google)
pip install -e ".[llm-apis]"
# HuggingFace transformers
pip install -e ".[transformers]"
# All LLM dependencies (llm-apis + transformers)
pip install -e ".[llms]"
# All optional dependencies (metrics, llm-apis, transformers, and psutil)
pip install -e ".[all]"
Set Up API Keys
To use LLM providers, you need to set up API keys as environment variables:
export OPENAI_API_KEY="your-openai-key"
export ANTHROPIC_API_KEY="your-anthropic-key"
export GOOGLE_API_KEY="your-google-key"
Verify Installation
Test that everything is installed correctly:
import benchwise
print(benchwise.__version__)
Next Steps
- Quickstart Guide - Run your first evaluation
- Core Concepts - Understand key concepts