We use Poetry for dependency management. Please make sure that you have installed Poetry and set up the environment correctly before starting development.
-
Install dependencies from the lock file:
poetry install
-
Use the environment: You can either run commands directly with
poetry run <command>
or open a shell withpoetry shell
and then run commands directly.
- copy
.env.template
and rename to.env
- in
.env
, set api key and model for the desired LLM (OpenAI, Gemini or Claude), such as
GEMINI_15_API_KEY=AIxxx
GEMINI_15_MODEL=gemini-1.5-pro-latest
- in
benchmark/test_benchmark_with_llm.py
, select desired assessing LLM
@pytest.fixture
def client():
return GeminiClient() # GptClient, GeminiClient and ClaudeClient are available
We can run benchmark either by execute poetry run pytest benchmark/test_benchmark_with_llm.py
directly or execute pytest benchmark/test_benchmark_with_llm.py
in poetry shell.
After benchmark completed, we can find the results in benchmark-result.log