Ollama

Run LLM tests locally — no API key, no cost.

pip install pytest-llmtest[ollama]
ollama pull llama3.2
ollama serve

Usage

@llm_test(
    expect.is_not_empty(),
    provider="ollama",
    model="llama3.2",
)
def test_local(llm):
    output = llm("What is 2+2?")
    assert "4" in output.content

Supported models

Any model in your local Ollama: llama3.2, mistral, codellama, phi3, gemma2, deepseek-r1

ollama list  # see available models

Custom host

export OLLAMA_HOST=http://my-server:11434

When to use

• Local development — no API costs, no rate limits
• CI/CD — run tests without API keys
• Privacy — data never leaves your machine
• Offline — no internet needed