Skip to navigation

nemoguardrails.testing.fixtures

View as MarkdownOpen in Claude

Pytest fixtures exposing the public testing surface.

To opt in, add the following to your project’s conftest.py:

pytest_plugins = ["nemoguardrails.testing.fixtures"]

The fixtures below provide reasonable defaults that cover the most common testing patterns: a scriptable FakeLLMModel and a TestChat factory that wires the fake model into an LLMRails app.

Module Contents

Functions

NameDescription
fake_llmReturn a FakeLLMModel with a single "Hello!" response.
make_fake_llmReturn a factory for building FakeLLMModel instances.
make_test_chatReturn a factory that builds TestChat instances bound to a

API

nemoguardrails.testing.fixtures.fake_llm() -> nemoguardrails.testing.fake_model.FakeLLMModel

Return a FakeLLMModel with a single "Hello!" response.

Override the responses by overriding this fixture in your own conftest, or instantiate FakeLLMModel directly when you need custom behaviour:

from nemoguardrails.testing import FakeLLMModel
llm = FakeLLMModel(responses=["My scripted answer"])
nemoguardrails.testing.fixtures.make_fake_llm() -> typing.Callable[..., nemoguardrails.testing.fake_model.FakeLLMModel]

Return a factory for building FakeLLMModel instances.

Useful when a single test needs more than one fake model, or when the set of responses is computed dynamically:

def test_two_calls(make_fake_llm):
llm_one = make_fake_llm(responses=["one"])
llm_two = make_fake_llm(responses=["two"])
nemoguardrails.testing.fixtures.make_test_chat() -> typing.Callable[..., nemoguardrails.testing.chat_harness.TestChat]

Return a factory that builds TestChat instances bound to a fake LLM.

Example:

def test_greeting(make_test_chat):
config = RailsConfig.from_path("./examples/bots/abc")
chat = make_test_chat(config, llm_completions=["Hi there!"])
chat.user("hi")
chat.bot("Hi there!")