Skip to content

armgit / dspy

Public Imported from GitHub
0

DSPy: The framework for programming—not prompting—language models

Python

fix litellm cache test (#8109)

Chen Qian661ddc93f941
diff --git a/tests/predict/test_predict.py b/tests/predict/test_predict.py
index 9384667b1..3461de479 100644
--- a/tests/predict/test_predict.py
+++ b/tests/predict/test_predict.py
@@ -541,36 +541,6 @@ def test_field_constraints(adapter_type):
assert "a multiple of the given number: 2" in system_message
[email protected](os.environ.get("OPENAI_API_KEY") is None, reason="Skipping if OPENAI_API_KEY is not set")
-def test_litellm_cache_initialization_failure():
- """Test that DSPy handles litellm cache initialization failure gracefully."""
- # Mock Cache to raise a permission error
- mock_cache = MagicMock()
- mock_cache.side_effect = PermissionError("Permission denied")
-
- import litellm
-
- # Normal import should have set litellm.cache
- assert litellm.cache is not None
-
- with patch("litellm.caching.Cache", mock_cache):
- import importlib
- import dspy.clients
-
- importlib.reload(dspy.clients)
-
- # On cache initialization failure, litellm.cache should be set to None
- assert litellm.cache is None
-
- # Create a simple predictor to verify it works without cache
- predictor = Predict("question -> answer")
- dspy.settings.configure(lm=dspy.LM(model="openai/gpt-4o-mini"))
-
- # No exception should be raised when litellm.cache is None even if we try to use the cache.
- assert dspy.settings.lm.cache == True
- predictor(question="test")
-
-
@pytest.mark.asyncio
async def test_async_predict():
program = Predict("question -> answer")