fix broken tests (#706)

Author: mahesh-bespoke

tests/code_executor/test_code_execution.py

@@ -32,6 +32,7 @@ def code_output(self, row, exec_output):
 @pytest.mark.asyncio
 async def test_simple_code_execution_ray():
     """Test simple code execution with basic input/output."""
+    pytest.importorskip("ray")
 
     # Initialize backend
     class TestCodeExecutor(curator.CodeExecutor):

tests/conftest.py

@@ -7,6 +7,7 @@
 from datasets import Dataset
 
 os.environ["TELEMETRY_ENABLED"] = "false"
+os.environ["CURATOR_VIEWER"] = "false"
 mode = os.environ.get("VCR_MODE", None)
 _KEY_MAP = {
     "openai": "OPENAI_API_KEY",
@@ -16,6 +17,10 @@
     "vllm": "VLLM_API_KEY",
     "gemini": "GEMINI_API_KEY",
 }
+# Map backend names to directory names (to avoid namespace collisions with packages)
+_DIR_MAP = {
+    "vllm": "vllm_backend",
+}
 
 
 @pytest.fixture
@@ -26,16 +31,18 @@ def temp_working_dir(request):
         os.environ[_KEY_MAP[backend.split("/")[-1]]] = "sk-mocked-**"
         backend = backend.split("/")[0]
     os.environ["HF_DATASETS_CACHE"] = "/dev/null"
-    temp_working_dir = f"tests/integrations/{backend}/fixtures/.test_cache"
+    # Use directory mapping to avoid namespace collisions with installed packages
+    backend_dir = _DIR_MAP.get(backend, backend)
+    temp_working_dir = f"tests/integrations/{backend_dir}/fixtures/.test_cache"
 
     if cached_working_dir:
-        working_dir_zip = f"tests/integrations/{backend}/fixtures/.test_cache.zip"
+        working_dir_zip = f"tests/integrations/{backend_dir}/fixtures/.test_cache.zip"
         with zipfile.ZipFile(working_dir_zip, "r") as zip_ref:
             zip_ref.extractall(os.path.split(temp_working_dir)[0])
 
     vcr_path = request.param.get("vcr_path", None)
     if not vcr_path:
-        vcr_path = f"tests/integrations/{backend}/fixtures"
+        vcr_path = f"tests/integrations/{backend_dir}/fixtures"
 
     vcr_config = vcr.VCR(
         serializer="yaml",

tests/integrations/test_all.py

@@ -565,13 +565,14 @@ def test_failed_requests_file_in_cache(temp_working_dir, mock_dataset):
 @pytest.mark.parametrize("temp_working_dir", ([{"integration": "vllm"}]), indirect=True)
 def test_basic_offline(temp_working_dir, mock_dataset):
     """Test basic completion with VLLM backend"""
+    pytest.importorskip("vllm")
     temp_working_dir, _, _ = temp_working_dir
 
     import json
     import os
 
     # Load mock responses from fixture file
-    fixture_path = os.path.join(os.path.dirname(__file__), "vllm", "fixtures", "basic_responses.json")
+    fixture_path = os.path.join(os.path.dirname(__file__), "vllm_backend", "fixtures", "basic_responses.json")
     with open(fixture_path) as f:
         mock_responses = json.load(f)
 

tests/integrations/test_push_to_viewer.py

@@ -1,10 +1,22 @@
 import os
 
+import pytest
 import vcr
 
 from bespokelabs.curator.utils import load_dataset, push_to_viewer
 
 
+@pytest.fixture(autouse=True)
+def reset_curator_viewer_env():
+    """Reset CURATOR_VIEWER to false after each test to avoid affecting other tests."""
+    original_value = os.environ.get("CURATOR_VIEWER")
+    yield
+    if original_value is not None:
+        os.environ["CURATOR_VIEWER"] = original_value
+    else:
+        os.environ.pop("CURATOR_VIEWER", None)
+
+
 def test_smoke(mock_dataset):
     vcr_path = "tests/integrations/common_fixtures"
     os.environ["CURATOR_VIEWER"] = "true"

…tions/vllm/fixtures/basic_responses.json …lm_backend/fixtures/basic_responses.jsontests/integrations/vllm/fixtures/basic_responses.json renamed to tests/integrations/vllm_backend/fixtures/basic_responses.json tests/integrations/vllm/fixtures/basic_responses.json renamed to tests/integrations/vllm_backend/fixtures/basic_responses.json

@@ -191,23 +191,26 @@ def test_disable_cache(tmp_path, temp_working_dir):
 
     os.environ["CURATOR_DISABLE_CACHE"] = "true"
 
-    prompter = curator.LLM(model_name="gpt-4o-mini")
+    try:
+        prompter = curator.LLM(model_name="gpt-4o-mini")
 
-    # Run twice and store results
-    with vcr_config.use_cassette("basic_diff_cache.yaml"):
-        result1 = prompter(["Say '1'. Do not explain."], working_dir=str(tmp_path)).dataset
+        # Run twice and store results
+        with vcr_config.use_cassette("basic_diff_cache.yaml"):
+            result1 = prompter(["Say '1'. Do not explain."], working_dir=str(tmp_path)).dataset
 
-    with vcr_config.use_cassette("basic_diff_cache.yaml"):
-        result2 = prompter(["Say '1'. Do not explain."], working_dir=str(tmp_path)).dataset
+        with vcr_config.use_cassette("basic_diff_cache.yaml"):
+            result2 = prompter(["Say '1'. Do not explain."], working_dir=str(tmp_path)).dataset
 
-    # Verify both runs produced the expected output
-    assert result1.to_pandas().iloc[0]["response"] == "1"
-    assert result2.to_pandas().iloc[0]["response"] == "1"
+        # Verify both runs produced the expected output
+        assert result1.to_pandas().iloc[0]["response"] == "1"
+        assert result2.to_pandas().iloc[0]["response"] == "1"
 
-    # Check cache directory, excluding metadata.db
-    cache_dirs = [d for d in tmp_path.glob("*") if d.name != "metadata.db"]
+        # Check cache directory, excluding metadata.db
+        cache_dirs = [d for d in tmp_path.glob("*") if d.name != "metadata.db"]
 
-    # Should have exactly 2 different cache directories
-    assert len(cache_dirs) == 2
-    # All directories should be different (no duplicates)
-    assert len({str(d) for d in cache_dirs}) == 2
+        # Should have exactly 2 different cache directories
+        assert len(cache_dirs) == 2
+        # All directories should be different (no duplicates)
+        assert len({str(d) for d in cache_dirs}) == 2
+    finally:
+        os.environ.pop("CURATOR_DISABLE_CACHE", None)