lib/crewai/tests/rag/embeddings/test_voyageai_embedding_callable.py (1)

30-49: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

Cover the single-document multi-chunk shape directly.

This test only proves concatenation across result.results. The new behavior that matters here is flattening result.results[*].embeddings when one input auto-chunks into multiple vectors, so a fixture with a single result containing two embeddings would make the regression target much clearer.

Suggested test tweak

     mock_client.contextualized_embed.return_value = MagicMock(
         results=[
-            MagicMock(embeddings=[[0.1, 0.2]]),
-            MagicMock(embeddings=[[0.3, 0.4]]),
+            MagicMock(embeddings=[[0.1, 0.2], [0.3, 0.4]]),
         ]
     )
@@
-    result = fn(["aa", "bb"])
+    result = fn(["aa"])
@@
     mock_client.embed.assert_not_called()
     mock_client.contextualized_embed.assert_called_once()
     assert np.allclose(result, [[0.1, 0.2], [0.3, 0.4]])

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@lib/crewai/tests/rag/embeddings/test_voyageai_embedding_callable.py` around
lines 30 - 49, Update the contextualized embedding test in
VoyageAIEmbeddingFunction so it directly covers the single-document multi-chunk
case: keep the existing VoyageAIEmbeddingFunction and contextualized_embed
setup, but make the mocked contextualized_embed response return one result whose
embeddings contains multiple vectors, then assert the callable flattens
result.results[*].embeddings into the final output. This should replace the
current multi-result concatenation check while still verifying embed is not
called and contextualized_embed is used.

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

lib/crewai/tests/rag/embeddings/test_voyageai_embedding_callable.py (1)

51-107: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

Add a regression test for input_type on voyage-context*.

These tests cover wrapping and flattening, but they never assert what gets passed when the function is configured with input_type="query". That gap is why the current hardcoded "document" behavior can slip through.

🧪 Minimal test to add

+    def test_contextualized_model_forwards_input_type(self):
+        with patch("voyageai.Client") as mock_client_class:
+            mock_client = MagicMock()
+            mock_client_class.return_value = mock_client
+            mock_client.contextualized_embed.return_value = MagicMock(
+                results=[MagicMock(embeddings=[[0.1, 0.2]])]
+            )
+
+            fn = VoyageAIEmbeddingFunction(
+                api_key="voyage-key",
+                model="voyage-context-4",
+                input_type="query",
+            )
+            fn(["aa"])
+
+            _, kwargs = mock_client.contextualized_embed.call_args
+            assert kwargs["input_type"] == "query"

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@lib/crewai/tests/rag/embeddings/test_voyageai_embedding_callable.py` around
lines 51 - 107, Add a regression test around VoyageAIEmbeddingFunction for
voyage-context* models when configured with input_type="query"; the current
contextualized embedding tests only cover document-style wrapping and miss this
branch. Extend the existing test module by creating a case that instantiates
VoyageAIEmbeddingFunction with input_type="query", invokes it, and asserts the
mocked voyageai.Client.contextualized_embed call receives the query-specific
input_type rather than the hardcoded document value. Keep the assertions aligned
with the existing contextualized_embed call_args checks so the new test clearly
guards the input_type wiring.