fix: Add metadata filtering for Chroma components (#9137)

* 🐛 (chroma.py, local_db.py): Handle ImportError when importing filter_complex_metadata function to avoid crashing the application
💡 (chroma.py, local_db.py): Add try-except block to gracefully handle ImportError and log a warning message if the function cannot be imported

* ♻️ (local_db.py): remove unnecessary try-except block and simplify document adding process in LocalDBComponent

* 🔧 (chroma.py): Remove unused import 'filter_complex_metadata' and add support for filtering complex metadata to prevent ChromaDB errors
🔧 (test_chroma_vector_store_component.py): Add test cases for filtering complex metadata, preserving simple metadata types, handling single file upload scenario, fallback behavior when import fails, and handling empty and None metadata values.

* [autofix.ci] apply automated fixes

* 🐛 (test_chroma_vector_store_component.py): fix failing test due to None value being filtered out by ChromaDB metadata handling

* 🔧 (test_chroma_vector_store_component.py): refactor test method names and data to improve clarity and consistency in metadata filtering and preservation logic

* 🐛 (test_chroma_vector_store_component.py): fix missing variable assignment for error message in mocked import error to improve error handling and debugging in tests

* [autofix.ci] apply automated fixes

* 🔧 (test_chroma_vector_store_component.py): delete the file test_chroma_vector_store_component.py as it is no longer needed in the project

* [autofix.ci] apply automated fixes

* ✨ (test_chroma_vector_store_component.py): add unit tests for ChromaVectorStoreComponent including create_db, create_collection_with_data, similarity_search, mmr_search, search_with_different_types, search_with_score, duplicate_handling, chroma_collection_to_data, metadata_filtering_with_complex_data, metadata_filtering_fallback

---------

Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
Co-authored-by: Edwin Jose <edwin.jose@datastax.com>

src/backend/base/langflow/components/vectorstores/chroma.py

@@ -154,6 +154,14 @@ class ChromaVectorStoreComponent(LCVectorStoreComponent):
 
         if documents and self.embedding is not None:
             self.log(f"Adding {len(documents)} documents to the Vector Store.")
-            vector_store.add_documents(documents)
+            # Filter complex metadata to prevent ChromaDB errors
+            try:
+                from langchain_community.vectorstores.utils import filter_complex_metadata
+
+                filtered_documents = filter_complex_metadata(documents)
+                vector_store.add_documents(filtered_documents)
+            except ImportError:
+                self.log("Warning: Could not import filter_complex_metadata. Adding documents without filtering.")
+                vector_store.add_documents(documents)
         else:
             self.log("No documents to add to the Vector Store.")

src/backend/tests/unit/components/vectorstores/test_chroma_vector_store_component.py

@@ -318,3 +318,84 @@ class TestChromaVectorStoreComponent(ComponentTestBaseWithoutClient):
 
         # Verify the conversion
         assert len(data_objects) == 0
+
+    def test_metadata_filtering_with_complex_data(
+        self, component_class: type[ChromaVectorStoreComponent], default_kwargs: dict[str, Any]
+    ) -> None:
+        """Test that complex metadata is properly filtered and simple types are preserved."""
+        from langflow.base.vectorstores.utils import chroma_collection_to_data
+
+        # Create test data that covers the original error scenario and validation
+        test_data = [
+            Data(
+                data={
+                    "text": "Document with mixed metadata",
+                    "files": [],  # This empty list was causing the original ChromaDB error
+                    "tags": ["tag1", "tag2"],  # Lists should be filtered out
+                    "nested": {"key": "value"},  # Nested objects should be filtered out
+                    "simple_string": "preserved",
+                    "simple_int": 42,
+                    "simple_bool": True,
+                    "empty_string": "",  # Edge case: empty but valid
+                    "zero_value": 0,  # Edge case: falsy but valid
+                }
+            )
+        ]
+
+        default_kwargs["ingest_data"] = test_data
+        default_kwargs["collection_name"] = "test_metadata_filtering"
+
+        # This should not raise an error despite the complex metadata
+        component: ChromaVectorStoreComponent = component_class().set(**default_kwargs)
+        vector_store = component.build_vector_store()
+
+        # Verify document was added successfully
+        collection_dict = vector_store.get()
+        assert len(collection_dict["documents"]) == 1
+        assert "Document with mixed metadata" in collection_dict["documents"][0]
+
+        # Verify metadata filtering: simple types preserved, complex types filtered out
+        data_objects = chroma_collection_to_data(collection_dict)
+        data_obj = data_objects[0]
+
+        # Simple types should be preserved
+        assert data_obj.data["simple_string"] == "preserved"
+        assert data_obj.data["simple_int"] == 42
+        assert data_obj.data["simple_bool"] is True
+        assert data_obj.data["empty_string"] == ""
+        assert data_obj.data["zero_value"] == 0
+
+        # Complex types should be filtered out
+        assert "files" not in data_obj.data
+        assert "tags" not in data_obj.data
+        assert "nested" not in data_obj.data
+
+    def test_metadata_filtering_fallback(
+        self, component_class: type[ChromaVectorStoreComponent], default_kwargs: dict[str, Any], monkeypatch
+    ) -> None:
+        """Test the fallback behavior when filter_complex_metadata import fails."""
+        import builtins
+
+        original_import = builtins.__import__
+
+        def mock_import(name, *args, **kwargs):
+            if name == "langchain_community.vectorstores.utils":
+                error_msg = "Mocked import error"
+                raise ImportError(error_msg)
+            return original_import(name, *args, **kwargs)
+
+        monkeypatch.setattr(builtins, "__import__", mock_import)
+
+        # Use simple test data to avoid ChromaDB errors when filtering is unavailable
+        test_data = [Data(data={"text": "Simple document", "simple_field": "simple_value"})]
+        default_kwargs["ingest_data"] = test_data
+        default_kwargs["collection_name"] = "test_fallback"
+
+        # Should work with fallback (no filtering)
+        component: ChromaVectorStoreComponent = component_class().set(**default_kwargs)
+        vector_store = component.build_vector_store()
+
+        # Verify document was added
+        collection_dict = vector_store.get()
+        assert len(collection_dict["documents"]) == 1
+        assert "Simple document" in collection_dict["documents"][0]