Merge pull request #2042 from weaviate/empty_default_vector_index

Send empty vector index type on new Weavaite versions

Author: dirkkul

.github/workflows/main.yaml

@@ -28,7 +28,7 @@ env:
   WEAVIATE_134: 1.34.19
   WEAVIATE_135: 1.35.18
   WEAVIATE_136: 1.36.12
-  WEAVIATE_137: 1.37.1-4e61e26.amd64
+  WEAVIATE_137: 1.37.5-e0fe0d5.amd64
 
 jobs:
   lint-and-format:

integration/test_collection_config.py

@@ -1025,9 +1025,7 @@ def test_config_export_and_recreate_from_dict(collection_factory: CollectionFact
             Property(name="booleans", data_type=DataType.BOOL_ARRAY),
             Property(name="geo", data_type=DataType.GEO_COORDINATES),
             Property(name="phone", data_type=DataType.PHONE_NUMBER),
-            Property(
-                name="field_index_searchable", data_type=DataType.TEXT, index_searchable=False
-            ),
+            Property(name="field_searchable_off", data_type=DataType.TEXT, index_searchable=False),
             Property(
                 name="field_index_range_filters_false",
                 data_type=DataType.INT,
@@ -1054,7 +1052,9 @@ def test_config_export_and_recreate_from_dict(collection_factory: CollectionFact
                         tokenization=Tokenization.FIELD,
                     ),
                     Property(
-                        name="nested_searchable", data_type=DataType.TEXT, index_searchable=False
+                        name="nested_searchable_off",
+                        data_type=DataType.TEXT,
+                        index_searchable=False,
                     ),
                     Property(
                         name="nested_filterable", data_type=DataType.TEXT, index_filterable=False
@@ -1598,7 +1598,7 @@ def test_replication_config_with_async_config(collection_factory: CollectionFact
             factor=1,
             async_enabled=True,
             async_config=Configure.Replication.async_config(
-                max_workers=8,
+                propagation_concurrency=4,
                 hashtree_height=20,
             ),
         ),
@@ -1608,8 +1608,12 @@ def test_replication_config_with_async_config(collection_factory: CollectionFact
     assert config.replication_config.async_enabled is True
     assert config.replication_config.async_config is not None
     ac = config.replication_config.async_config
-    assert ac.max_workers == 8
+    assert ac.propagation_concurrency == 4
     assert ac.hashtree_height == 20
+    if collection._connection._weaviate_version.is_at_least(1, 37, 3):
+        # Server removed max_workers / alive_nodes_checking_frequency from the schema in 1.37.3
+        assert ac.max_workers is None
+        assert ac.alive_nodes_checking_frequency is None
 
 
 def test_replication_config_remove_async_config_by_disabling_async_replication(
@@ -1624,14 +1628,14 @@ def test_replication_config_remove_async_config_by_disabling_async_replication(
             factor=1,
             async_enabled=True,
             async_config=Configure.Replication.async_config(
-                max_workers=8,
+                propagation_concurrency=4,
                 hashtree_height=20,
             ),
         ),
     )
     config = collection.config.get()
     assert config.replication_config.async_config is not None
-    assert config.replication_config.async_config.max_workers == 8
+    assert config.replication_config.async_config.propagation_concurrency == 4
 
     collection.config.update(
         replication_config=Reconfigure.replication(
@@ -1653,14 +1657,14 @@ def test_replication_config_remove_async_config(collection_factory: CollectionFa
             factor=1,
             async_enabled=True,
             async_config=Configure.Replication.async_config(
-                max_workers=8,
+                propagation_concurrency=4,
                 hashtree_height=20,
             ),
         ),
     )
     config = collection.config.get()
     assert config.replication_config.async_config is not None
-    assert config.replication_config.async_config.max_workers == 8
+    assert config.replication_config.async_config.propagation_concurrency == 4
 
     collection.config.update(
         replication_config=Reconfigure.replication(
@@ -1685,29 +1689,29 @@ def test_replication_config_unset_single_async_field(
             factor=1,
             async_enabled=True,
             async_config=Configure.Replication.async_config(
-                max_workers=8,
+                propagation_concurrency=4,
                 hashtree_height=20,
             ),
         ),
     )
     config = collection.config.get()
     ac = config.replication_config.async_config
     assert ac is not None
-    assert ac.max_workers == 8
+    assert ac.propagation_concurrency == 4
     assert ac.hashtree_height == 20
 
-    # Update with only max_workers — hashtree_height reverts to server default
+    # Update with only propagation_concurrency — hashtree_height reverts to server default
     collection.config.update(
         replication_config=Reconfigure.replication(
             async_config=Reconfigure.Replication.async_config(
-                max_workers=8,
+                propagation_concurrency=4,
             ),
         ),
     )
     config = collection.config.get()
     ac = config.replication_config.async_config
     assert ac is not None
-    assert ac.max_workers == 8
+    assert ac.propagation_concurrency == 4
     assert ac.hashtree_height != 20
 
 
@@ -1734,16 +1738,16 @@ def test_replication_config_add_async_config_to_existing_collection(
     collection.config.update(
         replication_config=Reconfigure.replication(
             async_config=Reconfigure.Replication.async_config(
-                max_workers=8,
                 propagation_concurrency=4,
+                hashtree_height=20,
             ),
         ),
     )
     config = collection.config.get()
     assert config.replication_config.async_config is not None
     ac = config.replication_config.async_config
-    assert ac.max_workers == 8
     assert ac.propagation_concurrency == 4
+    assert ac.hashtree_height == 20
 
 
 def test_update_property_descriptions(collection_factory: CollectionFactory) -> None:

weaviate/collections/classes/config.py

@@ -2473,7 +2473,7 @@ def __add_to_module_config(
         else:
             return_dict["moduleConfig"][addition_key] = addition_val
 
-    def _to_dict(self) -> Dict[str, Any]:
+    def _to_dict(self, *, emit_default_vector_index_type: bool = True) -> Dict[str, Any]:
         ret_dict: Dict[str, Any] = {}
 
         for cls_field in type(self).model_fields:
@@ -2494,14 +2494,23 @@ def _to_dict(self) -> Dict[str, Any]:
                 ret_dict["vectorIndexType"] = val.vector_index_type().value
                 ret_dict[cls_field] = val._to_dict()
             elif isinstance(val, _VectorConfigCreate):
-                ret_dict["vectorConfig"] = {val.name or "default": val._to_dict()}
+                ret_dict["vectorConfig"] = {
+                    val.name or "default": val._to_dict(
+                        emit_default_vector_index_type=emit_default_vector_index_type
+                    )
+                }
             elif (
                 isinstance(val, list)
                 and len(val) > 0
                 and all(isinstance(item, _NamedVectorConfigCreate) for item in val)
             ):
                 val = cast(List[_NamedVectorConfigCreate], val)
-                ret_dict["vectorConfig"] = {item.name: item._to_dict() for item in val}
+                ret_dict["vectorConfig"] = {
+                    item.name: item._to_dict(
+                        emit_default_vector_index_type=emit_default_vector_index_type
+                    )
+                    for item in val
+                }
             elif (
                 isinstance(val, list)
                 and len(val) > 0
@@ -2514,11 +2523,17 @@ def _to_dict(self) -> Dict[str, Any]:
                         raise WeaviateInvalidInputError(
                             "Vector config name must be set when specifying multiple vectors"
                         )
-                    ret_dict["vectorConfig"][item.name] = item._to_dict()
+                    ret_dict["vectorConfig"][item.name] = item._to_dict(
+                        emit_default_vector_index_type=emit_default_vector_index_type
+                    )
             else:
                 assert isinstance(val, _ConfigCreateModel)
                 ret_dict[cls_field] = val._to_dict()
-        if self.vectorIndexConfig is None and "vectorConfig" not in ret_dict:
+        if (
+            self.vectorIndexConfig is None
+            and "vectorConfig" not in ret_dict
+            and emit_default_vector_index_type
+        ):
             ret_dict["vectorIndexType"] = VectorIndexType.HNSW
 
         ret_dict["class"] = self.name
@@ -2583,7 +2598,17 @@ def async_config(
         """Create a configuration object create for async replication settings when creating a collection.
 
         This is only available with WeaviateDB `>=v1.36.0`.
+
+        Note:
+            `max_workers` and `alive_nodes_checking_frequency` were removed from the
+            Weaviate server schema in v1.37.3. Passing them has no effect against any
+            server `>=v1.37.3` (the server silently drops them) and emits a
+            ``DeprecationWarning``. Both arguments will be removed in a future release.
         """
+        if max_workers is not None:
+            _Warnings.async_replication_field_removed_server_side("max_workers")
+        if alive_nodes_checking_frequency is not None:
+            _Warnings.async_replication_field_removed_server_side("alive_nodes_checking_frequency")
         return _AsyncReplicationConfigCreate(
             maxWorkers=max_workers,
             hashtreeHeight=hashtree_height,
@@ -2624,7 +2649,17 @@ def async_config(
         """Create a configuration object for async replication settings when updating a collection.
 
         This is only available with WeaviateDB `>=v1.36.0`.
+
+        Note:
+            `max_workers` and `alive_nodes_checking_frequency` were removed from the
+            Weaviate server schema in v1.37.3. Passing them has no effect against any
+            server `>=v1.37.3` (the server silently drops them) and emits a
+            ``DeprecationWarning``. Both arguments will be removed in a future release.
         """
+        if max_workers is not None:
+            _Warnings.async_replication_field_removed_server_side("max_workers")
+        if alive_nodes_checking_frequency is not None:
+            _Warnings.async_replication_field_removed_server_side("alive_nodes_checking_frequency")
         return _AsyncReplicationConfigUpdate(
             maxWorkers=max_workers,
             hashtreeHeight=hashtree_height,

weaviate/collections/classes/config_named_vectors.py

@@ -77,12 +77,12 @@ class _NamedVectorConfigCreate(_ConfigCreateModel):
         default=None, alias="vector_index_config"
     )
 
-    def _to_dict(self) -> Dict[str, Any]:
+    def _to_dict(self, *, emit_default_vector_index_type: bool = True) -> Dict[str, Any]:
         ret_dict: Dict[str, Any] = self.__parse_vectorizer()
         if self.vectorIndexConfig is not None:
             ret_dict["vectorIndexType"] = self.vectorIndexConfig.vector_index_type().value
             ret_dict["vectorIndexConfig"] = self.vectorIndexConfig._to_dict()
-        else:
+        elif emit_default_vector_index_type:
             ret_dict["vectorIndexType"] = self.vectorIndexType.value
         return ret_dict
 

weaviate/collections/classes/config_vectors.py

@@ -87,12 +87,12 @@ class _VectorConfigCreate(_ConfigCreateModel):
         default=None, alias="vector_index_config"
     )
 
-    def _to_dict(self) -> Dict[str, Any]:
+    def _to_dict(self, *, emit_default_vector_index_type: bool = True) -> Dict[str, Any]:
         ret_dict: Dict[str, Any] = self.__parse_vectorizer()
         if self.vectorIndexConfig is not None:
             ret_dict["vectorIndexType"] = self.vectorIndexConfig.vector_index_type().value
             ret_dict["vectorIndexConfig"] = self.vectorIndexConfig._to_dict()
-        else:
+        elif emit_default_vector_index_type:
             ret_dict["vectorIndexType"] = self.vectorIndexType.value
         return ret_dict
 

weaviate/collections/collections/executor.py

@@ -253,8 +253,16 @@ def create(
                 f"Invalid collection config create parameters: {e}"
             ) from e
 
+        # Servers >= 1.37.5 apply DEFAULT_VECTOR_INDEX_TYPE to named-vector
+        # configs that omit `vectorIndexType`; older servers reject the empty
+        # field, so for them we keep emitting the client-side HNSW default.
+        emit_default_vector_index_type = not self._connection._weaviate_version.is_at_least(
+            1, 37, 5
+        )
         return self.__create(
-            config=config._to_dict(),
+            config=config._to_dict(
+                emit_default_vector_index_type=emit_default_vector_index_type,
+            ),
             data_model_properties=data_model_properties,
             data_model_references=data_model_references,
             skip_argument_validation=skip_argument_validation,

weaviate/warnings.py

@@ -251,6 +251,16 @@ def min_occurrences_metric_deprecated() -> None:
             stacklevel=1,
         )
 
+    @staticmethod
+    def async_replication_field_removed_server_side(argument: str) -> None:
+        warnings.warn(
+            message=f"""Dep029: The `{argument}` argument in `Configure.Replication.async_config` / `Reconfigure.Replication.async_config` is deprecated.
+            It was removed from the Weaviate server schema in v1.37.3 and is silently ignored by newer servers.
+            The argument has no effect against any server >= 1.37.3 and will be removed in a future release.""",
+            category=DeprecationWarning,
+            stacklevel=1,
+        )
+
     @staticmethod
     def datetime_insertion_with_no_specified_timezone(date: datetime) -> None:
         warnings.warn(