label-studio-sdk 1.0.7__py3-none-any.whl → 1.0.10__py3-none-any.whl

label_studio_sdk/core/jsonable_encoder.py

@@ -8,33 +8,27 @@ Taken from FastAPI, and made a bit simpler
 https://github.com/tiangolo/fastapi/blob/master/fastapi/encoders.py
 """
 
+import base64
 import dataclasses
 import datetime as dt
-from collections import defaultdict
 from enum import Enum
 from pathlib import PurePath
 from types import GeneratorType
-from typing import Any, Callable, Dict, List, Optional, Set, Tuple, Union
+from typing import Any, Callable, Dict, List, Optional, Set, Union
+
+import pydantic
 
 from .datetime_utils import serialize_datetime
-from .pydantic_utilities import pydantic_v1
+from .pydantic_utilities import (
+    IS_PYDANTIC_V2,
+    encode_by_type,
+    to_jsonable_with_fallback,
+)
 
 SetIntStr = Set[Union[int, str]]
 DictIntStrAny = Dict[Union[int, str], Any]
 
 
-def generate_encoders_by_class_tuples(
-    type_encoder_map: Dict[Any, Callable[[Any], Any]]
-) -> Dict[Callable[[Any], Any], Tuple[Any, ...]]:
-    encoders_by_class_tuples: Dict[Callable[[Any], Any], Tuple[Any, ...]] = defaultdict(tuple)
-    for type_, encoder in type_encoder_map.items():
-        encoders_by_class_tuples[encoder] += (type_,)
-    return encoders_by_class_tuples
-
-
-encoders_by_class_tuples = generate_encoders_by_class_tuples(pydantic_v1.json.ENCODERS_BY_TYPE)
-
-
 def jsonable_encoder(obj: Any, custom_encoder: Optional[Dict[Any, Callable[[Any], Any]]] = None) -> Any:
     custom_encoder = custom_encoder or {}
     if custom_encoder:
@@ -44,17 +38,24 @@ def jsonable_encoder(obj: Any, custom_encoder: Optional[Dict[Any, Callable[[Any]
             for encoder_type, encoder_instance in custom_encoder.items():
                 if isinstance(obj, encoder_type):
                     return encoder_instance(obj)
-    if isinstance(obj, pydantic_v1.BaseModel):
-        encoder = getattr(obj.__config__, "json_encoders", {})
+    if isinstance(obj, pydantic.BaseModel):
+        if IS_PYDANTIC_V2:
+            encoder = getattr(obj.model_config, "json_encoders", {})  # type: ignore # Pydantic v2
+        else:
+            encoder = getattr(obj.__config__, "json_encoders", {})  # type: ignore # Pydantic v1
         if custom_encoder:
             encoder.update(custom_encoder)
         obj_dict = obj.dict(by_alias=True)
         if "__root__" in obj_dict:
             obj_dict = obj_dict["__root__"]
+        if "root" in obj_dict:
+            obj_dict = obj_dict["root"]
         return jsonable_encoder(obj_dict, custom_encoder=encoder)
     if dataclasses.is_dataclass(obj):
-        obj_dict = dataclasses.asdict(obj)
+        obj_dict = dataclasses.asdict(obj)  # type: ignore
         return jsonable_encoder(obj_dict, custom_encoder=custom_encoder)
+    if isinstance(obj, bytes):
+        return base64.b64encode(obj).decode("utf-8")
     if isinstance(obj, Enum):
         return obj.value
     if isinstance(obj, PurePath):
@@ -80,20 +81,21 @@ def jsonable_encoder(obj: Any, custom_encoder: Optional[Dict[Any, Callable[[Any]
             encoded_list.append(jsonable_encoder(item, custom_encoder=custom_encoder))
         return encoded_list
 
-    if type(obj) in pydantic_v1.json.ENCODERS_BY_TYPE:
-        return pydantic_v1.json.ENCODERS_BY_TYPE[type(obj)](obj)
-    for encoder, classes_tuple in encoders_by_class_tuples.items():
-        if isinstance(obj, classes_tuple):
-            return encoder(obj)
+    def fallback_serializer(o: Any) -> Any:
+        attempt_encode = encode_by_type(o)
+        if attempt_encode is not None:
+            return attempt_encode
 
-    try:
-        data = dict(obj)
-    except Exception as e:
-        errors: List[Exception] = []
-        errors.append(e)
         try:
-            data = vars(obj)
+            data = dict(o)
         except Exception as e:
+            errors: List[Exception] = []
             errors.append(e)
-            raise ValueError(errors) from e
-    return jsonable_encoder(data, custom_encoder=custom_encoder)
+            try:
+                data = vars(o)
+            except Exception as e:
+                errors.append(e)
+                raise ValueError(errors) from e
+        return jsonable_encoder(data, custom_encoder=custom_encoder)
+
+    return to_jsonable_with_fallback(obj, fallback_serializer)

label_studio_sdk/core/pagination.py

@@ -4,11 +4,12 @@ import typing
 
 from typing_extensions import Self
 
-from .pydantic_utilities import pydantic_v1
+import pydantic
 
 # Generic to represent the underlying type of the results within a page
 T = typing.TypeVar("T")
 
+
 # SDKs implement a Page ABC per-pagination request, the endpoint then retuns a pager that wraps this type
 # for example, an endpoint will return SyncPager[UserPage] where UserPage implements the Page ABC. ex:
 #
@@ -18,16 +19,16 @@ T = typing.TypeVar("T")
 #     # This should be the outer function that returns the SyncPager again
 #     get_next=lambda: list(..., cursor: response.cursor) (or list(..., offset: offset + 1))
 # )
-class BasePage(pydantic_v1.BaseModel, typing.Generic[T]):
+class BasePage(pydantic.BaseModel, typing.Generic[T]):
     has_next: bool
     items: typing.Optional[typing.List[T]]
 
 
-class SyncPage(BasePage, typing.Generic[T]):
+class SyncPage(BasePage[T], typing.Generic[T]):
     get_next: typing.Optional[typing.Callable[[], typing.Optional[Self]]]
 
 
-class AsyncPage(BasePage, typing.Generic[T]):
+class AsyncPage(BasePage[T], typing.Generic[T]):
     get_next: typing.Optional[typing.Callable[[], typing.Awaitable[typing.Optional[Self]]]]
 
 

label_studio_sdk/core/pydantic_utilities.py

@@ -1,28 +1,296 @@
 # This file was auto-generated by Fern from our API Definition.
 
+# nopycln: file
+import datetime as dt
 import typing
+from collections import defaultdict
+
+import typing_extensions
 
 import pydantic
 
+from .datetime_utils import serialize_datetime
+from .serialization import convert_and_respect_annotation_metadata
+
 IS_PYDANTIC_V2 = pydantic.VERSION.startswith("2.")
 
 if IS_PYDANTIC_V2:
-    import pydantic.v1 as pydantic_v1  # type: ignore  # nopycln: import
+    # isort will try to reformat the comments on these imports, which breaks mypy
+    # isort: off
+    from pydantic.v1.datetime_parse import (  # type: ignore # pyright: ignore[reportMissingImports] # Pydantic v2
+        parse_date as parse_date,
+    )
+    from pydantic.v1.datetime_parse import (  # pyright: ignore[reportMissingImports] # Pydantic v2
+        parse_datetime as parse_datetime,
+    )
+    from pydantic.v1.json import (  # type: ignore # pyright: ignore[reportMissingImports] # Pydantic v2
+        ENCODERS_BY_TYPE as encoders_by_type,
+    )
+    from pydantic.v1.typing import (  # type: ignore # pyright: ignore[reportMissingImports] # Pydantic v2
+        get_args as get_args,
+    )
+    from pydantic.v1.typing import (  # pyright: ignore[reportMissingImports] # Pydantic v2
+        get_origin as get_origin,
+    )
+    from pydantic.v1.typing import (  # pyright: ignore[reportMissingImports] # Pydantic v2
+        is_literal_type as is_literal_type,
+    )
+    from pydantic.v1.typing import (  # pyright: ignore[reportMissingImports] # Pydantic v2
+        is_union as is_union,
+    )
+    from pydantic.v1.fields import ModelField as ModelField  # type: ignore # pyright: ignore[reportMissingImports] # Pydantic v2
 else:
-    import pydantic as pydantic_v1  # type: ignore  # nopycln: import
+    from pydantic.datetime_parse import parse_date as parse_date  # type: ignore # Pydantic v1
+    from pydantic.datetime_parse import parse_datetime as parse_datetime  # type: ignore # Pydantic v1
+    from pydantic.fields import ModelField as ModelField  # type: ignore # Pydantic v1
+    from pydantic.json import ENCODERS_BY_TYPE as encoders_by_type  # type: ignore # Pydantic v1
+    from pydantic.typing import get_args as get_args  # type: ignore # Pydantic v1
+    from pydantic.typing import get_origin as get_origin  # type: ignore # Pydantic v1
+    from pydantic.typing import is_literal_type as is_literal_type  # type: ignore # Pydantic v1
+    from pydantic.typing import is_union as is_union  # type: ignore # Pydantic v1
+
+    # isort: on
+
+
+T = typing.TypeVar("T")
+Model = typing.TypeVar("Model", bound=pydantic.BaseModel)
+
+
+def parse_obj_as(type_: typing.Type[T], object_: typing.Any) -> T:
+    dealiased_object = convert_and_respect_annotation_metadata(object_=object_, annotation=type_, direction="read")
+    if IS_PYDANTIC_V2:
+        adapter = pydantic.TypeAdapter(type_)  # type: ignore # Pydantic v2
+        return adapter.validate_python(dealiased_object)
+    else:
+        return pydantic.parse_obj_as(type_, dealiased_object)
+
+
+def to_jsonable_with_fallback(
+    obj: typing.Any, fallback_serializer: typing.Callable[[typing.Any], typing.Any]
+) -> typing.Any:
+    if IS_PYDANTIC_V2:
+        from pydantic_core import to_jsonable_python
+
+        return to_jsonable_python(obj, fallback=fallback_serializer)
+    else:
+        return fallback_serializer(obj)
+
+
+class UniversalBaseModel(pydantic.BaseModel):
+    if IS_PYDANTIC_V2:
+        model_config: typing.ClassVar[pydantic.ConfigDict] = pydantic.ConfigDict(
+            # Allow fields begining with `model_` to be used in the model
+            protected_namespaces=(),
+        )  # type: ignore # Pydantic v2
+
+        @pydantic.model_serializer(mode="wrap", when_used="json")  # type: ignore # Pydantic v2
+        def serialize_model(self, handler: pydantic.SerializerFunctionWrapHandler) -> typing.Any:  # type: ignore # Pydantic v2
+            serialized = handler(self)
+            data = {k: serialize_datetime(v) if isinstance(v, dt.datetime) else v for k, v in serialized.items()}
+            return data
+
+    else:
+
+        class Config:
+            smart_union = True
+            json_encoders = {dt.datetime: serialize_datetime}
+
+    @classmethod
+    def model_construct(
+        cls: typing.Type["Model"], _fields_set: typing.Optional[typing.Set[str]] = None, **values: typing.Any
+    ) -> "Model":
+        dealiased_object = convert_and_respect_annotation_metadata(object_=values, annotation=cls, direction="read")
+        return cls.construct(_fields_set, **dealiased_object)
+
+    @classmethod
+    def construct(
+        cls: typing.Type["Model"], _fields_set: typing.Optional[typing.Set[str]] = None, **values: typing.Any
+    ) -> "Model":
+        dealiased_object = convert_and_respect_annotation_metadata(object_=values, annotation=cls, direction="read")
+        if IS_PYDANTIC_V2:
+            return super().model_construct(_fields_set, **dealiased_object)  # type: ignore # Pydantic v2
+        else:
+            return super().construct(_fields_set, **dealiased_object)
+
+    def json(self, **kwargs: typing.Any) -> str:
+        kwargs_with_defaults: typing.Any = {
+            "by_alias": True,
+            "exclude_unset": True,
+            **kwargs,
+        }
+        if IS_PYDANTIC_V2:
+            return super().model_dump_json(**kwargs_with_defaults)  # type: ignore # Pydantic v2
+        else:
+            return super().json(**kwargs_with_defaults)
+
+    def dict(self, **kwargs: typing.Any) -> typing.Dict[str, typing.Any]:
+        """
+        Override the default dict method to `exclude_unset` by default. This function patches
+        `exclude_unset` to work include fields within non-None default values.
+        """
+        # Note: the logic here is multi-plexed given the levers exposed in Pydantic V1 vs V2
+        # Pydantic V1's .dict can be extremely slow, so we do not want to call it twice.
+        #
+        # We'd ideally do the same for Pydantic V2, but it shells out to a library to serialize models
+        # that we have less control over, and this is less intrusive than custom serializers for now.
+        if IS_PYDANTIC_V2:
+            kwargs_with_defaults_exclude_unset: typing.Any = {
+                **kwargs,
+                "by_alias": True,
+                "exclude_unset": True,
+                "exclude_none": False,
+            }
+            kwargs_with_defaults_exclude_none: typing.Any = {
+                **kwargs,
+                "by_alias": True,
+                "exclude_none": True,
+                "exclude_unset": False,
+            }
+            dict_dump = deep_union_pydantic_dicts(
+                super().model_dump(**kwargs_with_defaults_exclude_unset),  # type: ignore # Pydantic v2
+                super().model_dump(**kwargs_with_defaults_exclude_none),  # type: ignore # Pydantic v2
+            )
+
+        else:
+            _fields_set = self.__fields_set__.copy()
+
+            fields = _get_model_fields(self.__class__)
+            for name, field in fields.items():
+                if name not in _fields_set:
+                    default = _get_field_default(field)
+
+                    # If the default values are non-null act like they've been set
+                    # This effectively allows exclude_unset to work like exclude_none where
+                    # the latter passes through intentionally set none values.
+                    if default is not None or ("exclude_unset" in kwargs and not kwargs["exclude_unset"]):
+                        _fields_set.add(name)
+
+                        if default is not None:
+                            self.__fields_set__.add(name)
+
+            kwargs_with_defaults_exclude_unset_include_fields: typing.Any = {
+                "by_alias": True,
+                "exclude_unset": True,
+                "include": _fields_set,
+                **kwargs,
+            }
+
+            dict_dump = super().dict(**kwargs_with_defaults_exclude_unset_include_fields)
+
+        return convert_and_respect_annotation_metadata(object_=dict_dump, annotation=self.__class__, direction="write")
+
+
+def _union_list_of_pydantic_dicts(
+    source: typing.List[typing.Any], destination: typing.List[typing.Any]
+) -> typing.List[typing.Any]:
+    converted_list: typing.List[typing.Any] = []
+    for i, item in enumerate(source):
+        destination_value = destination[i]  # type: ignore
+        if isinstance(item, dict):
+            converted_list.append(deep_union_pydantic_dicts(item, destination_value))
+        elif isinstance(item, list):
+            converted_list.append(_union_list_of_pydantic_dicts(item, destination_value))
+        else:
+            converted_list.append(item)
+    return converted_list
 
 
 def deep_union_pydantic_dicts(
     source: typing.Dict[str, typing.Any], destination: typing.Dict[str, typing.Any]
 ) -> typing.Dict[str, typing.Any]:
     for key, value in source.items():
+        node = destination.setdefault(key, {})
         if isinstance(value, dict):
-            node = destination.setdefault(key, {})
             deep_union_pydantic_dicts(value, node)
+        # Note: we do not do this same processing for sets given we do not have sets of models
+        # and given the sets are unordered, the processing of the set and matching objects would
+        # be non-trivial.
+        elif isinstance(value, list):
+            destination[key] = _union_list_of_pydantic_dicts(value, node)
         else:
             destination[key] = value
 
     return destination
 
 
-__all__ = ["pydantic_v1"]
+if IS_PYDANTIC_V2:
+
+    class V2RootModel(UniversalBaseModel, pydantic.RootModel):  # type: ignore # Pydantic v2
+        pass
+
+    UniversalRootModel: typing_extensions.TypeAlias = V2RootModel  # type: ignore
+else:
+    UniversalRootModel: typing_extensions.TypeAlias = UniversalBaseModel  # type: ignore
+
+
+def encode_by_type(o: typing.Any) -> typing.Any:
+    encoders_by_class_tuples: typing.Dict[typing.Callable[[typing.Any], typing.Any], typing.Tuple[typing.Any, ...]] = (
+        defaultdict(tuple)
+    )
+    for type_, encoder in encoders_by_type.items():
+        encoders_by_class_tuples[encoder] += (type_,)
+
+    if type(o) in encoders_by_type:
+        return encoders_by_type[type(o)](o)
+    for encoder, classes_tuple in encoders_by_class_tuples.items():
+        if isinstance(o, classes_tuple):
+            return encoder(o)
+
+
+def update_forward_refs(model: typing.Type["Model"], **localns: typing.Any) -> None:
+    if IS_PYDANTIC_V2:
+        model.model_rebuild(raise_errors=False)  # type: ignore # Pydantic v2
+    else:
+        model.update_forward_refs(**localns)
+
+
+# Mirrors Pydantic's internal typing
+AnyCallable = typing.Callable[..., typing.Any]
+
+
+def universal_root_validator(
+    pre: bool = False,
+) -> typing.Callable[[AnyCallable], AnyCallable]:
+    def decorator(func: AnyCallable) -> AnyCallable:
+        if IS_PYDANTIC_V2:
+            return pydantic.model_validator(mode="before" if pre else "after")(func)  # type: ignore # Pydantic v2
+        else:
+            return pydantic.root_validator(pre=pre)(func)  # type: ignore # Pydantic v1
+
+    return decorator
+
+
+def universal_field_validator(field_name: str, pre: bool = False) -> typing.Callable[[AnyCallable], AnyCallable]:
+    def decorator(func: AnyCallable) -> AnyCallable:
+        if IS_PYDANTIC_V2:
+            return pydantic.field_validator(field_name, mode="before" if pre else "after")(func)  # type: ignore # Pydantic v2
+        else:
+            return pydantic.validator(field_name, pre=pre)(func)  # type: ignore # Pydantic v1
+
+    return decorator
+
+
+PydanticField = typing.Union[ModelField, pydantic.fields.FieldInfo]
+
+
+def _get_model_fields(
+    model: typing.Type["Model"],
+) -> typing.Mapping[str, PydanticField]:
+    if IS_PYDANTIC_V2:
+        return model.model_fields  # type: ignore # Pydantic v2
+    else:
+        return model.__fields__  # type: ignore # Pydantic v1
+
+
+def _get_field_default(field: PydanticField) -> typing.Any:
+    try:
+        value = field.get_default()  # type: ignore # Pydantic < v1.10.15
+    except:
+        value = field.default
+    if IS_PYDANTIC_V2:
+        from pydantic_core import PydanticUndefined
+
+        if value == PydanticUndefined:
+            return None
+        return value
+    return value

label_studio_sdk/core/query_encoder.py

@@ -1,33 +1,58 @@
 # This file was auto-generated by Fern from our API Definition.
 
-from collections import ChainMap
-from typing import Any, Dict, Optional
+from typing import Any, Dict, List, Optional, Tuple
 
-from .pydantic_utilities import pydantic_v1
+import pydantic
 
 
 # Flattens dicts to be of the form {"key[subkey][subkey2]": value} where value is not a dict
-def traverse_query_dict(dict_flat: Dict[str, Any], key_prefix: Optional[str] = None) -> Dict[str, Any]:
-    result = {}
+def traverse_query_dict(dict_flat: Dict[str, Any], key_prefix: Optional[str] = None) -> List[Tuple[str, Any]]:
+    result = []
     for k, v in dict_flat.items():
         key = f"{key_prefix}[{k}]" if key_prefix is not None else k
         if isinstance(v, dict):
-            result.update(traverse_query_dict(v, key))
+            result.extend(traverse_query_dict(v, key))
+        elif isinstance(v, list):
+            for arr_v in v:
+                if isinstance(arr_v, dict):
+                    result.extend(traverse_query_dict(arr_v, key))
+                else:
+                    result.append((key, arr_v))
         else:
-            result[key] = v
+            result.append((key, v))
     return result
 
 
-def single_query_encoder(query_key: str, query_value: Any) -> Dict[str, Any]:
-    if isinstance(query_value, pydantic_v1.BaseModel) or isinstance(query_value, dict):
-        if isinstance(query_value, pydantic_v1.BaseModel):
+def single_query_encoder(query_key: str, query_value: Any) -> List[Tuple[str, Any]]:
+    if isinstance(query_value, pydantic.BaseModel) or isinstance(query_value, dict):
+        if isinstance(query_value, pydantic.BaseModel):
             obj_dict = query_value.dict(by_alias=True)
         else:
             obj_dict = query_value
         return traverse_query_dict(obj_dict, query_key)
+    elif isinstance(query_value, list):
+        encoded_values: List[Tuple[str, Any]] = []
+        for value in query_value:
+            if isinstance(value, pydantic.BaseModel) or isinstance(value, dict):
+                if isinstance(value, pydantic.BaseModel):
+                    obj_dict = value.dict(by_alias=True)
+                elif isinstance(value, dict):
+                    obj_dict = value
 
-    return {query_key: query_value}
+                encoded_values.extend(single_query_encoder(query_key, obj_dict))
+            else:
+                encoded_values.append((query_key, value))
 
+        return encoded_values
 
-def encode_query(query: Optional[Dict[str, Any]]) -> Optional[Dict[str, Any]]:
-    return dict(ChainMap(*[single_query_encoder(k, v) for k, v in query.items()])) if query is not None else None
+    return [(query_key, query_value)]
+
+
+def encode_query(query: Optional[Dict[str, Any]]) -> Optional[List[Tuple[str, Any]]]:
+    if query is None:
+        return None
+
+    encoded_query = []
+    for k, v in query.items():
+        encoded_query.extend(single_query_encoder(k, v))
+    return encoded_query

label_studio_sdk/core/request_options.py

@@ -23,6 +23,8 @@ class RequestOptions(typing.TypedDict, total=False):
         - additional_query_parameters: typing.Dict[str, typing.Any]. A dictionary containing additional parameters to spread into the request's query parameters dict
 
         - additional_body_parameters: typing.Dict[str, typing.Any]. A dictionary containing additional parameters to spread into the request's body parameters dict
+
+        - chunk_size: int. The size, in bytes, to process each chunk of data being streamed back within the response. This equates to leveraging `chunk_size` within `requests` or `httpx`, and is only leveraged for file downloads.
     """
 
     timeout_in_seconds: NotRequired[int]
@@ -30,3 +32,4 @@ class RequestOptions(typing.TypedDict, total=False):
     additional_headers: NotRequired[typing.Dict[str, typing.Any]]
     additional_query_parameters: NotRequired[typing.Dict[str, typing.Any]]
     additional_body_parameters: NotRequired[typing.Dict[str, typing.Any]]
+    chunk_size: NotRequired[int]