from __future__ import annotations
import dataclasses
import json
import logging
import typing as t
import attr
from starlette.requests import Request
from starlette.responses import Response
from ...exceptions import BadInput
from ...exceptions import InvalidArgument
from ..service.openapi import REF_PREFIX
from ..service.openapi import SUCCESS_DESCRIPTION
from ..service.openapi.specification import MediaType
from ..service.openapi.specification import Schema
from ..types import LazyType
from ..utils import LazyLoader
from ..utils import bentoml_cattr
from ..utils.http import set_cookies
from ..utils.pkg import pkg_version_info
from .base import IODescriptor
EXC_MSG = "'pydantic' must be installed to use 'pydantic_model'. Install with 'pip install bentoml[io-json]'."
if t.TYPE_CHECKING:
from types import UnionType
import pydantic
import pydantic.schema as schema
if pkg_version_info("pydantic")[0] >= 2:
import pydantic.json_schema as jschema
from google.protobuf import message as _message
from google.protobuf import struct_pb2
from typing_extensions import Self
from ..context import ServiceContext as Context
from .base import OpenAPIResponse
else:
pydantic = LazyLoader("pydantic", globals(), "pydantic", exc_msg=EXC_MSG)
schema = LazyLoader("schema", globals(), "pydantic.schema", exc_msg=EXC_MSG)
jschema = LazyLoader(
"jschema", globals(), "pydantic.json_schema", exc_msg="Pydantic v2 is required."
)
# lazy load our proto generated.
struct_pb2 = LazyLoader("struct_pb2", globals(), "google.protobuf.struct_pb2")
# lazy load numpy for processing ndarray.
np = LazyLoader("np", globals(), "numpy")
JSONType = t.Union[str, t.Dict[str, t.Any], "pydantic.BaseModel", None]
logger = logging.getLogger(__name__)
class DefaultJsonEncoder(json.JSONEncoder):
def default(self, o: type) -> t.Any:
if dataclasses.is_dataclass(o):
return dataclasses.asdict(o)
if LazyType["ext.NpNDArray"]("numpy.ndarray").isinstance(o):
return o.tolist()
if LazyType["ext.NpGeneric"]("numpy.generic").isinstance(o):
return o.item()
if LazyType["ext.PdDataFrame"]("pandas.DataFrame").isinstance(o):
return o.to_dict() # type: ignore
if LazyType["ext.PdSeries"]("pandas.Series").isinstance(o):
return o.to_dict() # type: ignore
if LazyType["pydantic.BaseModel"]("pydantic.BaseModel").isinstance(o):
obj_dict = o.dict()
if "__root__" in obj_dict:
obj_dict = obj_dict.get("__root__")
return obj_dict
if attr.has(o):
return bentoml_cattr.unstructure(o)
return super().default(o)
[docs]class JSON(
IODescriptor[JSONType], descriptor_id="bentoml.io.JSON", proto_fields=("json",)
):
"""
:obj:`JSON` defines API specification for the inputs/outputs of a Service, where either
inputs will be converted to or outputs will be converted from a JSON representation
as specified in your API function signature.
A sample service implementation:
.. code-block:: python
:caption: `service.py`
from __future__ import annotations
import typing
from typing import TYPE_CHECKING
from typing import Any
from typing import Optional
import bentoml
from bentoml.io import NumpyNdarray
from bentoml.io import JSON
import numpy as np
import pandas as pd
from pydantic import BaseModel
iris_clf_runner = bentoml.sklearn.get("iris_clf_with_feature_names:latest").to_runner()
svc = bentoml.Service("iris_classifier_pydantic", runners=[iris_clf_runner])
class IrisFeatures(BaseModel):
sepal_len: float
sepal_width: float
petal_len: float
petal_width: float
# Optional field
request_id: Optional[int]
# Use custom Pydantic config for additional validation options
class Config:
extra = 'forbid'
input_spec = JSON(pydantic_model=IrisFeatures)
@svc.api(input=input_spec, output=NumpyNdarray())
def classify(input_data: IrisFeatures) -> NDArray[Any]:
if input_data.request_id is not None:
print("Received request ID: ", input_data.request_id)
input_df = pd.DataFrame([input_data.dict(exclude={"request_id"})])
return iris_clf_runner.run(input_df)
Users then can then serve this service with :code:`bentoml serve`:
.. code-block:: bash
% bentoml serve ./service.py:svc --reload
Users can then send requests to the newly started services with any client:
.. tab-set::
.. tab-item:: Bash
.. code-block:: bash
% curl -X POST -H "content-type: application/json" \\
--data '{"sepal_len": 6.2, "sepal_width": 3.2, "petal_len": 5.2, "petal_width": 2.2}' \\
http://127.0.0.1:3000/classify
# [2]%
.. tab-item:: Python
.. code-block:: python
:caption: `request.py`
import requests
requests.post(
"http://0.0.0.0:3000/predict",
headers={"content-type": "application/json"},
data='{"sepal_len": 6.2, "sepal_width": 3.2, "petal_len": 5.2, "petal_width": 2.2}'
).text
Args:
pydantic_model: Pydantic model schema. When used, inference API callback
will receive an instance of the specified ``pydantic_model`` class.
json_encoder: JSON encoder class. By default BentoML implements a custom JSON encoder that
provides additional serialization supports for numpy arrays, pandas dataframes,
dataclass-like (`attrs <https://www.attrs.org/en/stable/>`_, dataclass, etc.).
If you wish to use a custom encoder, make sure to support the aforementioned object.
Returns:
:obj:`JSON`: IO Descriptor that represents JSON format.
"""
# default mime type is application/json
_mime_type = "application/json"
def __init__(
self,
*,
pydantic_model: type[pydantic.BaseModel] | None = None,
validate_json: bool | None = None,
json_encoder: type[json.JSONEncoder] = DefaultJsonEncoder,
):
if pydantic_model is not None:
assert issubclass(
pydantic_model, pydantic.BaseModel
), "'pydantic_model' must be a subclass of 'pydantic.BaseModel'."
self._pydantic_model = pydantic_model
self._json_encoder = json_encoder
# Remove validate_json in version 1.0.2
if validate_json is not None:
logger.warning(
"'validate_json' option from 'bentoml.io.JSON' has been deprecated. Use a Pydantic model to specify validation options instead."
)
def _from_sample(self, sample: JSONType) -> JSONType:
"""
Create a :class:`~bentoml._internal.io_descriptors.json.JSON` IO Descriptor from given inputs.
Args:
sample: A JSON-like datatype, which can be either dict, str, list.
``sample`` will also accepting a Pydantic model.
.. code-block:: python
from pydantic import BaseModel
class IrisFeatures(BaseModel):
sepal_len: float
sepal_width: float
petal_len: float
petal_width: float
input_spec = JSON.from_sample(
IrisFeatures(sepal_len=1.0, sepal_width=2.0, petal_len=3.0, petal_width=4.0)
)
@svc.api(input=input_spec, output=NumpyNdarray())
async def predict(input: NDArray[np.int16]) -> NDArray[Any]:
return await runner.async_run(input)
json_encoder: Optional JSON encoder.
Returns:
:class:`~bentoml._internal.io_descriptors.json.JSON`: IODescriptor from given users inputs.
Example:
.. code-block:: python
:caption: `service.py`
from __future__ import annotations
import bentoml
from typing import Any
from bentoml.io import JSON
input_spec = JSON.from_sample({"Hello": "World", "foo": "bar"})
@svc.api(input=input_spec, output=JSON())
async def predict(input: dict[str, Any]) -> dict[str, Any]:
return await runner.async_run(input)
Raises:
:class:`BadInput`: Given sample is not a valid JSON string, bytes, or supported nest types.
"""
if LazyType["pydantic.BaseModel"]("pydantic.BaseModel").isinstance(sample):
self._pydantic_model = sample.__class__
elif isinstance(sample, str):
try:
sample = json.loads(sample)
except json.JSONDecodeError as e:
raise BadInput(
f"Unable to parse JSON string. Please make sure the input is a valid JSON string: {e}"
) from None
elif isinstance(sample, bytes):
try:
sample = json.loads(sample.decode())
except json.JSONDecodeError as e:
raise BadInput(
f"Unable to parse JSON bytes. Please make sure the input is a valid JSON bytes: {e}"
) from None
elif not isinstance(sample, (dict, list)):
raise BadInput(
f"Unable to infer JSON type from sample: {sample}. Please make sure the input is a valid JSON object."
)
return sample
def to_spec(self) -> dict[str, t.Any]:
return {
"id": self.descriptor_id,
"args": {
"has_pydantic_model": self._pydantic_model is not None,
"has_json_encoder": self._json_encoder is not DefaultJsonEncoder,
},
}
@classmethod
def from_spec(cls, spec: dict[str, t.Any]) -> Self:
if "args" not in spec:
raise InvalidArgument(f"Missing args key in JSON spec: {spec}")
if "has_pydantic_model" in spec["args"] and spec["args"]["has_pydantic_model"]:
logger.warning(
"BentoML does not support loading pydantic models from URLs; output will be a normal dictionary."
)
if "has_json_encoder" in spec["args"] and spec["args"]["has_json_encoder"]:
logger.warning(
"BentoML does not support loading JSON encoders from URLs; output will be a normal dictionary."
)
return cls()
def input_type(self) -> UnionType:
return JSONType
def openapi_schema(self) -> Schema:
if not self._pydantic_model:
return Schema(type="object")
# returns schemas from pydantic_model.
if pkg_version_info("pydantic")[0] >= 2:
json_schema = jschema.model_json_schema(
self._pydantic_model, ref_template=REF_PREFIX + "{model}"
)
# NOTE: we don't need def here, as these will be available in openapi.components.
if "$defs" in json_schema:
json_schema.pop("$defs", None)
return Schema(**json_schema)
else:
return Schema(
**schema.model_process_schema(
self._pydantic_model,
model_name_map=schema.get_model_name_map(
schema.get_flat_models_from_model(self._pydantic_model)
),
ref_prefix=REF_PREFIX,
)[0]
)
def openapi_components(self) -> dict[str, t.Any] | None:
if not self._pydantic_model:
return {}
from ..service.openapi.utils import pydantic_components_schema
return {"schemas": pydantic_components_schema(self._pydantic_model)}
def openapi_example(self):
if self.sample is not None:
if LazyType["pydantic.BaseModel"]("pydantic.BaseModel").isinstance(
self.sample
):
if pkg_version_info("pydantic")[0] >= 2:
return self.sample.model_dump()
else:
return self.sample.dict()
elif isinstance(self.sample, (str, list)):
return json.dumps(
self.sample,
cls=self._json_encoder,
ensure_ascii=False,
allow_nan=False,
indent=None,
separators=(",", ":"),
)
elif isinstance(self.sample, dict):
return self.sample
def openapi_request_body(self) -> dict[str, t.Any]:
return {
"content": {
self._mime_type: MediaType(
schema=self.openapi_schema(), example=self.openapi_example()
)
},
"required": True,
"x-bentoml-io-descriptor": self.to_spec(),
}
def openapi_responses(self) -> OpenAPIResponse:
return {
"description": SUCCESS_DESCRIPTION,
"content": {
self._mime_type: MediaType(
schema=self.openapi_schema(), example=self.openapi_example()
)
},
"x-bentoml-io-descriptor": self.to_spec(),
}
[docs] async def from_http_request(self, request: Request) -> JSONType:
json_str = await request.body()
try:
json_obj = json.loads(json_str)
except json.JSONDecodeError as e:
raise BadInput(f"Invalid JSON input received: {e}") from None
if self._pydantic_model:
try:
if pkg_version_info("pydantic")[0] >= 2:
pydantic_model = self._pydantic_model.model_validate(json_obj)
else:
pydantic_model = self._pydantic_model.parse_obj(json_obj)
return pydantic_model
except pydantic.ValidationError as e:
raise BadInput(f"Invalid JSON input received: {e}") from None
else:
return json_obj
[docs] async def to_http_response(
self, obj: JSONType | pydantic.BaseModel, ctx: Context | None = None
):
# This is to prevent cases where custom JSON encoder is used.
if LazyType["pydantic.BaseModel"]("pydantic.BaseModel").isinstance(obj):
if pkg_version_info("pydantic")[0] >= 2:
obj = obj.model_dump()
else:
obj = obj.dict()
json_str = (
json.dumps(
obj,
cls=self._json_encoder,
ensure_ascii=False,
allow_nan=False,
indent=None,
separators=(",", ":"),
)
if obj is not None
else None
)
if ctx is not None:
res = Response(
json_str,
media_type=self._mime_type,
headers=ctx.response.metadata, # type: ignore (bad starlette types)
status_code=ctx.response.status_code,
)
set_cookies(res, ctx.response.cookies)
return res
else:
return Response(json_str, media_type=self._mime_type)
[docs] async def from_proto(self, field: struct_pb2.Value | bytes) -> JSONType:
from google.protobuf.json_format import MessageToDict
if isinstance(field, bytes):
content = field
if self._pydantic_model:
try:
if pkg_version_info("pydantic")[0] >= 2:
return self._pydantic_model.model_validate_json(
json.loads(content)
)
else:
return self._pydantic_model.parse_raw(content)
except pydantic.ValidationError as e:
raise BadInput(f"Invalid JSON input received: {e}") from None
try:
parsed = json.loads(content)
except json.JSONDecodeError as e:
raise BadInput(f"Invalid JSON input received: {e}") from None
else:
assert isinstance(field, struct_pb2.Value)
parsed = MessageToDict(field, preserving_proto_field_name=True)
if self._pydantic_model:
try:
if pkg_version_info("pydantic")[0] >= 2:
return self._pydantic_model.model_validate(parsed)
else:
return self._pydantic_model.parse_obj(parsed)
except pydantic.ValidationError as e:
raise BadInput(f"Invalid JSON input received: {e}") from None
return parsed
[docs] async def to_proto(self, obj: JSONType) -> struct_pb2.Value:
if LazyType["pydantic.BaseModel"]("pydantic.BaseModel").isinstance(obj):
if pkg_version_info("pydantic")[0] >= 2:
obj = obj.model_dump()
else:
obj = obj.dict()
msg = struct_pb2.Value()
return parse_dict_to_proto(obj, msg, json_encoder=self._json_encoder)
def parse_dict_to_proto(
obj: JSONType,
msg: _message.Message,
json_encoder: type[json.JSONEncoder] = DefaultJsonEncoder,
) -> t.Any:
if obj is None:
# this function is an identity op for the msg if obj is None.
return msg
from google.protobuf.json_format import ParseDict
if isinstance(obj, (dict, str, list, float, int, bool)):
# ParseDict handles google.protobuf.Struct type
# directly if given object has a supported type
ParseDict(obj, msg)
else:
# If given object doesn't have a supported type, we will
# use given JSON encoder to convert it to dictionary
# and then parse it to google.protobuf.Struct.
# Note that if a custom JSON encoder is used, it mustn't
# take any arguments.
ParseDict(json_encoder().default(obj), msg)
return msg