Source code for langchain_core.output_parsers.json

from __future__ import annotations

import json
from json import JSONDecodeError
from typing import Any, List, Optional, Type, TypeVar, Union

import jsonpatch  # type: ignore[import]
import pydantic  # pydantic: ignore

from langchain_core.exceptions import OutputParserException
from langchain_core.output_parsers.format_instructions import JSON_FORMAT_INSTRUCTIONS
from langchain_core.output_parsers.transform import BaseCumulativeTransformOutputParser
from langchain_core.outputs import Generation
from langchain_core.utils.json import (
    parse_and_check_json_markdown,
    parse_json_markdown,
    parse_partial_json,
)
from langchain_core.utils.pydantic import PYDANTIC_MAJOR_VERSION

if PYDANTIC_MAJOR_VERSION < 2:
    PydanticBaseModel = pydantic.BaseModel

else:
    from pydantic.v1 import BaseModel  # pydantic: ignore

    # Union type needs to be last assignment to PydanticBaseModel to make mypy happy.
    PydanticBaseModel = Union[BaseModel, pydantic.BaseModel]  # type: ignore

TBaseModel = TypeVar("TBaseModel", bound=PydanticBaseModel)


class JsonOutputParser(BaseCumulativeTransformOutputParser[Any]):
    """将LLM调用的输出解析为JSON对象。

在流模式下使用时，它将生成包含到目前为止返回的所有键的部分JSON对象。

在流模式下，如果`diff`设置为`True`，则生成描述前一个对象和当前对象之间差异的JSONPatch操作。
"""

    pydantic_object: Optional[Type[TBaseModel]] = None  # type: ignore

    def _diff(self, prev: Optional[Any], next: Any) -> Any:
        return jsonpatch.make_patch(prev, next).patch

    def _get_schema(self, pydantic_object: Type[TBaseModel]) -> dict[str, Any]:
        if PYDANTIC_MAJOR_VERSION == 2:
            if issubclass(pydantic_object, pydantic.BaseModel):
                return pydantic_object.model_json_schema()
            elif issubclass(pydantic_object, pydantic.v1.BaseModel):
                return pydantic_object.schema()
        return pydantic_object.schema()

    def parse_result(self, result: List[Generation], *, partial: bool = False) -> Any:
        text = result[0].text
        text = text.strip()
        if partial:
            try:
                return parse_json_markdown(text)
            except JSONDecodeError:
                return None
        else:
            try:
                return parse_json_markdown(text)
            except JSONDecodeError as e:
                msg = f"Invalid json output: {text}"
                raise OutputParserException(msg, llm_output=text) from e

    def parse(self, text: str) -> Any:
        return self.parse_result([Generation(text=text)])

    def get_format_instructions(self) -> str:
        if self.pydantic_object is None:
            return "Return a JSON object."
        else:
            # Copy schema to avoid altering original Pydantic schema.
            schema = {k: v for k, v in self._get_schema(self.pydantic_object).items()}

            # Remove extraneous fields.
            reduced_schema = schema
            if "title" in reduced_schema:
                del reduced_schema["title"]
            if "type" in reduced_schema:
                del reduced_schema["type"]
            # Ensure json in context is well-formed with double quotes.
            schema_str = json.dumps(reduced_schema)
            return JSON_FORMAT_INSTRUCTIONS.format(schema=schema_str)

    @property
    def _type(self) -> str:
        return "simple_json_output_parser"


# For backwards compatibility
SimpleJsonOutputParser = JsonOutputParser
parse_partial_json = parse_partial_json
parse_and_check_json_markdown = parse_and_check_json_markdown