如何从模型返回结构化数据

先决条件

本指南假定您熟悉以下概念：

让模型返回与特定架构匹配的输出通常很有用。一个常见的用例是从文本中提取数据以插入数据库或与其他下游系统一起使用。本指南介绍了从模型获取结构化输出的几种策略。

这`.with_structured_output()`方法

支持的型号

您可以在此处找到支持此方法的模型列表。

这是获得结构化输出的最简单、最可靠的方法。with_structured_output()针对提供用于构建输出的本机 API（如工具/函数调用或 JSON 模式）的模型实现，并在后台利用这些功能。

此方法将架构作为输入，该架构指定所需输出属性的名称、类型和描述。该方法返回一个类似模型的 Runnable，不同之处在于它不是输出字符串或消息，而是输出与给定架构对应的对象。该模式可以指定为 TypedDict 类、JSON Schema 或 Pydantic 类。如果使用 TypedDict 或 JSON Schema，则 Runnable 将返回一个字典，如果使用 Pydantic 类，则将返回一个 Pydantic 对象。

例如，让我们获取一个模型来生成一个笑话，并将设置与妙语分开：

选择聊天模式：

pip install -qU "langchain[openai]"

import getpass
import os

if not os.environ.get("OPENAI_API_KEY"):
  os.environ["OPENAI_API_KEY"] = getpass.getpass("Enter API key for OpenAI: ")

from langchain.chat_models import init_chat_model

llm = init_chat_model("gpt-4o-mini", model_provider="openai")

Pydantic 类

如果我们希望模型返回一个 Pydantic 对象，我们只需要传入所需的 Pydantic 类。使用 Pydantic 的主要优点是模型生成的输出将得到验证。如果缺少任何必填字段或任何字段的类型错误，Pydantic 将引发错误。

from typing import Optional

from pydantic import BaseModel, Field


# Pydantic
class Joke(BaseModel):
    """Joke to tell user."""

    setup: str = Field(description="The setup of the joke")
    punchline: str = Field(description="The punchline to the joke")
    rating: Optional[int] = Field(
        default=None, description="How funny the joke is, from 1 to 10"
    )


structured_llm = llm.with_structured_output(Joke)

structured_llm.invoke("Tell me a joke about cats")

Joke(setup='Why was the cat sitting on the computer?', punchline='Because it wanted to keep an eye on the mouse!', rating=7)

提示

除了 Pydantic 类的结构之外，Pydantic 类的名称、文档字符串以及参数的名称和提供的描述也非常重要。大多数时候with_structured_output正在使用模型的函数/工具调用 API，您可以有效地将所有这些信息视为已添加到模型提示符中。

TypedDict 或 JSON 架构

如果你不想使用 Pydantic，明确不想验证参数，或者希望能够流式传输模型输出，你可以使用 TypedDict 类定义你的模式。我们可以选择性地使用特殊的Annotated语法，允许您指定字段的默认值和描述。请注意，如果模型未生成默认值，则不会自动填充默认值，它仅用于定义传递给模型的架构。

要求

核心：langchain-core>=0.2.26
键入扩展：强烈建议导入Annotated和TypedDict从typing_extensions而不是typing以确保跨 Python 版本的行为一致。

from typing import Optional

from typing_extensions import Annotated, TypedDict

# TypedDict
class Joke(TypedDict):
    """Joke to tell user."""

    setup: Annotated[str, ..., "The setup of the joke"]

    # Alternatively, we could have specified setup as:

    # setup: str                    # no default, no description
    # setup: Annotated[str, ...]    # no default, no description
    # setup: Annotated[str, "foo"]  # default, no description

    punchline: Annotated[str, ..., "The punchline of the joke"]
    rating: Annotated[Optional[int], None, "How funny the joke is, from 1 to 10"]

structured_llm = llm.with_structured_output(Joke)

structured_llm.invoke("Tell me a joke about cats")

{'setup': 'Why was the cat sitting on the computer?',
 'punchline': 'Because it wanted to keep an eye on the mouse!',
 'rating': 7}

等效地，我们可以传入一个 JSON Schema dict。这不需要导入或类，并且非常清楚地说明了每个参数是如何记录的，但代价是更加冗长。

json_schema = {
    "title": "joke",
    "description": "Joke to tell user.",
    "type": "object",
    "properties": {
        "setup": {
            "type": "string",
            "description": "The setup of the joke",
        },
        "punchline": {
            "type": "string",
            "description": "The punchline to the joke",
        },
        "rating": {
            "type": "integer",
            "description": "How funny the joke is, from 1 to 10",
            "default": None,
        },
    },
    "required": ["setup", "punchline"],
}
structured_llm = llm.with_structured_output(json_schema)

structured_llm.invoke("Tell me a joke about cats")

{'setup': 'Why was the cat sitting on the computer?',
 'punchline': 'Because it wanted to keep an eye on the mouse!',
 'rating': 7}

在多个架构之间进行选择

让模型从多个架构中进行选择的最简单方法是创建具有 Union 类型属性的父架构。

使用 Pydantic

from typing import Union

class Joke(BaseModel):
    """Joke to tell user."""

    setup: str = Field(description="The setup of the joke")
    punchline: str = Field(description="The punchline to the joke")
    rating: Optional[int] = Field(
        default=None, description="How funny the joke is, from 1 to 10"
    )

class ConversationalResponse(BaseModel):
    """Respond in a conversational manner. Be kind and helpful."""

    response: str = Field(description="A conversational response to the user's query")

class FinalResponse(BaseModel):
    final_output: Union[Joke, ConversationalResponse]

structured_llm = llm.with_structured_output(FinalResponse)

structured_llm.invoke("Tell me a joke about cats")

FinalResponse(final_output=Joke(setup='Why was the cat sitting on the computer?', punchline='Because it wanted to keep an eye on the mouse!', rating=7))

structured_llm.invoke("How are you today?")

FinalResponse(final_output=ConversationalResponse(response="I'm just a computer program, so I don't have feelings, but I'm here and ready to help you with whatever you need!"))

使用 TypedDict

from typing import Optional, Union

from typing_extensions import Annotated, TypedDict

class Joke(TypedDict):
    """Joke to tell user."""

    setup: Annotated[str, ..., "The setup of the joke"]
    punchline: Annotated[str, ..., "The punchline of the joke"]
    rating: Annotated[Optional[int], None, "How funny the joke is, from 1 to 10"]

class ConversationalResponse(TypedDict):
    """Respond in a conversational manner. Be kind and helpful."""

    response: Annotated[str, ..., "A conversational response to the user's query"]

class FinalResponse(TypedDict):
    final_output: Union[Joke, ConversationalResponse]

structured_llm = llm.with_structured_output(FinalResponse)

structured_llm.invoke("Tell me a joke about cats")

{'final_output': {'setup': 'Why was the cat sitting on the computer?',
  'punchline': 'Because it wanted to keep an eye on the mouse!',
  'rating': 7}}

structured_llm.invoke("How are you today?")

{'final_output': {'response': "I'm just a computer program, so I don't have feelings, but I'm here and ready to help you with whatever you need!"}}

响应应与 Pydantic 示例中所示的响应相同。

或者，如果所选模型支持，您可以直接使用工具调用来允许模型在选项之间进行选择。这涉及更多的解析和设置，但在某些情况下会带来更好的性能，因为您不必使用嵌套架构。有关更多详细信息，请参阅此操作指南。

流

当输出类型为 dict 时（即，当 schema 指定为 TypedDict 类或 JSON Schema dict 时），我们可以流式传输来自结构化模型的输出。

信息

请注意，生成的已经是聚合的 chunks，而不是 deltas。

from typing_extensions import Annotated, TypedDict


# TypedDict
class Joke(TypedDict):
    """Joke to tell user."""

    setup: Annotated[str, ..., "The setup of the joke"]
    punchline: Annotated[str, ..., "The punchline of the joke"]
    rating: Annotated[Optional[int], None, "How funny the joke is, from 1 to 10"]


structured_llm = llm.with_structured_output(Joke)

for chunk in structured_llm.stream("Tell me a joke about cats"):
    print(chunk)

{}
{'setup': ''}
{'setup': 'Why'}
{'setup': 'Why was'}
{'setup': 'Why was the'}
{'setup': 'Why was the cat'}
{'setup': 'Why was the cat sitting'}
{'setup': 'Why was the cat sitting on'}
{'setup': 'Why was the cat sitting on the'}
{'setup': 'Why was the cat sitting on the computer'}
{'setup': 'Why was the cat sitting on the computer?'}
{'setup': 'Why was the cat sitting on the computer?', 'punchline': ''}
{'setup': 'Why was the cat sitting on the computer?', 'punchline': 'Because'}
{'setup': 'Why was the cat sitting on the computer?', 'punchline': 'Because it'}
{'setup': 'Why was the cat sitting on the computer?', 'punchline': 'Because it wanted'}
{'setup': 'Why was the cat sitting on the computer?', 'punchline': 'Because it wanted to'}
{'setup': 'Why was the cat sitting on the computer?', 'punchline': 'Because it wanted to keep'}
{'setup': 'Why was the cat sitting on the computer?', 'punchline': 'Because it wanted to keep an'}
{'setup': 'Why was the cat sitting on the computer?', 'punchline': 'Because it wanted to keep an eye'}
{'setup': 'Why was the cat sitting on the computer?', 'punchline': 'Because it wanted to keep an eye on'}
{'setup': 'Why was the cat sitting on the computer?', 'punchline': 'Because it wanted to keep an eye on the'}
{'setup': 'Why was the cat sitting on the computer?', 'punchline': 'Because it wanted to keep an eye on the mouse'}
{'setup': 'Why was the cat sitting on the computer?', 'punchline': 'Because it wanted to keep an eye on the mouse!'}
{'setup': 'Why was the cat sitting on the computer?', 'punchline': 'Because it wanted to keep an eye on the mouse!', 'rating': 7}

Few-shot 提示

对于更复杂的 schema，在 prompt 中添加 few-shot 示例非常有用。这可以通过几种方式完成。

最简单、最通用的方法是在提示符中向系统消息添加示例：

from langchain_core.prompts import ChatPromptTemplate

system = """You are a hilarious comedian. Your specialty is knock-knock jokes. \
Return a joke which has the setup (the response to "Who's there?") and the final punchline (the response to "<setup> who?").

Here are some examples of jokes:

example_user: Tell me a joke about planes
example_assistant: {{"setup": "Why don't planes ever get tired?", "punchline": "Because they have rest wings!", "rating": 2}}

example_user: Tell me another joke about planes
example_assistant: {{"setup": "Cargo", "punchline": "Cargo 'vroom vroom', but planes go 'zoom zoom'!", "rating": 10}}

example_user: Now about caterpillars
example_assistant: {{"setup": "Caterpillar", "punchline": "Caterpillar really slow, but watch me turn into a butterfly and steal the show!", "rating": 5}}"""

prompt = ChatPromptTemplate.from_messages([("system", system), ("human", "{input}")])

few_shot_structured_llm = prompt | structured_llm
few_shot_structured_llm.invoke("what's something funny about woodpeckers")

API 参考：ChatPromptTemplate

{'setup': 'Woodpecker',
 'punchline': "Woodpecker you a joke, but I'm afraid it might be too 'hole-some'!",
 'rating': 7}

当构建输出的底层方法是工具调用时，我们可以将示例作为显式工具调用传入。您可以在其 API 参考中检查您正在使用的模型是否使用了工具调用。

from langchain_core.messages import AIMessage, HumanMessage, ToolMessage

examples = [
    HumanMessage("Tell me a joke about planes", name="example_user"),
    AIMessage(
        "",
        name="example_assistant",
        tool_calls=[
            {
                "name": "joke",
                "args": {
                    "setup": "Why don't planes ever get tired?",
                    "punchline": "Because they have rest wings!",
                    "rating": 2,
                },
                "id": "1",
            }
        ],
    ),
    # Most tool-calling models expect a ToolMessage(s) to follow an AIMessage with tool calls.
    ToolMessage("", tool_call_id="1"),
    # Some models also expect an AIMessage to follow any ToolMessages,
    # so you may need to add an AIMessage here.
    HumanMessage("Tell me another joke about planes", name="example_user"),
    AIMessage(
        "",
        name="example_assistant",
        tool_calls=[
            {
                "name": "joke",
                "args": {
                    "setup": "Cargo",
                    "punchline": "Cargo 'vroom vroom', but planes go 'zoom zoom'!",
                    "rating": 10,
                },
                "id": "2",
            }
        ],
    ),
    ToolMessage("", tool_call_id="2"),
    HumanMessage("Now about caterpillars", name="example_user"),
    AIMessage(
        "",
        tool_calls=[
            {
                "name": "joke",
                "args": {
                    "setup": "Caterpillar",
                    "punchline": "Caterpillar really slow, but watch me turn into a butterfly and steal the show!",
                    "rating": 5,
                },
                "id": "3",
            }
        ],
    ),
    ToolMessage("", tool_call_id="3"),
]
system = """You are a hilarious comedian. Your specialty is knock-knock jokes. \
Return a joke which has the setup (the response to "Who's there?") \
and the final punchline (the response to "<setup> who?")."""

prompt = ChatPromptTemplate.from_messages(
    [("system", system), ("placeholder", "{examples}"), ("human", "{input}")]
)
few_shot_structured_llm = prompt | structured_llm
few_shot_structured_llm.invoke({"input": "crocodiles", "examples": examples})

API 参考：AIMessage | HumanMessage （人工消息） | 工具消息

{'setup': 'Crocodile',
 'punchline': 'Crocodile be seeing you later, alligator!',
 'rating': 6}

有关使用工具调用时出现少量镜头提示的更多信息，请参阅此处。

（高级）指定构建输出的方法

对于支持多种构建输出方式的模型（即，它们同时支持工具调用和 JSON 模式），您可以指定要使用method=论点。

JSON 模式

如果使用 JSON 模式，您仍必须在模型提示符中指定所需的架构。您传递给with_structured_output将仅用于解析模型输出，不会像使用工具调用那样传递给模型。

要查看您使用的模型是否支持 JSON 模式，请检查 API 参考中的其条目。

structured_llm = llm.with_structured_output(None, method="json_mode")

structured_llm.invoke(
    "Tell me a joke about cats, respond in JSON with `setup` and `punchline` keys"
)

{'setup': 'Why was the cat sitting on the computer?',
 'punchline': 'Because it wanted to keep an eye on the mouse!'}

（高级）原始输出

LLM 在生成结构化输出方面并不完美，尤其是在架构变得复杂时。您可以避免引发异常，并通过将include_raw=True.这会更改输出格式以包含原始消息输出parsed值（如果成功）和任何由此产生的错误：

structured_llm = llm.with_structured_output(Joke, include_raw=True)

structured_llm.invoke("Tell me a joke about cats")

{'raw': AIMessage(content='', additional_kwargs={'tool_calls': [{'id': 'call_f25ZRmh8u5vHlOWfTUw8sJFZ', 'function': {'arguments': '{"setup":"Why was the cat sitting on the computer?","punchline":"Because it wanted to keep an eye on the mouse!","rating":7}', 'name': 'Joke'}, 'type': 'function'}]}, response_metadata={'token_usage': {'completion_tokens': 33, 'prompt_tokens': 93, 'total_tokens': 126}, 'model_name': 'gpt-4o-2024-05-13', 'system_fingerprint': 'fp_4e2b2da518', 'finish_reason': 'stop', 'logprobs': None}, id='run-d880d7e2-df08-4e9e-ad92-dfc29f2fd52f-0', tool_calls=[{'name': 'Joke', 'args': {'setup': 'Why was the cat sitting on the computer?', 'punchline': 'Because it wanted to keep an eye on the mouse!', 'rating': 7}, 'id': 'call_f25ZRmh8u5vHlOWfTUw8sJFZ', 'type': 'tool_call'}], usage_metadata={'input_tokens': 93, 'output_tokens': 33, 'total_tokens': 126}),
 'parsed': {'setup': 'Why was the cat sitting on the computer?',
  'punchline': 'Because it wanted to keep an eye on the mouse!',
  'rating': 7},
 'parsing_error': None}

直接提示和解析模型输出

并非所有型号都支持.with_structured_output()，因为并非所有模型都支持工具调用或 JSON 模式。对于此类模型，您需要直接提示模型使用特定格式，并使用输出解析器从原始模型输出中提取结构化响应。

用`PydanticOutputParser`

以下示例使用内置的PydanticOutputParser来解析提示匹配给定 Pydantic 架构的聊天模型的输出。请注意，我们正在添加format_instructions直接从解析器上的方法发送到提示符：

from typing import List

from langchain_core.output_parsers import PydanticOutputParser
from langchain_core.prompts import ChatPromptTemplate
from pydantic import BaseModel, Field


class Person(BaseModel):
    """Information about a person."""

    name: str = Field(..., description="The name of the person")
    height_in_meters: float = Field(
        ..., description="The height of the person expressed in meters."
    )


class People(BaseModel):
    """Identifying information about all people in a text."""

    people: List[Person]


# Set up a parser
parser = PydanticOutputParser(pydantic_object=People)

# Prompt
prompt = ChatPromptTemplate.from_messages(
    [
        (
            "system",
            "Answer the user query. Wrap the output in `json` tags\n{format_instructions}",
        ),
        ("human", "{query}"),
    ]
).partial(format_instructions=parser.get_format_instructions())

API 参考：PydanticOutputParser | 聊天提示模板

我们来看看向模型发送了哪些信息：

query = "Anna is 23 years old and she is 6 feet tall"

print(prompt.invoke({"query": query}).to_string())

System: Answer the user query. Wrap the output in `json` tags
The output should be formatted as a JSON instance that conforms to the JSON schema below.

As an example, for the schema {"properties": {"foo": {"title": "Foo", "description": "a list of strings", "type": "array", "items": {"type": "string"}}}, "required": ["foo"]}
the object {"foo": ["bar", "baz"]} is a well-formatted instance of the schema. The object {"properties": {"foo": ["bar", "baz"]}} is not well-formatted.

Here is the output schema:
\`\`\`
{"description": "Identifying information about all people in a text.", "properties": {"people": {"title": "People", "type": "array", "items": {"$ref": "#/definitions/Person"}}}, "required": ["people"], "definitions": {"Person": {"title": "Person", "description": "Information about a person.", "type": "object", "properties": {"name": {"title": "Name", "description": "The name of the person", "type": "string"}, "height_in_meters": {"title": "Height In Meters", "description": "The height of the person expressed in meters.", "type": "number"}}, "required": ["name", "height_in_meters"]}}}
\`\`\`
Human: Anna is 23 years old and she is 6 feet tall

现在让我们调用它：

chain = prompt | llm | parser

chain.invoke({"query": query})

People(people=[Person(name='Anna', height_in_meters=1.8288)])

要更深入地了解如何将输出解析器与结构化输出的提示技术结合使用，请参阅本指南。

自定义解析

您还可以使用 LangChain 表达式语言（LCEL）创建自定义提示和解析器，使用 plain 函数解析模型的输出：

import json
import re
from typing import List

from langchain_core.messages import AIMessage
from langchain_core.prompts import ChatPromptTemplate
from pydantic import BaseModel, Field


class Person(BaseModel):
    """Information about a person."""

    name: str = Field(..., description="The name of the person")
    height_in_meters: float = Field(
        ..., description="The height of the person expressed in meters."
    )


class People(BaseModel):
    """Identifying information about all people in a text."""

    people: List[Person]


# Prompt
prompt = ChatPromptTemplate.from_messages(
    [
        (
            "system",
            "Answer the user query. Output your answer as JSON that  "
            "matches the given schema: \`\`\`json\n{schema}\n\`\`\`. "
            "Make sure to wrap the answer in \`\`\`json and \`\`\` tags",
        ),
        ("human", "{query}"),
    ]
).partial(schema=People.schema())


# Custom parser
def extract_json(message: AIMessage) -> List[dict]:
    """Extracts JSON content from a string where JSON is embedded between \`\`\`json and \`\`\` tags.

    Parameters:
        text (str): The text containing the JSON content.

    Returns:
        list: A list of extracted JSON strings.
    """
    text = message.content
    # Define the regular expression pattern to match JSON blocks
    pattern = r"\`\`\`json(.*?)\`\`\`"

    # Find all non-overlapping matches of the pattern in the string
    matches = re.findall(pattern, text, re.DOTALL)

    # Return the list of matched JSON strings, stripping any leading or trailing whitespace
    try:
        return [json.loads(match.strip()) for match in matches]
    except Exception:
        raise ValueError(f"Failed to parse: {message}")

API 参考：AIMessage | 聊天提示模板

以下是发送给模型的提示：

query = "Anna is 23 years old and she is 6 feet tall"

print(prompt.format_prompt(query=query).to_string())

System: Answer the user query. Output your answer as JSON that  matches the given schema: \`\`\`json
{'title': 'People', 'description': 'Identifying information about all people in a text.', 'type': 'object', 'properties': {'people': {'title': 'People', 'type': 'array', 'items': {'$ref': '#/definitions/Person'}}}, 'required': ['people'], 'definitions': {'Person': {'title': 'Person', 'description': 'Information about a person.', 'type': 'object', 'properties': {'name': {'title': 'Name', 'description': 'The name of the person', 'type': 'string'}, 'height_in_meters': {'title': 'Height In Meters', 'description': 'The height of the person expressed in meters.', 'type': 'number'}}, 'required': ['name', 'height_in_meters']}}}
\`\`\`. Make sure to wrap the answer in \`\`\`json and \`\`\` tags
Human: Anna is 23 years old and she is 6 feet tall

当我们调用它时，它看起来是这样的：

chain = prompt | llm | extract_json

chain.invoke({"query": query})

[{'people': [{'name': 'Anna', 'height_in_meters': 1.8288}]}]

这.with_structured_output()方法