如何解析 XML 输出

前置条件

本指南假设您熟悉以下概念：

来自不同提供商的大型语言模型（LLMs）往往因其训练数据的具体差异而各具优势。这也意味着，在生成非 JSON 格式的输出时，某些模型可能表现更佳且更为可靠。

本指南展示了如何使用 XMLOutputParser 向模型提示以获取 XML 输出，然后解析该输出为可用格式。

注意

请记住，大型语言模型是存在信息泄露的抽象层！你必须使用具有足够容量以生成格式良好 XML 的 LLM。

在以下示例中，我们使用 Anthropic 的 Claude-2 模型（https://docs.anthropic.com/claude/docs），该模型是专为 XML 标签优化的模型之一。

%pip install -qU langchain langchain-anthropic

import os
from getpass import getpass

if "ANTHROPIC_API_KEY" not in os.environ:
    os.environ["ANTHROPIC_API_KEY"] = getpass()

让我们从向模型发出一个简单的请求开始。

from langchain_anthropic import ChatAnthropic
from langchain_core.output_parsers import XMLOutputParser
from langchain_core.prompts import PromptTemplate

model = ChatAnthropic(model="claude-2.1", max_tokens_to_sample=512, temperature=0.1)

actor_query = "Generate the shortened filmography for Tom Hanks."

output = model.invoke(
    f"""{actor_query}
Please enclose the movies in <movie></movie> tags"""
)

print(output.content)

API 参考:ChatAnthropic |XML 输出解析器 |提示模板

Here is the shortened filmography for Tom Hanks, with movies enclosed in XML tags:

<movie>Splash</movie>
<movie>Big</movie>
<movie>A League of Their Own</movie>
<movie>Sleepless in Seattle</movie>
<movie>Forrest Gump</movie>
<movie>Toy Story</movie>
<movie>Apollo 13</movie>
<movie>Saving Private Ryan</movie>
<movie>Cast Away</movie>
<movie>The Da Vinci Code</movie>

这实际上效果相当不错！但如果能将那个 XML 解析成更易用的格式就更好了。我们可以使用 XMLOutputParser 来同时向提示词添加默认格式指令，并将输出的 XML 解析为字典：

parser = XMLOutputParser()

# We will add these instructions to the prompt below
parser.get_format_instructions()

'The output should be formatted as a XML file.\n1. Output should conform to the tags below. \n2. If tags are not given, make them on your own.\n3. Remember to always open and close all the tags.\n\nAs an example, for the tags ["foo", "bar", "baz"]:\n1. String "<foo>\n   <bar>\n      <baz></baz>\n   </bar>\n</foo>" is a well-formatted instance of the schema. \n2. String "<foo>\n   <bar>\n   </foo>" is a badly-formatted instance.\n3. String "<foo>\n   <tag>\n   </tag>\n</foo>" is a badly-formatted instance.\n\nHere are the output tags:\n\`\`\`\nNone\n\`\`\`'

prompt = PromptTemplate(
    template="""{query}\n{format_instructions}""",
    input_variables=["query"],
    partial_variables={"format_instructions": parser.get_format_instructions()},
)

chain = prompt | model | parser

output = chain.invoke({"query": actor_query})
print(output)

{'filmography': [{'movie': [{'title': 'Big'}, {'year': '1988'}]}, {'movie': [{'title': 'Forrest Gump'}, {'year': '1994'}]}, {'movie': [{'title': 'Toy Story'}, {'year': '1995'}]}, {'movie': [{'title': 'Saving Private Ryan'}, {'year': '1998'}]}, {'movie': [{'title': 'Cast Away'}, {'year': '2000'}]}]}

我们还可以添加一些标签，以使输出符合我们的需求。你可以在提示的其他部分尝试添加自己的格式提示，以增强或替换默认指令：

parser = XMLOutputParser(tags=["movies", "actor", "film", "name", "genre"])

# We will add these instructions to the prompt below
parser.get_format_instructions()

'The output should be formatted as a XML file.\n1. Output should conform to the tags below. \n2. If tags are not given, make them on your own.\n3. Remember to always open and close all the tags.\n\nAs an example, for the tags ["foo", "bar", "baz"]:\n1. String "<foo>\n   <bar>\n      <baz></baz>\n   </bar>\n</foo>" is a well-formatted instance of the schema. \n2. String "<foo>\n   <bar>\n   </foo>" is a badly-formatted instance.\n3. String "<foo>\n   <tag>\n   </tag>\n</foo>" is a badly-formatted instance.\n\nHere are the output tags:\n\`\`\`\n[\'movies\', \'actor\', \'film\', \'name\', \'genre\']\n\`\`\`'

prompt = PromptTemplate(
    template="""{query}\n{format_instructions}""",
    input_variables=["query"],
    partial_variables={"format_instructions": parser.get_format_instructions()},
)


chain = prompt | model | parser

output = chain.invoke({"query": actor_query})

print(output)

{'movies': [{'actor': [{'name': 'Tom Hanks'}, {'film': [{'name': 'Forrest Gump'}, {'genre': 'Drama'}]}, {'film': [{'name': 'Cast Away'}, {'genre': 'Adventure'}]}, {'film': [{'name': 'Saving Private Ryan'}, {'genre': 'War'}]}]}]}

此输出解析器还支持部分块的分流传输。以下是一个示例：

for s in chain.stream({"query": actor_query}):
    print(s)

{'movies': [{'actor': [{'name': 'Tom Hanks'}]}]}
{'movies': [{'actor': [{'film': [{'name': 'Forrest Gump'}]}]}]}
{'movies': [{'actor': [{'film': [{'genre': 'Drama'}]}]}]}
{'movies': [{'actor': [{'film': [{'name': 'Cast Away'}]}]}]}
{'movies': [{'actor': [{'film': [{'genre': 'Adventure'}]}]}]}
{'movies': [{'actor': [{'film': [{'name': 'Saving Private Ryan'}]}]}]}
{'movies': [{'actor': [{'film': [{'genre': 'War'}]}]}]}

下一步

您现在已学会如何提示模型返回 XML。接下来，请查看关于获取结构化输出的更广泛指南，以了解其他相关技术。

下一步​

下一步