llmxml - XML parsing for LLM outputs

Sponsored by Speck

XML outputs are a common format for LLM outputs, but they are not always easy to parse. This package provides a way to parse XML outputs into Pydantic models.

This supports nested models, lists, and unions as well as partial parsing for streamed responses. llmxml also supports generating a prompt template for the XML schema to pass to the LLM.

Usage

Parsing XML

from llmxml import parse_xml, generate_prompt_template
from pydantic import BaseModel, Field

class Movie(BaseModel):
    title: str = Field(..., description="The title of the movie")
    director: str = Field(..., description="The director of the movie")

class Response(BaseModel):
    movies: list[Movie] = Field(
        ..., description="A list of movies that match the query"
    )

class ResponseObject(BaseModel):
    response: Response = Field(
        ..., description="The response object that contains the movies"
    )

xml: str = """
<response>
    <movies>
        <movie>
            <title>The Matrix</title>
            <director>The Wachowskis</director>
        </movie>
    </movies>
</response>
"""

result: ResponseObject = parse_xml(ResponseObject, xml)
print(result)

Output:

response=Response(movies=[Movie(title='The Matrix', director='The Wachowskis')])

Generating a prompt template

prompt: str = generate_prompt_template(
    model=Response, include_instructions=True  # Default is true
)
print(prompt)

Output:

<response_instructions>
You are to understand the content and provide the parsed objects in xml that match the following xml_schema:

Make sure to return an instance of the XML, not the schema itself

Each field in the field_schema has a description and a type.
Example:
<field_name>
[type]
[description]
</field_name>

Schema:
<movies>
[type: list]
[A list of movies]
</movies>
</response_instructions>

Development

Running tests

$ uv run -m pytest -n auto tests/test_parser.py

License

This project is licensed under the MIT License. See the LICENSE file for details.