类class agentscope.rag

WordParser

class WordParser(ParserBase)

源码Source agentscope/rag/_parser/_word.py:288–502

Word 解析器，把私有文档转化为可检索的知识。位于 agentscope.rag 包，继承自 ParserBase。核心成员包括 supported_extensions、parse。

Document order is preserved. Elements are visited sequentially and grouped into a minimum number of sections:

Paragraphs contribute to one running text section.
Tables are rendered as Markdown or JSON and, depending on separate_table, either merged into the running text or emitted as standalone text sections.
Images emit standalone :class:Section instances holding a :class:DataBlock with the base64-encoded image bytes.

构造器摘要Constructor Summary

构造器Constructor	说明Description
`def __init__(include_image: bool = True, separate_table: bool = False, table_format: Literal['markdown', 'json'] = 'mar…`	初始化 `WordParser` 实例。 Initialize the Word parser.

属性Attribute	类型Type	说明Description	值Value
`supported_media_types`	list[str]	`supported_media_types` 字段。	`['application/vnd.openxmlformats-officedocument.wordprocessingml.document']`

方法Method	说明Description
`def supported_extensions() -> list[str]`	返回 `supported_extensions` 结果。 Return ``[".docx"]`` — the only format ``python-docx`` reads.
`async def parse(file: bytes \| str, filename: str) -> list[Section]`async	执行解析相关操作。 Parse a DOCX file into a list of :class:`Section` objects.

def __init__(include_image: bool = True, separate_table: bool = False, table_format: Literal['markdown', 'json'] = 'markdown') -> None

初始化 WordParser 实例。

Initialize the Word parser.

参数Parameters

名称Name	类型Type	说明Description
`include_image`默认default `True`	bool	include 图像（类型 `bool`）。When `True`, embedded images are emitted as :class:`DataBlock` sections. Set to `False` to keep a text-only index.
`separate_table`默认default `False`	bool	separate table（类型 `bool`）。When `True`, each table becomes its own text section, never merged with surrounding paragraphs. table_format (`Literal["markdown", "json"]`, defaults to `"markdown"`): How to render tables. `"markdown"` uses pipe-table syntax, escaping pipes and rendering cell line breaks as ` `; `"json"` emits a JSON array prefixed with a ` ` marker and preserves extracted cell strings without Markdown rendering.
`table_format`默认default `'markdown'`	Literal['markdown', 'json']	table 格式（类型 `Literal['markdown', 'json']`）。The table format.

异常Raises

源码Source agentscope/rag/_parser/_word.py:312–347

list[str] = ['application/vnd.openxmlformats-officedocument.wordprocessingml.document']

supported_media_types 字段。声明为 list[str]。

The supported media types field.

源码Source agentscope/rag/_parser/_word.py:302

def supported_extensions() -> list[str]

返回 supported_extensions 结果。

Return [".docx"] — the only format python-docx reads.

返回Returns list[str]

返回str列表。Returns a list of str.

源码Source agentscope/rag/_parser/_word.py:308–310

async def parse(file: bytes | str, filename: str) -> list[Section]

async

执行解析相关操作。

Parse a DOCX file into a list of :class:Section objects.

参数Parameters

名称Name	类型Type	说明Description
`file`	bytes \| str	文件（类型 `bytes \| str`）。Either the raw DOCX bytes, or a filesystem path to the DOCX file.
`filename`	str	filename（类型 `str`）。The source filename, copied into each Section's :attr:`Section.source`.

返回Returns list[Section]

返回Section列表。`list[Section]`:

异常Raises

源码Source agentscope/rag/_parser/_word.py:349–502