Skip to main content
RichDocument, Table, and related helpers backed by Docling. RichDocument wraps a DoclingDocument (e.g. produced by converting a PDF or Markdown file) and renders it as Markdown for a language model. Table represents a single table within a Docling document and provides transpose, to_markdown, and query/transform helpers. Use RichDocument.from_document_file to convert a PDF or other supported format, and get_tables() to extract structured table data for downstream LLM-driven Q&A or transformation tasks.

​
Classes

​
CLASS RichDocument

A RichDocument is a block of content backed by a DoclingDocument. Provides helper functions for working with the document and extracting parts such as tables. Use from_document_file to convert PDFs or other formats, and save/load for persistence. Args:
  • doc: The underlying Docling document to wrap.
Methods:

​
FUNC parts 

parts(self) -> list[Component | CBlock]
Return the constituent parts of this document. Currently always returns an empty list. Future versions may support chunking the document into constituent parts. Returns:
  • list[Component | CBlock]: Always an empty list.

​
FUNC format_for_llm 

format_for_llm(self) -> TemplateRepresentation | str
Return the document content as a Markdown string. No template is needed; the full document is exported to Markdown directly. Returns:
  • TemplateRepresentation | str: The full document rendered as Markdown.

​
FUNC docling 

docling(self) -> DoclingDocument
Return the underlying DoclingDocument. Returns:
  • The wrapped Docling document instance.

​
FUNC to_markdown 

to_markdown(self)
Get the full text of the document as markdown.

​
FUNC get_tables 

get_tables(self) -> list[Table]
Return all tables found in this document. Returns:
  • list[Table]: A list of Table objects extracted from the document.

​
FUNC save 

save(self, filename: str | Path) -> None
Save the underlying DoclingDocument to a JSON file for later reuse. Args:
  • filename: Destination file path for the serialized document.

​
FUNC load 

load(cls, filename: str | Path) -> RichDocument
Load a RichDocument from a previously saved DoclingDocument JSON file. Args:
  • filename: Path to a JSON file previously created by RichDocument.save.
Returns:
  • A new RichDocument wrapping the loaded document.

​
FUNC from_document_file 

from_document_file(cls, source: str | Path | DocumentStream) -> RichDocument
Convert a document file to a RichDocument using Docling. Args:
  • source: Path or stream for the source document (e.g. a PDF or Markdown file).
Returns:
  • A new RichDocument wrapping the converted document.

​
CLASS TableQuery

A Query component specialised for Table objects. Formats the table as Markdown alongside the query string so the LLM receives both the structured table content and the natural-language question. Args:
  • obj: The table to query.
  • query: The natural-language question to ask about the table.
Methods:

​
FUNC parts 

parts(self) -> list[Component | CBlock]
Return the constituent parts of this table query. Returns:
  • list[Component | CBlock]: A list containing the wrapped Table
  • object.

​
FUNC format_for_llm 

format_for_llm(self) -> TemplateRepresentation
Format this table query for the language model. Renders the table as Markdown alongside the query string, and forwards any tools and fields from the table’s own representation. Returns:
  • Template args containing the query string
  • and the Markdown-rendered table.

​
CLASS TableTransform

A Transform component specialised for Table objects. Formats the table as Markdown alongside the transformation instruction so the LLM receives both the structured table content and the mutation description. Args:
  • obj: The table to transform.
  • transformation: Natural-language description of the desired mutation.
Methods:

​
FUNC parts 

parts(self) -> list[Component | CBlock]
Return the constituent parts of this table transform. Returns:
  • list[Component | CBlock]: A list containing the wrapped Table
  • object.

​
FUNC format_for_llm 

format_for_llm(self) -> TemplateRepresentation
Format this table transform for the language model. Renders the table as Markdown alongside the transformation description, and forwards any tools and fields from the table’s own representation. Returns:
  • Template args containing the transformation
  • description and the Markdown-rendered table.

​
CLASS Table

A Table represents a single table within a larger Docling Document. Args:
  • ti: The Docling TableItem extracted from the document.
  • doc: The parent DoclingDocument. Passing None may cause downstream Docling functions to fail.
Methods:

​
FUNC from_markdown 

from_markdown(cls, md: str) -> Table | None
Create a Table from a Markdown string by round-tripping through Docling. Wraps the Markdown in a minimal document, converts it with Docling, and returns the first table found. Args:
  • md: A Markdown string containing at least one table.
Returns:
  • Table | None: The first Table extracted from the Markdown, or
  • None if no table could be found.

​
FUNC parts 

parts(self)
Return the constituent parts of this table component. The current implementation always returns an empty list because the table is rendered entirely through format_for_llm. Returns:
  • list[Component | CBlock]: Always an empty list.

​
FUNC to_markdown 

to_markdown(self) -> str
Export this table as a Markdown string. Returns:
  • The Markdown representation of this table.

​
FUNC transpose 

transpose(self) -> Table | None
Transpose this table and return the result as a new Table. Returns:
  • Table | None: A new transposed Table, or None if the
  • transposed Markdown cannot be parsed back into a Table.

​
FUNC format_for_llm 

format_for_llm(self) -> TemplateRepresentation | str
Return the table representation for the Formatter. Returns:
  • TemplateRepresentation | str: A [TemplateRepresentation](../../../core/base#class-templaterepresentation) that
  • renders the table as its Markdown string using a \{\{table\}\}
  • template.