Chunk

A semantically coherent slice of a document, sized to fit an embedding model's context window (sizes are set by chunking parameters). result.chunks is a list of these, ordered as they appear in the document.

result = docslicer.parse_document("report.pdf")
 
for chunk in result.chunks:
    print(chunk.text)
    print(chunk.page_number, chunk.heading)

Fields

Field	Type	Description
`id`	`str`	Unique identifier (`uuid`) for this chunk within the result.
`chunk_index`	`int`	Zero-based position of this chunk in the document.
`text`	`str`	The chunk text, including any serialised table content.
`char_count`	`int`	Character count of `text`.
`token_count`	`int`	Token count. Exact when `exact_tokens=True` was passed; otherwise a `chars / 4` estimate.
`page_number`	`int`	1-based physical page number where the chunk starts.
`page_label`	`str \| None`	Printed page label where the chunk starts, e.g. `"iv"`, `"A-6"`. Differs from `page_number` when documents use Roman numerals, lettered appendices, or custom page numbering. `None` when no label was detected.
`section`	`str`	Document region this chunk belongs to (`body`, `front_matter`, an appendix, …). See Sections for the full list.
`heading`	`str \| None`	Text of the active heading at the point this chunk was created — the nearest heading above the chunk in document order. `None` for content before the first heading.
`path`	`list[str]`	Full heading path from the document root to the active heading, e.g. `["Part II", "Item 1A", "Risk Factors"]`. Empty list when no heading context is available.
`parent_chunk_id`	`str \| None`	ID of the parent chunk when hierarchical chunking produced nested chunks. `None` for top-level chunks.
`bbox`	`BBox \| None`	Bounding box of the chunk on the page. Populated for PDF and for HTML rendered via Playwright; `None` for DOCX, PPTX, and HTML parsed without Playwright.
`link_url`	`list[str]`	Unique URLs found within this chunk's text. Empty list when none were detected.
`table_ids`	`list[str]`	IDs of tables whose content appears in this chunk. Use these to look up the full `Table` object from `result.tables`.
`chart_ids`	`list[str]`	IDs of charts whose content appears in this chunk. Use these to look up the full `Chart` object from `result.charts`. DOCX and PPTX only — always empty for PDF and HTML.
`extra`	`dict`	Additional pipeline fields requested via `extra_fields` on the parse call. Empty dict by default.

Examples

Filter by section

chunk.section names the document region the chunk came from. Keep body content only:

body_chunks = [c for c in result.chunks if c.section == "body"]

See Sections for the full list of values and how they're detected.

Access table content

tables_by_id = {t.id: t for t in result.tables}
 
for chunk in result.chunks:
    if chunk.table_ids:
        table = tables_by_id[chunk.table_ids[0]]
        print(table.markdown)

Use extra fields

result = docslicer.parse_document("report.pdf", extra_fields=["is_bold", "font_size"])
 
for chunk in result.chunks:
    print(chunk.extra["font_size"], chunk.text[:80])

Methods

`to_dict()`

Serialize the chunk to a plain dict. All fields are included.

d = chunk.to_dict()

`Chunk.from_dict(d)`

Reconstruct a Chunk from a dict produced by to_dict().

chunk = Chunk.from_dict(d)