ONLYOFFICE Docs icon

ONLYOFFICE Docs

Consume ONLYOFFICE Document Server API (Conversion, Document Builder)

Actions21

Document Builder → Extract Chunks

AI-generated

Summary

Extracts document content into chunks for downstream processing such as vectorization. Choose whether to group content by heading sections or emit finer-grained chunks by paragraph or table row.

Inputs

  • File URL (required) — URL of the input document; it must be reachable from the ONLYOFFICE Document Server. Supports expressions, including a previous Document Builder step’s output URL.
  • Chunk By — Chunking strategy. By Headings (default) combines content between headings into section chunks; By Paragraphs emits one chunk per non-empty paragraph or table.

Output shape

A list of items, generally one per chunk, each with a JSON text field and metadata.

Heading chunks include metadata.section, metadata.headingLevel, metadata.elementStart, and metadata.elementEnd. Paragraph mode includes metadata.section, metadata.headingLevel, metadata.isHeading, and metadata.elementIndex. Chunk text can include table rows, whose cells are joined with |. Only non-empty paragraph and table content is emitted; if no chunks are found, the operation returns one item with an empty JSON object.

Links

Discussion