Nesting depth; 0 for top-level segments.
OptionalEndSegment end in milliseconds, for audio/video.
OptionalEndExclusive end character offset within the source text.
OptionalMediaMedia payload pointer, for non-text segments.
Payload modality.
OptionalPage1-based page number, for paginated sources (PDF, slides).
OptionalParentSequence of this segment's parent, for chapter -> sub-chapter hierarchies.
Registration key of the segmenter that produced this segment — provenance.
0-based position within the full segment list. Assigned by BaseSegmenter.
OptionalSpeakerSpeaker label carried through from transcript cues, when known.
OptionalStartSegment start in milliseconds, for audio/video.
OptionalStartInclusive start character offset within the source text.
OptionalTextTextual payload — the extracted/transcribed text for this segment.
OptionalTitleHuman-readable label, e.g. a heading or a generated chapter title.
Estimated token count of Text (0 for pure-media segments).
One embeddable unit of content produced by a segmenter.
A segment carries
Text,Media, or both (the "dual representation" case: a video chapter with a native media reference and its transcript, so it can be embedded natively for retrieval while remaining readable for an agent).