Member Junction
    Preparing search index...

    Interface MJContentSourceEntity_IContentSegmentationOptions

    Options for the segmentation strategy. All optional; each segmenter ignores options that don't apply to it.

    interface MJContentSourceEntity_IContentSegmentationOptions {
        BoundaryGapMs?: number;
        EmitSubChapters?: boolean;
        MaxChapterMs?: number;
        MaxSegmentTokens?: number;
        MinSegmentTokens?: number;
        MinTokensForLLM?: number;
        NoSplitPercent?: number;
        OverlapTokens?: number;
        OvershootPercent?: number;
        TargetTokens?: number;
        UndershootPercent?: number;
        WindowMs?: number;
    }
    Index

    Properties

    BoundaryGapMs?: number

    Transcript: a silence gap at least this long starts a new chapter.

    EmitSubChapters?: boolean

    Transcript: also emit one child segment per speaker turn within each chapter.

    MaxChapterMs?: number

    Transcript: maximum wall-clock length of one chapter, in milliseconds.

    MaxSegmentTokens?: number

    Hard ceiling on tokens per segment. Segments larger than this are split.

    MinSegmentTokens?: number

    Merge adjacent text segments estimating below this many tokens.

    MinTokensForLLM?: number

    SemanticText: skip the LLM boundary pass for documents below this token count.

    NoSplitPercent?: number

    AdaptiveBoundary: if the whole text is within this percent of target, don't split at all.

    OverlapTokens?: number

    Overlap tokens applied when an oversized segment must be split.

    OvershootPercent?: number

    AdaptiveBoundary: percent above target to keep looking for a sentence/word break.

    TargetTokens?: number

    AdaptiveBoundary: desired segment size — size this to your queries, not to the model.

    UndershootPercent?: number

    AdaptiveBoundary: percent below target at which a paragraph break is accepted.

    WindowMs?: number

    FixedWindow: window length in milliseconds for audio/video with no transcript.