Class ChunkingConfig (2.32.0)

ChunkingConfig(mapping=None, *, ignore_unknown_fields=False, **kwargs)

Serving config for chunking.

Attributes

Name Description
chunk_size int
Optional. The chunk sizes to use when splitting documents, in order of level.
include_ancestor_headings bool
Optional. Whether or not to include ancestor headings when splitting.
semantic_chunking_group_size bool
Optional. The number of tokens to group together when evaluating semantic similarity.
breakpoint_percentile_threshold int
Optional. The percentile of cosine dissimilarity that must be exceeded between a group of tokens and the next. The smaller this number is, the more chunks will be generated.