semantic-chunking
Version:
Semantically create chunks from large texts. Useful for workflows involving large language models (LLMs).
18 lines (17 loc) • 507 B
JavaScript
const defaultFormValues = {
maxTokenSize: 500,
similarityThreshold: 0.5,
dynamicThresholdLowerBound: 0.4,
dynamicThresholdUpperBound: 0.8,
numSimilaritySentencesLookahead: 3,
combineChunks: true,
combineChunksSimilarityThreshold: 0.5,
onnxEmbeddingModel: "Xenova/all-MiniLM-L6-v2",
dtype: 'q8',
device: 'cpu',
returnEmbedding: false,
returnTokenLength: true,
chunkPrefix: null,
excludeChunkPrefixInResults: false,
};
export default defaultFormValues;