perf: chunk trigger and paragraph split (#4893)

* perf: chunk trigger and paragraph split

* update max size computed

* perf: i18n

* remove table
This commit is contained in:
Archer
2025-05-26 18:57:22 +08:00
committed by GitHub
parent 874300a56a
commit c25cd48e72
23 changed files with 859 additions and 164 deletions

View File

@@ -120,7 +120,6 @@ export const computeChunkSize = (params: {
return Math.min(params.chunkSize ?? chunkAutoChunkSize, getLLMMaxChunkSize(params.llmModel));
};
export const computeChunkSplitter = (params: {
chunkSettingMode?: ChunkSettingModeEnum;
chunkSplitMode?: DataChunkSplitModeEnum;
@@ -129,8 +128,21 @@ export const computeChunkSplitter = (params: {
if (params.chunkSettingMode === ChunkSettingModeEnum.auto) {
return undefined;
}
if (params.chunkSplitMode === DataChunkSplitModeEnum.size) {
if (params.chunkSplitMode !== DataChunkSplitModeEnum.char) {
return undefined;
}
return params.chunkSplitter;
};
export const computeParagraphChunkDeep = (params: {
chunkSettingMode?: ChunkSettingModeEnum;
chunkSplitMode?: DataChunkSplitModeEnum;
paragraphChunkDeep?: number;
}) => {
if (params.chunkSettingMode === ChunkSettingModeEnum.auto) {
return 5;
}
if (params.chunkSplitMode === DataChunkSplitModeEnum.paragraph) {
return params.paragraphChunkDeep;
}
return 0;
};