infiniflow/ragflow · error · Error

dataset_id (or kb_id/knowledge_id) is required for retrieval

Error message

dataset_id (or kb_id/knowledge_id) is required for retrievalTest

What it means

Thrown in web/src/services/knowledge-service.ts:151 by chunkService.retrievalTest when the params object contains none of dataset_id, kb_id, or knowledge_id. The service normalizes those legacy aliases into dataset_ids for the POST to api.retrievalTest; with no dataset identifier the request cannot be built, so it fails fast client-side before any network call.

Source

Thrown at web/src/services/knowledge-service.ts:151

    payload.available ??
    (payload.available_int === undefined
      ? undefined
      : payload.available_int === 1),
  image_base64: payload.image_base64,
});

const getAvailableParam = (available?: number) => {
  if (available === undefined) {
    return undefined;
  }
  return available === 1 ? 'true' : 'false';
};

const chunkService = {
  retrievalTest: async (params: Record<string, any>) => {
    const datasetId = params.dataset_id || params.kb_id || params.knowledge_id;
    if (!datasetId) {
      throw new Error(
        'dataset_id (or kb_id/knowledge_id) is required for retrievalTest',
      );
    }
    const datasetIds = Array.isArray(datasetId) ? datasetId : [datasetId];
    const rest = { ...params };
    delete rest.dataset_id;
    delete rest.kb_id;
    delete rest.knowledge_id;
    return request.post(api.retrievalTest, {
      data: { ...rest, dataset_ids: datasetIds },
    });
  },
  chunkList: async (params: Record<string, any>) => {
    const datasetId = getDatasetId(params);
    const documentId = getDocumentId(params);
    const response = await request.get(api.chunkList(datasetId, documentId), {
      params: {
        page: params.page,

View on GitHub (pinned to 554fb1133a)

Solutions

1. Pass dataset_id (the current canonical key) explicitly in the params object
2. Make the UI require a dataset selection before the test button is enabled
3. Type the params as { dataset_id: string; [k: string]: unknown } so TypeScript enforces the key
4. Extend the alias check if callers still legitimately use kb_id/knowledge_id

Example fix

// before
await chunkService.retrievalTest({
  question: query,
  page: 1,
  page_size: 10,
});

// after
await chunkService.retrievalTest({
  dataset_id: selectedKnowledgeBaseId,
  question: query,
  page: 1,
  page_size: 10,
});

const hasDatasetId = (p: Record<string, any>) =>
  Boolean(p.dataset_id || p.kb_id || p.knowledge_id);

const isRetrievalParams = (
  p: unknown,
): p is { dataset_id: string } & Record<string, unknown> =>
  typeof p === 'object' && p !== null &&
  typeof (p as any).dataset_id === 'string' && (p as any).dataset_id.length > 0;

When it happens

Trigger: Calling retrievalTest({question, ...}) and forgetting the dataset key; using a renamed key the normalizer does not know (e.g. datasetId camelCase); the form/dropdown value bound to dataset_id is undefined because no knowledge base was selected; data flowing from a context where kb_id was dropped during a refactor.

Common situations: Testing retrieval before selecting a dataset in the UI. Key drift across API versions (kb_id → dataset_id) leaving some callers on the old name that was removed from the alias list. Optional-chaining somewhere upstream producing undefined silently.

Related errors

• 101
• Cannot search across datasets where some have embedding mode
• Datasets use different embedding models: {[kb.embd_id for kb
• Invalid {label} format: {value!r}
• main() must be defined or exported.

AI-assisted analysis of infiniflow/ragflow@554fb1133a (2026-08-15). Data as JSON: /api/errors/27e18bde8b5e4c45. Report an issue: GitHub.