Skip to content

Enterprise agent platform for building, deploying, and operating AI teammates.

6.9 KB
Raw Blame
import { describe, expect, it } from "vitest"

import {
  buildKnowledgeBaseErrorResult,
  getKnowledgeBaseErrorToastContent,
  normalizeKnowledgeBaseIngestionResult,
} from "./kb-ingest-feedback"

const COPY = {
  genericTitle: "上传失败",
  nameUnavailableTitle: "该知识库名称不可用",
  nameUnavailableDescription: "请换一个知识库名称后重试。",
  embeddingTitle: "知识库导入失败:未配置可用的嵌入模型",
  embeddingDescription: "请先配置默认嵌入模型,或选择一个可用的嵌入模型后重试。",
  rollbackTitle: "知识库导入失败,清理未完全完成",
  rollbackDescription: "系统已尝试回滚本次导入,请查看处理结果中的详细错误。",
}

describe("getKnowledgeBaseErrorToastContent", () => {
  it("maps embedding configuration errors to a concise actionable toast", () => {
    const result = getKnowledgeBaseErrorToastContent(
      "Model 'text-embedding-v4' not found in hub and no environment configuration available for embedding.",
      COPY,
      { status: undefined, adviseRename: true }
    )

    expect(result).toEqual({
      title: COPY.embeddingTitle,
      description: COPY.embeddingDescription,
    })
  })

  it("maps an advise-rename 409 to actionable copy without leaking the raw detail", () => {
    const result = getKnowledgeBaseErrorToastContent(
      "Knowledge base name unavailable: test.",
      COPY,
      { status: 409, adviseRename: true }
    )

    expect(result).toEqual({
      title: COPY.nameUnavailableTitle,
      description: COPY.nameUnavailableDescription,
    })
  })

  it("classifies a 409 by status alone, whatever wording the backend sends", () => {
    const result = getKnowledgeBaseErrorToastContent(
      "some entirely different sentence the backend may switch to",
      COPY,
      { status: 409, adviseRename: true }
    )

    expect(result).toEqual({
      title: COPY.nameUnavailableTitle,
      description: COPY.nameUnavailableDescription,
    })
  })

  it("does not treat the conflict wording as a conflict without a 409", () => {
    const result = getKnowledgeBaseErrorToastContent(
      "Knowledge base name unavailable: test.",
      COPY,
      { status: 500, adviseRename: true }
    )

    expect(result.title).toBe(COPY.genericTitle)
  })

  // Every 409 the backend can raise, checked against what the user is shown.
  // A "neutral" conflict repeats the backend's own sentence, because that
  // sentence is the only thing true of the cause that actually fired. The
  // strings below are the verbatim output of the rename dialog's four 409 exits in kb.py.
  it.each([
    [
      "the caller's own collection holds the name",
      "Target collection already exists: docs",
    ],
    [
      "another tenant holds the name",
      "Knowledge base name unavailable: docs. Please choose a different name.",
    ],
    [
      "the target name already has stored files",
      "Cannot rename to 'docs': that name already has stored files. Please choose a different name.",
    ],
    [
      "another operation holds the lock",
      "Another operation is in progress; please try again later.",
    ],
  ])("relays the backend verdict verbatim when %s", (_case, detail) => {
    const result = getKnowledgeBaseErrorToastContent(detail, COPY, {
      status: 409,
      adviseRename: false,
    })

    expect(result.title).toBe(COPY.genericTitle)
    expect(result.description).toBe(detail)
  })

  it("does not tell a screen without a name field to pick another name", () => {
    // What the detail page and the settings panel actually receive. The backend
    // states the fact only, so no rename advice can reach a screen that has no
    // name to change.
    const result = getKnowledgeBaseErrorToastContent(
      "Knowledge base name unavailable: demo.",
      COPY,
      { status: 409, adviseRename: false }
    )

    expect(result.title).toBe(COPY.genericTitle)
    expect(result.description).toBe("Knowledge base name unavailable: demo.")
  })

  it("keeps rollback toasts concise while preserving rollback context", () => {
    const result = getKnowledgeBaseErrorToastContent(
      "Failed to fully roll back ingest for demo/file.txt: delete failed. Original ingestion error: Model 'text-embedding-v4' not found in hub and no environment configuration available for embedding.",
      COPY,
      { status: undefined, adviseRename: true }
    )

    expect(result).toEqual({
      title: COPY.embeddingTitle,
      description: `${COPY.embeddingDescription} ${COPY.rollbackDescription}`,
    })
  })

  it("falls back to a generic title with a truncated description for unknown errors", () => {
    const result = getKnowledgeBaseErrorToastContent(
      "A very long upload failure happened while processing the document and there are many more technical details that should not become the toast title for end users.",
      COPY,
      { status: undefined, adviseRename: true }
    )

    expect(result.title).toBe(COPY.genericTitle)
    expect(result.description).toContain("A very long upload failure happened")
  })
})

describe("buildKnowledgeBaseErrorResult", () => {
  it("builds a synthetic error result for processing details panels", () => {
    expect(
      buildKnowledgeBaseErrorResult(
        "demo",
        "Failed to upload file",
        "resolve_embedding_adapter",
        "failed.xlsx"
      )
    ).toEqual({
      collection: "demo",
      document_count: 0,
      chunks_count: 0,
      status: "error",
      message: "Failed to upload file",
      failed_step: "resolve_embedding_adapter",
      file_name: "failed.xlsx",
    })
  })
})

describe("normalizeKnowledgeBaseIngestionResult", () => {
  it("maps backend ingest fields into the display shape used by result panels", () => {
    expect(
      normalizeKnowledgeBaseIngestionResult(
        {
          status: "success",
          message: "done",
          doc_id: "doc-123",
          chunk_count: 8,
          embedding_count: 8,
          vector_count: 8,
        },
        {
          collection: "demo",
          fileName: "report.pdf",
        }
      )
    ).toEqual({
      collection: "demo",
      document_count: 1,
      chunks_count: 8,
      status: "success",
      message: "done",
      failed_step: undefined,
      file_name: "report.pdf",
      parses_completed: 1,
      doc_id: "doc-123",
      embedding_count: 8,
      vector_count: 8,
      embeddings_created: undefined,
      error: undefined,
    })
  })

  it("falls back to embedding-derived counts when vector_count is absent", () => {
    expect(
      normalizeKnowledgeBaseIngestionResult(
        {
          status: "success",
          message: "done",
          doc_id: "doc-456",
          chunk_count: 3,
          embeddings_created: 5,
        },
        {
          collection: "demo",
          fileName: "report.pdf",
        }
      )
    ).toMatchObject({
      collection: "demo",
      document_count: 1,
      chunks_count: 3,
      parses_completed: 1,
      file_name: "report.pdf",
      vector_count: 5,
    })
  })
})