| 1 | import { describe, expect, it } from 'vitest' |
| 2 | import { findUnsupportedPrecisionClaims } from '../../../src/main/thinking/content-credibility' |
| 3 | |
| 4 | describe('thinking content credibility', () => { |
| 5 | it('flags unsupported exact metrics when no sources exist', () => { |
| 6 | const issues = findUnsupportedPrecisionClaims({ |
| 7 | hasSources: false, |
| 8 | markdown: [ |
| 9 | '# Thinking Brief', |
| 10 | '', |
| 11 | '## Page 1: 性能跃迁', |
| 12 | '- Role: data', |
| 13 | '- Objective: 说明性能变化', |
| 14 | '', |
| 15 | '- 单token推理成本下降70%', |
| 16 | '- HumanEval基准从67%提升至89%', |
| 17 | '- 工程侧应关注推理效率和部署复杂度' |
| 18 | ].join('\n') |
| 19 | }) |
| 20 | |
| 21 | expect(issues).toHaveLength(2) |
| 22 | expect(issues.map((issue) => issue.text)).toEqual([ |
| 23 | '- 单token推理成本下降70%', |
| 24 | '- HumanEval基准从67%提升至89%' |
| 25 | ]) |
| 26 | }) |
| 27 | |
| 28 | it('allows structural numbers and sourced exact metrics', () => { |
| 29 | expect( |
| 30 | findUnsupportedPrecisionClaims({ |
| 31 | hasSources: false, |
| 32 | markdown: [ |
| 33 | '# Thinking Brief', |
| 34 | '', |
| 35 | '## Topic', |
| 36 | '2026 AI模型的进化', |
| 37 | '', |
| 38 | '## Setting', |
| 39 | '技术分享会，时长约10分钟', |
| 40 | '', |
| 41 | '## Page Count', |
| 42 | '9' |
| 43 | ].join('\n') |
| 44 | }) |
| 45 | ).toHaveLength(0) |
| 46 | |
| 47 | expect( |
| 48 | findUnsupportedPrecisionClaims({ |
| 49 | hasSources: true, |
| 50 | markdown: '- 单token推理成本下降70%' |
| 51 | }) |
| 52 | ).toHaveLength(0) |
| 53 | }) |
| 54 | }) |
| 55 |