eval-summary

To validate translation quality, compare readings.ts summary fields against original Korean Markdown to automatically identify gaps, track issues, and propose improvements for translation consistency.

You can audit readings.ts summaries by running an evaluation over weekly readings or individual slugs to verify required fields, record quality issues, and output an evaluation-report.json for review.

Automated comparison of tldr and chapterSummaries works by scanning readings.ts, verifying required fields exist, and comparing their content directly against the original Korean Markdown to flag discrepancies.

No, suggested corrections to readings.ts require user confirmation before applying changes, ensuring you maintain manual oversight of translation quality updates proposed during the evaluation.

The validation supports both weekly batches and individual reading slugs, allowing you to target an entire week or a specific reading path like week1/deep-dive-llms for issue tracking.

You should generate an evaluation report when maintaining translation consistency across reading materials, identifying gaps in KR translations, or iteratively refining readings.ts summaries based on identified issues.