A single pre-deployment command can catch many of the mechanical errors that undermine a math-problem release: malformed data, duplicate questions, mismatched answers, translation drift, broken SQL output and inconsistent sitemaps. In a project case study published September 26, 2026, MozgoQuest contributor Ivan Nedomolkov describes using that approach for 180 Russian-English math problems for grades 1–6. It is a project-specific account, not an independent audit; the pipeline checks defined conditions, while people still judge clarity and teaching quality. Read Nedomolkov’s DEV Community article.
What the pipeline checks
MozgoQuest keeps reviewed problem data in YAML as the editable source of truth. A content check validates that source, then generates SQL migrations and a JavaScript translation bundle. The intent is to find errors before deployment rather than discover them after content reaches the public site.
As an Amazon Associate I earn from qualifying purchases.
The author reports that the command, npm run content:check, validated four YAML sets and 180 original questions—30 for each grade from 1 through 6. It also verified 180 numeric answers, compiled 180 self-reviewed English translations, built 180 problems and 360 hints, and generated two sitemaps with reciprocal hreflang links. These are reported results from the project, not independently verified benchmarks.
How the checks work
1. Validate the authored data first
The structural validator checks required fields and editorial constraints before later checks depend on those fields. Among its reported checks are unique slugs and IDs, valid grade and difficulty ranges, approved topic names, statement and explanation lengths, two distinct substantial hints, numeric answers, authorship metadata and prohibited competition names.
#1 Best Overall
That order matters: downstream answer, translation and generation checks are useful only when the source records are complete and consistently structured.
2. Flag possible duplicates
Statements are normalized for case and punctuation before comparison. The author reports failure thresholds of 0.86 similarity among authored statements and 0.70 when comparing against recovered legacy material. These are project guardrails for identifying likely overlaps—not proof that a question is original or copied. Ownership metadata and editorial review still matter.
3. Recalculate answers safely
Each problem stores an expected answer and a separate verification expression. Rather than evaluate arbitrary Python, the checker parses a restricted Python abstract syntax tree and exposes only sum, range, gcd and lcm as callable names. The author says the pipeline applies exact numeric tolerance rules and stops when the computed and stored answers disagree.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →This makes answer checking a build-time gate, but only for expressions the restricted evaluator can handle. It does not by itself establish that a problem is well designed or that its intended interpretation is unambiguous.
4. Compare Russian and English versions
The two language sets must use the same slugs and form an exact one-to-one match. The validator compares grade, topic, answer and two-hint structure, and checks numbers in statements, explanations and hints to catch numerical drift. Translations also need an explicit review status; missing or unreviewed entries are excluded from the public runtime bundle.
Matching numbers can reveal that a value changed in translation, but it cannot tell whether the English is natural, age-appropriate or faithful in meaning. Those judgments need a reader.
5. Test generated database and sitemap output
The pipeline applies generated SQL to an in-memory SQLite database to check problem and hint counts, intended IDs and inactive status. The described release flow inserts rows and hints as inactive, checks the results, then activates only the intended ID range. This reduces the chance that a partial or unintended set becomes public during a release.
The site output is checked too. In the reported run, the Russian sitemap had 230 URLs and the English sitemap had 231; each included 180 task pages and 16 populated grade-topic hubs. The pipeline also validated reciprocal hreflang links. These counts describe that run, not permanent site totals.
Best Value
- Full of different activities to help your child develop their skills
- Contains one sixty-four page workbook
- Available in a variety of different age groups
- Available in different themed activity books
- Made in USA
What a passing run does—and does not—mean
A successful content check means the content passed the specific structural, originality-similarity, answer, translation-parity, SQL and sitemap conditions encoded in this project’s pipeline. It does not mean every defect has been found, nor does it establish learning outcomes. Nedomolkov summarizes the human limit this way: “Automation can prove that two stored numbers match. It cannot prove that a problem is interesting, age-appropriate, clearly worded, or pedagogically useful.”
Human reviewers still need to consider whether a child can understand a task without hidden context; whether the first hint leaves room to solve it; whether the second suggests a method without giving away the answer; whether the explanation teaches a reusable idea; and whether the English reads naturally.
What remains outside this content check
The author distinguishes the content-specific pipeline from broader deployment checks. Unit tests, browser scenarios, a Worker dry run and public health checks are described as separate work, not as part of npm run content:check. A passing content run therefore should not be mistaken for a complete deployment test.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
