We have released v1.0.245. It comes in two halves. The first is an exam coach API for learning apps: a study diary, a log of how the day was spent, target scores and test results, today's study plan with a weekly review, and a way to export it all. The second is about counting correctly and protecting what should be protected: one set of scoring rules, invitations that count seats correctly, sign-in protection for pages that need it, and automated tests that check billing, scoring and the separation between organizations.
Highlights
Exam coach API
A study diary, a daily time log, target scores and results, today's plan and a weekly review, all through the API.
One set of scoring rules
A pass is judged on the exact score, not the rounded one, and a question with no correct option no longer gives free marks.
Invitations and seats
Invitations respect the seat limit, and inviting back someone who was removed never counts them twice.
Page protection
Account settings pages now require sign-in, and unpublished lessons in public courses are no longer shown.
Finish tests over MCP
Put a test's questions in order and read back every setting, time limit included.
Automated tests, English pages
403 automated tests check billing, scoring and organization separation before every release. English help and news now show.
1. The exam coach API, for learning apps
For learning apps such as WakaRoute, a new API helps plan each day of study from what learners record themselves and what Manabu2 has recorded. Only the learner can read and write this data; their organization cannot see it.
1.1 Study diary and daily time log
- One diary entry per day: what went well, what was hard, what's next, plus focus and tiredness
- A log of how the day was spent, in categories such as school, club, cram school, self-study, homework, screen time and sleep. Today and the previous 31 days can be edited
- The day view shows study the learner logged (self-reported) next to, not added to, study done in Manabu2 (recorded automatically), so the same time is never counted twice
1.2 Target scores and test results
- Set a target score per subject for each kind of test: regular tests, mock exams, quizzes and others
- Record the scores that come back from school or a mock-exam provider, together with the maximum each was marked out of
- The gap to target is calculated only when the kind of test and the maximum match. Scores on different scales are never compared, and there is no pass/fail verdict
1.3 Today's plan and the weekly review
- Up to five items for today, in this order: anything carried over, a step back to a missed prerequisite, re-checks that are due, mistakes to fix, then new ground. The subject furthest from its target comes first
- The day's study time comes from the learner's own last week (20 to 120 minutes). Within 30 days of an exam, new ground waits while anything needs fixing
- The learner decides: accept, change, postpone, skip or mark done. The coach never decides for them. Its reasons come from the learner's own data, and it never promises a score
- A Monday-to-Sunday weekly review covers the gaps to target, the plan against what was actually done, the concepts reached this week, and next week's first suggestions. Diary text is never included
1.4 Data export and outcome metrics
- Learners can export their goals, targets, results, diary, time log, plans and decisions, study records and mastery as a single JSON document
- Organizations get aggregate outcome metrics: unaided correct answers after a hint, similar-problem success, re-check retention, plan completion and more. Nothing is reported per learner
- Closing an account deletes the diary, time log, targets, test results and plans
| Addition | What it does |
|---|---|
/api/v1/me/journal |
Diary, daily time log, day view, summary |
/api/v1/me/scores |
Target scores, test results, gap summary |
/api/v1/me/coach |
Today's plan, decisions on it, weekly review |
GET /api/v1/me/export |
Export of the learner's data (JSON) |
/api/v1/organizations/{id}/ai-ops/learning-metrics |
Outcome metrics for the organization (aggregate only) |
2. One set of scoring rules
Lesson quizzes and learning-path tests used to be scored by separate code. They now share one set of rules, and two of those rules have changed.
| Rule | Before | Now |
|---|---|---|
| Passing | Judged on the rounded score. 16 correct out of 23 (69.57%) showed as "70%" and passed a 70% pass mark | Judged on the exact score before rounding. It still shows as "70%", but it does not reach a 70% pass mark |
| A question with no correct option | If no option was marked correct, everyone got the mark, even for a blank answer | Nobody gets the mark. The question still counts towards the total, which makes a missing answer key easier to spot |
Automated tests also confirm that a multiple-choice question is right only when exactly the correct options are chosen, no more and no fewer, and that the same answer sent twice counts once.
3. Invitations and seats
- Invitations respect the seat limit: a pending invitation holds a seat. The limit used to be checked by the screen that sent invitations; it is now part of the invitation itself, so no route can invite past it
- Inviting back never counts a person twice: if HR removed a member, the member was invited back, and HR then added them again, the same person could end up with two memberships and use two seats. An invitation now reuses the earlier membership. Earlier roles, such as administrator, do not come back; the person starts as a learner, like any new invitation
- Invited-back members can sign in: someone removed from their last organization cannot sign in. Being added back by HR lifted that; accepting an invitation did not, so they still could not sign in after accepting. Accepting an invitation now lifts it too
- Closed accounts cannot be invited: a person who deleted their account cannot be invited, and an invitation sent before they did can no longer be accepted. A withdrawn invitation now shows as invalid, where it used to say "you have joined"
4. Page protection
We went through every page that can be reached without signing in and recorded why each one may be public. Automated tests will flag any new public page from now on. The review found two issues, both now fixed.
- Account settings pages: pages such as Change password could be opened without signing in. No personal information was shown; the page said it could not load the user. They now send visitors to the sign-in page
- Unpublished lessons in public courses: a public course listed its unpublished lessons too, and their content could be read from their URL. Only published lessons are shown now. The sitemap already listed only published lessons, so search results are not affected
5. MCP and API changes
| Item | Change |
|---|---|
MCP tool reorder_test_questions |
New. Sets the order of a test's questions. Every question must be listed exactly once; a partial list, or a question from another test, is refused. Reordering a test that has already been taken does not change past results |
MCP tool get_test_answer_key |
Now returns the description, pass mark, whether the test is required, the time limit and the learning path, with questions in the order learners see them, so what was saved can be checked |
MCP tool list_tests |
Shows each test's time limit |
courseCount in GET /api/v1/paths |
Counts only the courses the caller can see, so it matches the number shown in the path's detail. It used to include private and draft courses |
6. Reliability, quality and other changes
- API rate limits shared across servers: the per-minute limit is now counted correctly however many servers are running. If the shared counter's storage is briefly unreachable, the API keeps running and each server counts on its own
- More automated tests: seat calculations, quiz and test scoring, separation between organizations (another organization's courses, members, materials, tests and reports come back as "not found" through the API) and the public catalog rules are checked by 403 automated tests before anything reaches production
- English help and news:
/en-US/helpand/en-US/newsnow show English articles. Until now they could not find any
What is next
The exam coach API will keep improving as WakaRoute puts it to use. Next up: refreshed English help content, in-app notifications for due dates and renewals, and single sign-on (Entra ID and Google) for organizations.