Repository navigation
Releases
Releases · apitlekays/alms-releases
SuluhALMS 0.1.36
New
- A project that failed (marked ERROR on the projects screen) now has a Delete button under its card. It asks before deleting, and never touches the original file or link.
- Each project card now has a label at the top right of its cover saying where the course came from: PDF (blue), a document such as DOCX (purple), WEB (green) or YOUTUBE (red).
SuluhALMS 0.1.35
Fixed
- The build countdown no longer sits at "0:01" while a unit is still being made. When a step takes longer than usual (a long unit, or a lesson being rewritten after a check), the countdown now says "taking longer than usual" until a real estimate is available again.
SuluhALMS 0.1.34
Fixed
- "Start learning" on the Building screen now opens a new unit at 1. Your task, not at 3. Read. The course map's Learn button and the Back buttons from practice, the unit test and Apply it also open a unit where you left off.
- The menu bar and the Building countdown now count a unit as ready only once its task is written, the same moment the unit opens.
- From the next update on, the update window shows emphasised words in its notes in italics.
Improved
- Course screens stay quick as more courses are built: the app no longer reads the history of every finished job to tell what is still being made.
SuluhALMS 0.1.33
Fixed
- A new unit now always opens at 1. Your task. Before, a unit could open at 3. Read when its task was still being written and its video was still being made.
Improved
- While a unit's video is being made, the Watch step stays in the path with a turning mark and the words "Video being made", and the way forward goes on to Read. The mark goes away by itself once the video is ready.
- The panel beside a unit (what it teaches, and the Tutor) now keeps a margin from the right edge of the window.
SuluhALMS 0.1.32
Fixed
- Words the agents emphasise now appear in italics in lessons, corrective lessons and Tutor answers, instead of between asterisks.
SuluhALMS 0.1.31
Improved
- Video big pictures now use the scene that shows how the ideas connect (a cycle, a chain of causes, a timeline, a side-by-side comparison), and vary from one unit to the next, instead of usually showing the same tree of parts.
SuluhALMS 0.1.30
New
- Unit videos have a new look, in two styles: Studio dark (the default) and Paper light, which suits bright rooms and projectors. Choose in Settings, Background and video, "Look of videos". Every scene now shows the unit, how far through the video you are, and what kind of scene it is.
- Videos can now show designed scenes for the big picture and for key facts: comparison tables, timelines, key numbers, bar and line charts, parts of a whole, quotes, myths and facts, analogies, cause and effect chains, cycles, 2 by 2 grids, overlaps, trees, checklists, glossaries, formulas and code. The Media Producer picks the one that fits and each part appears as it is spoken. Charts and numbers only ever show numbers found in your source.
Improved
- English videos are now narrated by a native English HeyGen voice; Bahasa Malaysia videos keep the course owner's own voice.
- Narration reads numbers, percentages, money, units, dates and short forms properly in Bahasa Malaysia and English ("80%" is said "lapan puluh peratus", "RM2.50" "dua ringgit lima puluh sen", "hlm. 12" "halaman dua belas"), while the captions still show them as written.
- Every video is now at the same comfortable loudness, so you no longer need to change the volume between videos.
- Videos are drawn scene by scene, two at a time, so they are ready sooner; when a video is made again, only the scenes that changed are drawn again.
SuluhALMS 0.1.29
Improved
- A safety net for the agents' AI decisions: if the AI answers outside the allowed choices twice in a course, that course goes back to the app's fixed rules until you restart SuluhALMS, so no more AI calls are spent on it. The agent log tells you when this happens.
SuluhALMS 0.1.28
New
- The agents now use judgement where it helps. If you miss the same objective again after a corrective lesson, the Orchestrator chooses what to try next: another way of teaching it, smaller steps, a recap of an earlier unit it builds on, going back to the right section or moment in the video, or a short break before the retest. The corrective screen tells you why and gives you a button for it.
- The Learning Analyst reads your wrong answers for a misunderstanding they share. The corrective lesson then teaches against it first, and the Tutor knows about it too.
- When a lesson or question keeps failing the Quality Auditor's checks, the Orchestrator decides how to finish it: one last rewrite with clear advice, a different kind of question, or leaving the question out when others still cover what it tests. A question is never kept with a wrong answer key.
- The agent log now says who made each of these decisions: the AI model, or a rule. Settings, AI explains that scheduling stays rule-based and the AI decides only where judgement helps.
Improved
- While the agents are working on something you are waiting for (a corrective lesson, retest questions, practice questions, integration activities or the capstone), you now always see a moving status, so the app never looks stuck. Practice questions appear by themselves once they are written.
Fixed
- Opening the retest again while its questions were still being written could start the same work several times, which cost more and made the corrective lesson wait longer.
SuluhALMS 0.1.27
Improved
- The Tutor now opens with a simple greeting ("Hi, I'm SuluhTutor. How can I help you?") instead of a list of ready-made questions. Just ask in your own words.
- Bahasa Malaysia now addresses you politely as "anda" everywhere: on every screen, and in new lessons, questions and videos the agents write. Courses built before this update keep their wording.