Principles first
Three to eight principles, each with a worked phrase and a plausible pitfall.

Decision-based scenario training a teacher builds in minutes: illustrated scenes, spoken narration, choices with consequences, a world that remembers, and reports that show a trainer what to teach next.
A choice can diverge, converge at a deliberate bottleneck, or lead to a distinct outcome.
Release access is checked before download
Commercial access is read from the live release manifest. Prices, credit access and archive availability are never guessed here.
Policies, procedures and notes become a context for decisions.
Learners see what follows a choice before the explanation.
Skills and principles return to the material that shaped them.
Complete buyer and teacher guide
This is the complete v3.0.0 release guide. The live release manifest remains authoritative for the downloadable package and currently identifies it here as Checking release….
1 · The player
Three to eight principles, each with a worked phrase and a plausible pitfall.
An establishing frame and hook put the learner in the place.
The scene and question fit together. Generated scenarios use three choices; authored scenarios may use two to four options.
What happened appears before the explanation; Engagement, Trust and Tension respond.
Settled decisions can be re-read without resubmission or re-scoring.
Authored or imported scenarios can use flags, readable documents, picture hotspots and remembered visual tags.
High tension can switch to crisis words and an alternate picture.
Achievements can recognise holding ground, recovery, a faultless route and leaving the room settled.
Every decision is compared with the roads not taken before formal source-grounded knowledge returns.
ROLE · SUPPORT WORKER
Context: an anxious family at a medication round. Principle: acknowledge concern without guessing. Pitfall: promise an answer before checking the record.
A family member asks why a resident’s medication is late. Which response protects both the relationship and the safety process?
3 · Billing contract
| Plugin-displayed quote | Credits | Approx. USD |
|---|---|---|
| Scenario only | 100 | US$10 |
| Scenario + illustrations | 150 | US$15 |
| Scenario + narration | 150 | US$15 |
| Scenario + both media types | 200 | US$20 |
| Billed runtime operation | Credits | USD |
|---|---|---|
| Scenario generation | 20 | US$2.00 |
| Wizard population | 3 | US$0.30 |
| Each suggestion | 1 | US$0.10 |
| Each image request | 5 | US$0.50 |
| Each narration request | 5 | US$0.50 |
Ten credits equal one US dollar. Learner play and replay use no generation credits. A confirmed refunded run spends nothing against the daily allowance. Until confirmed, treat a refund as support-required. The 50-credit one-time plugin acquisition is separate from these usage charges and is not part of this mismatch; access remains controlled by the live Marketplace controls above.
4 · Frames and voice
Every screen that needs a picture can have its own frame. Six treatments are documented: photorealistic, cinematic, illustration, watercolour, noir and oil. Repeated cast sheets and treatment instructions support a coherent series, without guaranteeing identical provider output. Briefs exclude readable text, captions, logos, identifiable real people, children, injury, violence and weapons. Documents remain structured text rather than unreadable generated text inside an image.
Text-to-speech supports en-AU · en-GB · en-US · en-NZ · es-ES · fr-FR · de-DE · it-IT · pt-BR · nl-NL · hi-IN · id-ID · ja-JP · ko-KR · cmn-CN · ar-XA · vi-VN · th-TH. Narrator, male-character and female-character voices are configurable separately. Narration can cover the title, situation, document, question and lettered options; it is never the only reading route.
Optional synthesised cues distinguish ordinary progress, strong calls under pressure and poor outcomes. There is no game-show buzzer. Sound is a separate site-wide control from narration, costs no credits, carries no unique information and can be muted by learners.
5 · Training-manager evidence
A score says who may need help. Decision evidence says what to teach on Monday.
A poor choice made while the learner marked “I am not sure”. They recognise the edge of their knowledge; a rule or example can close it.
A poor choice made without uncertainty. The learner believes the answer is sound; seeing the consequence is more useful than repeating the rule.
The optional uncertainty tick is read before the outcome is shown and is never required. Cohort bars use the scenario’s decision quality, rank teachable decisions worst first and report decisions rather than profiling people. Below five answers, the report says there is too little data rather than presenting a percentage as fact. The attempt report separately shows each learner, attempt, score, outcome band, start time and decision route. Confirmed deletion removes answers, event log, gradebook mark and completion together.
6 · Administration
| Site setting | Reference |
|---|---|
| Credentials and connection | LMS Labs site ID/API key via Central Config when installed; API host, timeout and connection status |
| Spend and source | Daily allowance per user/24 hours; maximum source characters; generation-job retention |
| Media controls | Allow illustrations, narration and scenario sounds; narrator, male and female voices |
| Learner signal | Allow learners to mark that they are not sure |
| Defaults | Language, theme and every activity default pre-fill new activities |
| Activity setting | What it controls |
|---|---|
| Illustrations / narration | Generate scene images / spoken audio |
| Show progress / readings / debrief | Position, Engagement/Trust/Tension and full debrief |
| Attempts / replay | Attempts may be spent; abandoned runs do not count |
| Grading / completion | Highest, last, first or average; finish and minimum-score rules; point grades and scales |
Find settings at Site administration → Plugins → Activity modules → AI Branched Scenario. Eight capabilities are checked at entry points: view · attempt · addinstance · manage · publish · generate · viewreports · deleteattempts. Keeping generate separate from manage lets a teacher edit without permission to spend credits.
7 · RTO buyer reference
No AI-provider credential is stored in the plugin.
The plugin authenticates to LMS Labs with a site ID and API key. Supplied source content, wizard context, media-generation text and site authentication leave Moodle for generation. LMS Labs carries out external processing; this guide makes no external retention or erasure promise. Do not paste personal, learner or commercially sensitive material into source content.
8 · Operational release
| Requirement | Current reference |
|---|---|
| Promoted release | Checking release… |
| Moodle | Moodle™ 4.4 - 5.2 |
| PHP / databases | As required and supported by the installed Moodle |
| Component | mod_aibranchedscenario |
| Maturity / licence | Stable / GPL v3 or later |
| Cron | Generation uses an ad-hoc task; standard Moodle cron must run |
| Backup / restore | Activity content, revisions and generated media are supported |
9 · v3.0.0 editorial evidence
The author’s v3.0.0 build report records these checks for that build. LMS Labs did not independently rerun them for this documentation update. They are build-report evidence, not live-service success, uptime evidence or accessibility certification.
2,295
automated checks across Moodle 4.4, 4.5 and 5.2
45 / 364
unit tests / assertions
Moodle CS
0 errors reported
96 rendered cells
three frame heights × four widths × light and dark
7 sweeps
alignment, contrast, hostile themes and mobile
Negative control
layout sweep must fail with its fix disabled
78 invariants
audit baseline across content, billing and Privacy API
When a fault is found, ask which surface it lives on and whether any check can see that surface. If nothing can, build the check before the fix.
The supplied audit history records faults that coding rules alone would miss: wire fields dropped by a mapper, an attempt-table alignment error, a report query producing a confidently wrong 100%, and schema coverage that reached only one supported Moodle. This documentation does not claim a current external-provider run has succeeded.
10 · Practical answers
Not inside the plugin. Route B lets you use an external assistant and paste the result back; its provider fees may apply, and LMS Labs media requests are still charged.
Yes in an authored or imported scenario written with hotspots. Lettered, keyboard-reachable options remain available. Automatic generation is not currently claimed to emit hotspots.
No. Learner-earned state is filtered server-side. Individual attempts are available only in the protected attempt report.
No. It identifies decisions with gaps or misconceptions. The separate attempt report contains individual routes behind the reports capability.
The job records failure. Only a confirmed refund can be described as uncharged; an unconfirmed refund is support-required. Missing media can be requested individually.
No. Published scenes and media play from Moodle. Generation requires outbound LMS Labs access.
Yes. Review every route and edit the draft before publishing. Publishing creates an immutable revision for learners already in progress.
It preserves source-grounded decisions and per-decision records, but does not guarantee compliance or an audit outcome.
Keyboard reachability, visible focus, optional narration, words as well as colour, and reduced-motion support are documented. No manual assistive-technology audit or certification is claimed here.
11 · Moodle administration
/mod/aibranchedscenario and visit Notifications.12 · Before learners arrive