Learners can now bring their own material into a tutoring conversation. They attach a document or an image inside the chat and ask Walter to explain or summarise it — in both Walter web tutoring and NoodleMe.
How it works:
⚠️ This is for understanding material, not marking it. Walter won't review, correct, rewrite or grade a learner's own work through this. Essay feedback stays where it is — in the essay activity you set up.
💡 Available per organisation and per agent, so you decide where learners can attach files.
A learner can now say what they're working towards when they first set up a Space, and every tutoring session in that Space is steered by it.
What's new:
💡 Spaces without a stated goal carry on exactly as they do today, and a topic whose intent isn't clear defaults to learning mode.
Quiz generation now runs through a reworked pipeline, giving more consistent turnaround — including for short, plain-text documents where the wait was previously less predictable than it should have been.
✅ No action needed — applies automatically.
When a document doesn't pass the pre-upload check, the Admin Portal now tells you the actual reason — an unsupported file type, a file over the size limit, or a quota that's been reached — rather than a single general message.
You can also retry the one file that failed, instead of starting the whole batch again. The rest of the set uploads as normal.
The platform's type scale, colour tokens and activity icons have been updated as part of our design system work, for a more consistent look across the Admin Portal and learner-facing screens.
✅ No action needed — applies automatically.
Teaching the same subject across several LMS courses? You can now point a new course at an agent you've already built.
When you add a Walter activity through deep linking, you choose:
The choice happens in your LMS, as part of the setup you're already doing — so a new semester's course can pick up everything you've built, without leaving the LMS.
💡 This links a course to an existing agent; it doesn't merge or move content between agents. The course-mapping screen in the Admin Portal is still there if you prefer it.
When the agent quizzes a learner mid-conversation — because they asked, or because it's checking understanding — two things have improved: the questions it asks, and how it handles the answer.
Better questions
Better marking and feedback
Why it helps: marking that a learner can rely on is what makes the check on understanding worth doing — and it means a demanding question can be asked with confidence.
💡 Where questions come from is set per agent in the Admin Portal. In agentic mode the agent draws on the knowledge group the learner is working in; otherwise it asks what they'd like to be quizzed on first, and scopes to that.
A NoodleMe tab now sits beside Overview in the Organisation Dashboard, with seven measures of adoption and sharing across your own members — so you can see whether Spaces are being created, and whether learners are sharing them with each other.
Useful when you've enabled NoodleMe for a cohort or a faculty and want to know how it's landing before opening it up more widely.
⚠️ Visible to superadmins in organisations with NoodleMe enabled.
If your learners reach the chatbot mainly through your LMS, the LMS already supplies your branding and course context — so you may prefer a clean, logo-free header, and the vertical space it gives back to the conversation.
A new toggle in white-label settings shows or hides the brand logo on the chatbot sidebar. With it off, the header collapses and the navigation moves up to reclaim the space.
💡 On by default, so nothing changes unless you turn it off.
.pptx files whose embedded images can't be read, and still generates questions from the rest of the deck.
A role play used to be driven by a single fixed AI character. You can now define several personas on the same role play, so one scenario can be practised against different people.
How it works:
Why it helps: the same scenario lands differently depending on who's on the other side. A nursing student can practise the same conversation with a distressed patient, a calm patient, and a doctor — building judgement rather than a single rehearsed script.
💡 In this first release, learners practise with one persona at a time. Several personas in the same scene, taking turns, or switching part-way through are not supported yet.
Moodle content integration was limited to a single instance per organisation, while D2L Brightspace already supported several.
What's changed:
Why it helps: institutions running separate Moodle instances per faculty, campus or cohort no longer have to choose one.
Following the move to Avaturn, the Create your 3D Avatar link opened straight into selfie capture, with no visible way to upload an existing photo. It now opens the full choice screen, so you can use your camera or upload from your device.
💡 If you tried to recreate an avatar after our 21 July update and could only take a selfie, this is why — try again and the upload option will be there.
The Chat Sessions card now counts unique sessions across the period you've selected, and the Usage Trends chart continues to show per-day uniques — which is the right basis for a trend.
Both now say which they are, so the two figures read as two deliberate measures rather than a discrepancy.
