Frequently asked questions

No. Teams define organisational and manager scope. Learning groups are cross-team cohorts; membership grants no roles or automatic enrolment.

Select this page selects the displayed rows. You can retain selections across pages up to 100 people per reviewed action. Whole-filter background jobs are not implemented.

Keep the action dialog open and choose Retry same request. The same request ID and per-person receipts prevent duplicate effects. Confirmed people remain confirmed; failures are reported individually.

SCIM owns that group and its memberships. Change them in the identity provider. Company-managed learning groups can be edited by a company administrator.

No. It highlights the collection for the recipients without following it for them or creating enrolments. Use Assign a course or Assign a learning pathway when learning is required.

No. Following saves your interest. Individual course enrolment and lesson access remain separate.

No. Each assignment pins a published pathway version and its exact course versions. New publications apply to new assignments.

Only if prior credit is enabled for the assigned pathway, the course version matches exactly and completion is within its configured age window. Credit is selected at assignment time; certificate expiry is not used as a completion rule.

No. Only required courses determine progress and unlock sequential stages. Optional learning remains available according to the stage gate.

No. It stops new launches through that assignment but preserves existing individual enrolments and learning history.

Sequential pathways require every required course in the preceding stages to be complete. Your manager can inspect the same timeline. A separate course assignment is independent of this programme gate.

Yes. Company settings → AI & automation supports OpenRouter, direct OpenAI, Anthropic / Claude and DeepSeek. Your operator must approve the provider and company-owned keys in Platform console → Organisations → AI governance. Custom API URLs need a reviewed adapter; changing a model ID does not change the provider.

Keys are encrypted on the server using the existing key infrastructure and never returned to the browser. Recent sign-in and company administrator authority are required for changes. Editing or rotating disables a connection; test and activate it again. Removing a saved key retains audit and usage history; revoke it at the provider separately.

No. Each feature has one explicitly selected connection. A failed or disabled connection never silently switches providers or billable accounts. Changing provider settings requires refreshing the generation screen and reviewing the destination again.

Company reference guides, checklists, templates and policy examples. Current resource entries are plain text with preserved paragraphs and downloadable text. Reading a resource does not complete a course.

In Course Studio choose Draft with AI, enter a goal and audience, optionally select permitted resources, then generate and review the proposal. Acknowledge review to open an editable draft; publish separately.

The operator must enable the feature for an explicit company allowlist. Active organisation-wide authors, administrators and owners can use the author pilot. Learners and managers without author permission cannot invoke it.

The current pilot answers from one to three explicitly selected text resources, within 32,000 combined UTF-8 body bytes. PDF/DOCX ingestion, OCR, web crawling and whole-library retrieval are not included.

The server verifies that quotations occur in the selected source text and uses server-supplied titles and revisions. This does not prove that an answer’s interpretation follows from the quote. Review the evidence; the assistant can abstain when sources do not support an answer.

The brief/audience or question and explicitly selected resource titles/text go to the explicitly configured AI provider. Source selection requires consent. Other resources, learner records and existing courses are not automatically attached.

The author pilot defaults to 20 shared course/answer/test attempts per company per UTC day, within company and installation limits, with at most two concurrent requests per API process. Failed or cancelled provider attempts count. Requests have bounded duration and output size; there are no automatic retries.

No. AI produces an answer or an unsaved proposal. Accepting a proposal uses normal draft saving; publication remains explicit. The model cannot select tools, enrol learners or send messages.

These frameworks were assessed but are not installed. Current bounded generation, quotation validation and review flows use Go. Durable branching agents and resumable workflows would justify a separate evaluation rather than adding a runtime for its name.

The server rechecks access and source revisions after generation and rejects a stale result. The course acceptance UI checks again before opening the proposal as a draft. Refresh and regenerate against the current source; the check is not a permanent lock on future edits.

ZubLMS is a company learning workspace for creating, assigning and completing training, managing reference resources, communicating and reviewing outcomes. Available features depend on role, company configuration and the installed release.

The public website explains the product and offers sample learning. Your signed-in workspace contains private company records and role-specific tools. You can return to the public site without making company content public.

The explicit demo uses fictional people and learning records. Demo actions do not prove that an installation’s identity, email, storage or worker services are configured.

Open My learning in Learning space. The catalogue helps you discover published learning; it is separate from your assignments and from public sample courses.

Start with your company learning administrator using the support details in your workspace. Include the screen, action, time and a non-sensitive error description. Never send passwords, tokens or private learner exports.

Use the contact page to agree a pilot, features, hosting, migration, support and commercial terms. The website is not a self-service subscription checkout and does not establish an SLA or fixed price.

Keycloak is the identity service used for password and federated sign-in. The browser redirects to the approved authentication hostname, then returns to the LMS. Branding can be consistent while authentication remains a separate security boundary.

Use the account issued by your organisation or its approved SSO provider. Public documentation does not publish demo credentials or grant accounts. Ask your administrator if you have not received access.

Open your profile menu and choose Sign out. The LMS ends its session. Identity-provider and upstream company sessions can have separate logout behaviour; verify the configured SSO logout flow for shared devices.

Use the profile/account option in the user menu for supported profile fields. Changing a contact email does not change an approved identity-provider issuer/subject link or grant a different membership.

Use the recovery option offered by your configured identity provider, if enabled and tested. Company SSO passwords are managed by that company’s provider. Branded templates alone do not mean password recovery is enabled.

Learning, authoring, team management and company administration are separate permissions. Owners manage sensitive access. Platform operator access additionally requires an explicit deployment allowlist; being a company administrator does not make you a platform operator.

Yes, when it has approved memberships. Use the workspace/company switcher. The selected company determines the session scope; one membership does not grant access to another company.

Menus reflect role and configuration. The server also checks permissions, so manually entering a hidden URL does not grant access. Ask an owner to review your intended role and team scope.

Company access is restricted and affected company browser sessions are ended through the administration workflow. Reactivation requires a recovery-qualified operator procedure; it is not a casual undo switch.

Reopen the assigned course from My learning. The app restores acknowledged progress for the assigned version. SCORM packages also decide how to use their saved bookmark and suspend data.

Completion follows the supported activities and server-validated progress/assessment rules. Opening a page, watching a few seconds or saving a bookmark is not automatically course completion.

Native quiz activities use the authored question, options and correct answer. Submit through the player and use the recorded result. Authors must review quiz accuracy before publishing, including AI-generated questions.

Evidence rendering can run in a background job after completion. Inspect the achievement’s state. Queued or rendering means the file is not ready; a failed render needs investigation rather than repeated course completion.

Assignments are tied to published course versions. Editing a draft or publishing a newer version does not silently replace content already assigned. Coordinate new assignments or refresher cycles with your manager.

No. Bookmarks are shortcuts. They do not enrol you, complete learning or preserve access after a resource is archived or your permissions change.

A complete offline learning/synchronisation mode is not promised. Keep a connection for acknowledged progress, assessments and package saves. Do not assume a screen displayed from browser memory has saved your work.

The web interface is responsive and supports smaller screens. A native mobile application and its authentication contract are separate work. A third-party embedded course may still have its own fixed-width or accessibility limitations.

It helps you see assignment deadlines from the authorised workspace. It is not a promise of live-class scheduling, calendar-provider sync or automatic reminder email delivery.

Authors organise courses into modules with goals and sections/activities. Supported native sections include reading and knowledge checks, with configured video and approved SCORM package activities available through their respective workflows.

Yes. Course Studio provides module/section collapse controls and section movement within modules. Module controls arrange the outline; the section’s module field changes its placement. Sidebar section reordering is not the course-building feature.

Yes. Supported rich text can be edited visually or through HTML source mode. Content is sanitised; source mode does not grant arbitrary scripts, unsafe embeds or unrestricted HTML.

Complete drafts autosave after a pause when there are no blocking errors. Check the saving/saved state. Incomplete content, pending source edits, conflicts or connectivity problems can prevent saving.

No. Draft preview is for review and creates no learner credit. Publishing is a separate explicit action that creates a course version.

Revision checks reject stale changes rather than silently replacing a newer revision. Preserve your text, reload the latest draft and reconcile the changes before saving again.

Use only material you are entitled to upload, embed and distribute. A technically supported import or embed does not grant copyright, redistribution rights or accreditation.

Authors can use the supported YouTube/Vimeo embeds or configured Vimeo API playback. Vimeo API resolution supports progressive MP4 and active captions, subject to provider permissions. Arbitrary iframe HTML, generic signed-URL pasting, HLS/DASH and DRM are not implemented in this flow.

Supported players can report playback events and checkpoints through the activity system. Detailed events depend on the integration. Optional analytics delivery is separate from server-authorised learning progress and completion.

The delivered import/player flow supports self-contained single-SCO SCORM 1.2 packages with configured scanning and isolated hosting. SCORM 2004 and multi-SCO sequencing are not supported by this flow.

Open Learning packages, import the ZIP, resolve validation/scanning errors, preview the exact export and approve it. Attach the approved package version to a course activity, publish the course and assign it.

The current flow allows a ZIP up to 50 MiB, expanded content up to 256 MiB total and 64 MiB per file, within the company’s 1 GiB package storage budget. Unsafe paths, nested archives and unsupported external assets are rejected.

A replacement is a new immutable package version. Existing assigned course versions retain their original package and progress. This does not migrate suspend data between incompatible exports.

Authors can download the original uploaded ZIP where authorised. Generating a portable SCORM package from native ZubLMS content and hosted launcher distribution are separate capabilities still requiring delivery; Original ZIP does not perform either.

No blanket compatibility or certification is claimed. Qualify the exact export with representative browser, mobile, resume, scoring and completion tests. Authoring-tool branding alone is not compatibility evidence.

Authorised managers and administrators can assign published learning within their current company/team scope. The server checks recipients; selecting a broader UI filter cannot widen that authority.

The current customer UI includes an existing-member directory and scoped access controls. New-member invitation/provisioning UI and editable team hierarchy remain separate delivery work. Agree onboarding with your operator.

Company and manager insights include completion summaries, assignment status, deadline windows, course completion bars and team/topic heatmaps. Definitions and counts accompany the charts so colour alone is not required.

No. They use the authorised workspace snapshot and show a limitation notice when data is bounded. Learners started is not daily active users or minutes watched. Do not treat a filtered snapshot as an unlimited historical warehouse.

Dashboard filters narrow the visible charts. They do not automatically alter the scope of a separately requested export. Confirm the report’s own scope and current permissions.

Exports are background work. Inspect request history and worker health if the state persists. A queued request does not mean a file is available, and repeatedly requesting the same report can create unnecessary work.

Report downloads require current authentication, scope and an unexpired entitlement. The report centre is not a public file-sharing service. Handle any downloaded personal information under your organisation’s rules.

Access can change after the file was generated. Expiry, role changes, narrowed team scope or revocation can block a download. Request a fresh report only if you still have the required authority.

Messages contains human-authored announcements with recipient and delivery controls. Learning inbox is an action list built from learning records. An item in Learning inbox is not proof that an email was sent.

In-app messaging is included. Optional SMTP alerts require operator setup and recipient opt-in. SMS, WhatsApp, push, Teams/Slack, threaded chat and attachments are not delivered by the current announcement centre.

Recipients do not see the audience list. Senders can inspect their own delivery ledger. Administrators cannot browse arbitrary private message bodies simply because of their role.

Optional email requires a configured adapter, a usable contact address, recipient opt-in and valid current access/scope. Missing consent or changed authority can skip delivery rather than silently overriding preferences.

No. Accepted confirms SMTP submission acceptance. It does not establish mailbox delivery, opening or reading. Provider delivery/bounce integration is separate work.

The worker cannot prove whether an external hand-off was accepted, for example after interruption. Do not blindly retry: an operator should inspect provider records using the stable Message-ID before deciding on another alert.

Prepared templates support branded HTML and plain text. Application mail and Keycloak identity mail have separate configuration. Templates do not, by themselves, enable automated assignment/reminder workflows or password recovery.

The configured local setup uses a mail catcher for application and identity email. Production SMTP is separate. A captured local message is test evidence, not an email delivered to an external recipient.

Yes. A company can configure up to 20 independent SAML/OIDC connections, each with its own name, callback, status and routing configuration. Display order does not silently make a provider the default.

Company owners use the SSO area in Company settings to prepare and validate a connection and request activation. A platform operator reviews and activates the exact revision; the browser form alone does not make it live.

Copy the exact connection-specific value displayed in the setup workflow into your identity provider. Do not guess a path or reuse another company/connection’s callback.

No. Authentication is separate from approved LMS identity linking, active membership and role grants. Matching email addresses alone are not sufficient for automatic account linking. SCIM lifecycle provisioning is separately scoped.

Validated means configuration checks passed. Active means the operator activation step is complete. Verified means the intended provider callback succeeded for the expected connection/revision. These are distinct milestones.

SSO enforcement and automatic account linking are not enabled by this setup. Keep a tested recovery route. Explicit provider selection avoids guessing; a failed provider is not silently replaced with another company connection.

Supported company settings include identity/support fields and approved logo/accent values. Branding controls preserve the product navigation design. Arbitrary HTML, remote SVG logos or unrestricted theme injection are not accepted.

Yes. Explicitly authorised platform operators use Platform console to list clients and control slug/domain self-service per client. It is not yet a complete billing, subscription or self-service company provisioning portal.

When the operator allows it, an authorised company administrator can set a unique slug through Workspace addresses. Use the generated sign-in link and review the current revision before changing it.

The local feature supports domain claims and DNS ownership verification. The UI then shows awaiting hosting setup. Certificate issuance, proxy routing, custom-host sessions and identity callbacks require a separate activation workflow.

A TXT challenge proves control of a hostname at the time of checking. It does not create a certificate, route traffic or configure secure cookies and callbacks. Your operator must qualify those hosting steps separately.

Yes. Each control can be disabled independently. Disabling freezes self-service while preserving existing slugs and domain claims; it is not a company suspension or a domain disconnection.

Configured destinations include Google Analytics 4, Mixpanel, Klaviyo and Meta Conversions API. Available adapters are not automatically active, and Meta server-side delivery is not the same as installing an unrestricted browser pixel.

Analytics and marketing choices are separate and default to off. Use Privacy choices to change them. Withdrawal stops future optional submissions through those controls; it cannot recall events already delivered.

Open Documentation → API reference to search actual operations, inspect request/response schemas and download the Go-generated OpenAPI contract. Examples are inspection aids, not an unauthenticated live request console.

A general customer API-key/client-credentials interface and credential-management UI are not available through the current browser contract. Agree and qualify machine-access scopes and authentication before building an integration.

Internal foundations do not equal a qualified self-service product. Customer-facing SCIM lifecycle management, general webhook subscriptions and integration credential issuance remain separate delivery work.

Normal internal navigation uses the React/TypeScript app and client routing for a continuous experience. Identity redirects, external destinations and downloads legitimately use browser navigation. Network or initial-load waits should still be handled clearly.

Public product and documentation pages are prerendered with metadata and a generated sitemap. Private workspaces and company records must remain outside public indexing. Local/preview hosts are not intended for search submission; inspect the deployed release before submitting its production sitemap.

The core uses Go with modular domain services, PostgreSQL, River background jobs and a React/TypeScript frontend. Keycloak handles identity; OpenBao supports protected key operations. Object storage is adapter-backed, with deployment-specific configuration.

The server resolves company/user authority and checks roles and scope; database row-level policies provide another boundary. Frontend filters are not the security boundary. Operators require separately configured authority for cross-client controls.

Do not describe the platform that way. It uses HTTPS and protected storage/key handling for defined sensitive fields and objects. Authorised server workflows can process data. The exact deployment, fields and key configuration must be reviewed separately.

Not in the current hybrid local setup: infrastructure runs in containers while the Go API and Vite app use local processes, with an HTTPS proxy in front. All required parts must be running. Production has its own deployment configuration.

Automatic OpenBao unsealing is still deferred. Protected operations may require manual unsealing after a restart. Process restart policies do not remove this dependency.

Off-server backups remain a TODO by the current delivery decision. Local snapshots and persistent container volumes are not substitutes for independently protected backups and tested restoration.

No certification, legal compliance or training accreditation is established by this documentation. Agree applicable obligations, evidence, controls and independent review for your deployment and course content.

The public site includes privacy, terms, cookies, acceptable-use and trust information. Legal documents remain review drafts until operator identity, jurisdiction, contacts and contractual details are confirmed.

No. This handbook describes the current local implementation snapshot and identifies setup and remaining work. Production may run an earlier release. Review the actual deployment and acceptance evidence before promising a feature to clients.

Keep unsaved text, inspect the error and check whether the change already committed. A 409 generally needs a refreshed revision and reconciliation. Do not blindly repeat writes or disable CSRF, permissions or encryption to make them succeed.

Examples include general document ingestion and learner AI tutoring, autonomous agent workflows, SCORM 2004/portable exports/hosted distribution, broad machine APIs, billing/provisioning UI, full custom-domain activation, automatic learning-event emails and independent recovery requirements. Consult the capability guide for the scoped snapshot.