Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 11 additions & 1 deletion docs/help/course-prompts.md
Original file line number Diff line number Diff line change
Expand Up @@ -6,13 +6,23 @@ Course prompts let instructors customize reusable AI behavior for one course wit

Learning Objectives controls objective generation and is independent of teaching purpose. Quiz Blueprint and Question Generation can vary by Support Learning, Assess Understanding, or Gamify Learning. Coverage, question-history summary, and validation prompts have separate categories because they perform different tasks.

Quiz Blueprint decides what should be generated: the fixed coverage budget, question-type allocation, objective and subpoint coverage, Bloom levels, difficulty, and rationale. Question Generation writes the actual stem, answer, distractors, feedback, or other type-specific content for one approved Blueprint allocation. They can share the same teaching-purpose strategy, but their locked task responsibilities are different.

## Editable and locked instructions

The **Editable course instructions** box stores course-specific teaching context and reusable preferences. **Locked CREATE instructions** shows the workflow responsibility that CREATE always applies. Runtime guardrails summarize additional constraints such as output schemas, delivery compatibility, evidence retrieval, and question-history memory. Locked instructions are read-only so a course override cannot break saved-data contracts or make an unsupported package.

The panel shows the real course-level layer used by the workflow. Current objectives, evidence excerpts, Blueprint rows, history, counts, and output schemas are inserted dynamically when a request runs, so they are summarized rather than displayed as one stale prompt in settings.

## Course prompts versus one-time instructions

A course prompt defines repeatable behavior for future operations in that course. Instruction boxes in Learning Objectives, AI Blueprint, regeneration, and question generation apply only to the current request. Put stable terminology, audience assumptions, and course-wide constraints in a course prompt; put a one-off focus or exception in the local instruction box.

## Validate before saving

Validation checks required placeholders, length, grounding guidance, duplication guidance, and risky instructions. A warning does not always block saving, but missing required structure can prevent the prompt from supplying data the workflow needs.
Validation checks required placeholders, length, grounding guidance, duplication guidance, and risky instructions. It reviews the editable course layer together with the locked CREATE instructions and runtime context, so it should not report missing objectives, evidence, question schemas, or counts that CREATE injects automatically. An unchanged system default is recognized as valid without asking the AI reviewer to invent improvements.

When AI review identifies a meaningful, fixable issue, the validation result includes **Apply Changes**. This replaces only the text in the editable draft; it does not save or activate the prompt. Review the revised text, run **Validate** again, and then select **Save Course Prompt** if the result matches your intent.

Do not remove placeholders simply because their raw names look technical. They are replaced with the current objectives, materials, history, or requested total when an operation runs.

Expand Down
6 changes: 5 additions & 1 deletion docs/help/coverage-and-references.md
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,10 @@ Coverage tools show how course evidence flows into objectives and questions. The

Open the learning object's Coverage Map tab to inspect relationships among assigned materials, learning objectives, evidence, and generated questions. Use it after objectives are created and again after question generation.

In Graph view, drag nodes to rearrange the map, pan or zoom the canvas, select a node to inspect its properties and direct relationships, and double-click a node to isolate its one-hop neighborhood. Use Node Labels and Relationship Types to filter the map, search for a visible entity, and choose **Show complete map** to reset exploration. Evidence nodes with a resolved material can open the cited source.

The lines are generated from saved evidence and coverage relationships. CREATE does not let an instructor draw an arbitrary new line in this view because an unsaved line would falsely represent source support; update the objective, Blueprint, question, or source reference that owns the relationship instead.

Look for objectives with no evidence, objectives with no questions, heavily reused source passages, and clusters of questions that test the same subpoint. A sparse map can indicate missing coverage; an extremely dense area can indicate duplication.

## Learning-objective references
Expand All @@ -14,7 +18,7 @@ Objective references point to excerpts used during objective generation or enric

## Question evidence graph

In Review & Edit, open the evidence graph for one question. The graph connects the question to its learning objective, relevant subpoint, source material, and resolved evidence. Select an evidence node to inspect the source preview.
In Review & Edit, open the evidence graph for one question. The graph uses the same drag, zoom, filter, selection, and relationship-line interactions as the full Coverage Map, but starts with only that question's relevant neighborhood. Select an evidence node and choose **Open cited source** to inspect the source preview.

The evidence graph is question-specific, while Coverage Map summarizes the learning object. Use the graph for close review and the map for balance across the whole design.

Expand Down
20 changes: 18 additions & 2 deletions docs/help/learning-objectives.md
Original file line number Diff line number Diff line change
Expand Up @@ -8,10 +8,18 @@ Assign and finish processing the relevant materials first. Leave the instruction

Generation is a draft, not approval. Check every objective against the intended curriculum and the source material.

## Understand the generation log

The **Live generation log** explains what CREATE is doing before model text appears. It first resolves generation settings, reads the assigned material chunks, builds a source inventory, groups related sections into instructional clusters, and cleans document noise into a teaching-focused digest. The log reports safe aggregate results such as section, chunk, cluster, and topic counts; it does not display course source text or private model reasoning.

When drafting starts, the same log adds a **Live model draft** section. After the draft returns, CREATE checks whether the required source sections are covered and may run a targeted repair pass before saving the objectives. A long preparation stage does not by itself mean generation is stuck, but an error event in the log identifies the stage that needs attention.

## Add existing or manual objectives

Paste existing objectives into the generation instructions when you want the model to preserve or refine a provided set. Use **Add Manually** when wording must be stored exactly as entered. Manual entry is also useful when a required objective is not stated explicitly in the uploaded material.

When objectives already exist, select **Add New** to open the complete manual editor. Enter the objective text, optionally select Remember, Understand, Apply, Analyze, Evaluate, or Create, and add, edit, or remove custom subpoints before saving. The new objective is not written to the course until **Add Learning Objective** succeeds. Editing an existing objective opens the same fields.

An objective should use an observable verb and describe one coherent outcome. Avoid objectives that only say “understand” without showing what learners will demonstrate.

## Subpoints
Expand All @@ -26,7 +34,13 @@ Bloom level describes the cognitive action expected by the objective. Higher-ord

## Evidence and enrichment

Generated objectives can include source references and generation metadata. A manually added objective can be enriched from assigned materials; enrichment adds relevant subpoints and references without replacing useful existing metadata.
Generated objectives already include source references and generation metadata, so they do not show a separate Enrich action. A manually added objective can be enriched from assigned materials; enrichment adds relevant metadata and references without replacing an instructor-entered Bloom level or instructor-entered subpoints.

Use the **AI Enrich** action beside a manually created objective when only that LO should be enriched. Enrichment keeps the objective wording and manual Bloom/subpoints. Regeneration may rewrite the objective and then refreshes its AI-generated subpoints and source evidence to match the new wording. Hover or focus an icon to read its action label before selecting it. Processed, assigned materials are required for both actions.

After you manually add an objective, a short tutorial highlights the sparkle button on the first manually added LO. AI-generated or imported/classified objectives do not trigger this tutorial. Select **Enrich this LO** to run enrichment immediately, dismiss the tutorial to keep the objective unchanged, or replay the tutorial later from **User Account → Restart Feature Tutorials**.

Use **AI Link Missing** when an existing objective has no subpoints or source references. CREATE first asks the configured model for structured enrichment and then uses a source-grounded inventory fallback if the model is unavailable or returns incomplete JSON. If neither path can retrieve source evidence, the error identifies material processing or empty preview text as the next thing to repair; retrying a model connection alone is not sufficient in that case.

Open a reference to confirm the excerpt actually supports the objective. A citation proves where evidence came from, not that the objective is pedagogically correct.

Expand All @@ -41,4 +55,6 @@ Open a reference to confirm the excerpt actually supports the objective. A citat

## Editing and deleting objectives

Edit weak objectives before building the Blueprint. Deleting an objective that already has questions can also delete its dependent questions after confirmation. Review the warning carefully; this is a structural change, not only a text edit.
Edit weak objectives before building the Blueprint. If an objective already has linked questions, editing or regenerating it asks whether those questions should also be regenerated. Select **OK** to update the linked questions after the objective is saved, or **Cancel** to save the objective while keeping the existing questions unchanged. Review any regenerated questions and their evidence before export.

Deleting an objective that already has questions can also delete its dependent questions after confirmation. Review the warning carefully; this is a structural change, not only a text edit.
4 changes: 3 additions & 1 deletion docs/help/materials.md
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,9 @@ Choose the smallest relevant set. Unrelated, outdated, or duplicate materials ca

## Preview extracted content

Use the eye icon on a material card to inspect the content CREATE extracted. The preview is useful for confirming that headings, page text, and important passages are present before generation. A visually correct original file can still produce incomplete extracted text, especially when a PDF contains scanned images rather than selectable text.
Use the eye icon on a material card to inspect a processed source. PDF materials open in CREATE's existing PDF viewer, starting at page 1 with page navigation and an **Open full PDF** action. URL, DOC/DOCX, and pasted-text materials show the cleaned text CREATE extracted; URL previews also provide an **Open original webpage** action, and Word previews can download the original file.

The extracted-text preview is useful for confirming that headings and important passages are present before generation. A visually correct original file can still produce incomplete extracted text, especially when a PDF contains scanned images rather than selectable text.

## Page numbers and source locations

Expand Down
8 changes: 7 additions & 1 deletion docs/help/overview.md
Original file line number Diff line number Diff line change
Expand Up @@ -11,11 +11,15 @@ CREATE helps instructors turn course materials into evidence-grounded learning o
- **AI Blueprint:** an editable plan that allocates question types, counts, difficulty, Bloom level, focus areas, and rationale before generation.
- **Question:** a generated or manually created activity that can be reviewed, traced to evidence, and exported.

## Create and open quizzes

The course page lists saved learning objects in the **Quizzes** section using their editable names. Select **Add Quiz** to create another quiz; this button remains available when the course already has one or more quizzes. A newly created quiz appears in the course list and left sidebar immediately. Open a quiz to assign materials, define learning objectives, generate questions, review coverage, and export.

## Recommended end-to-end workflow

1. From the Dashboard, create a course for the subject or teaching context.
2. Open the course and add PDF or DOCX files, approved URLs, or pasted text.
3. Create a learning object, open it, and assign only the relevant course materials.
3. In Quizzes, select Add Quiz, open it, and assign only the relevant course materials.
4. Generate learning objectives from the materials, paste existing objectives, or add them manually.
5. Review the objectives and their subpoints, Bloom levels, and references before continuing.
6. Open Generate Questions, choose a delivery target, format, teaching purpose, and automatic or fixed question count.
Expand All @@ -36,6 +40,8 @@ Courses, learning objects, assigned materials, objectives, plans, and generated

CREATE Guide is the floating AI help chat available throughout the signed-in application. Ask about the page you are viewing, a workflow decision, question compatibility, exports, or troubleshooting. Its answers are grounded in this manual and include source cards. Selecting a source opens this Help Center and highlights the cited section.

On first use, a short tutorial highlights the round chat button in the bottom-right corner. Select **Try CREATE Guide** to open the chat immediately, dismiss the tutorial to keep the button closed, or select **Skip tutorials** to stop all quick-tour prompts. You can replay it later from **User Account → Restart Feature Tutorials**.

CREATE Guide is read-only. It cannot click controls, change a course, regenerate a question, delete data, or see information that is not included in its help sources and the current page context.

## If the guide and interface disagree
Expand Down
2 changes: 2 additions & 0 deletions docs/help/review-and-export.md
Original file line number Diff line number Diff line change
Expand Up @@ -38,6 +38,8 @@ H5P export creates a downloadable `.h5p` package. Column and Interactive Book su

PDF and Markdown exports can include questions only, answers only, or a combined version. Use questions-only for a learner handout and answers/combined for review or facilitation. These are snapshots: edits made after download require a new export.

PDF exports preserve the learner-facing content for every supported type. Mark the Words includes the statement without revealing the marked answers; Fill in the Blank uses printed blank lines instead of internal `$$` markers; Single Choice Set includes every subquestion and option; Documentation Tool includes its pages and response fields; Summary includes its knowledge points; and Essay includes the essay topic. Answer and combined exports add the corresponding answer or instructor-review guidance.

## Canvas export

Canvas export opens the Canvas connection and destination workflow. It requires a valid Canvas connection and appropriate permissions. Canvas LTI uses CREATE's player for Mixed Activity content; this is different from uploading a standard H5P package into Canvas.
Expand Down
4 changes: 4 additions & 0 deletions docs/help/troubleshooting.md
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,10 @@ Start with the smallest safe recovery step: preserve your work, read the visible

Complete the prerequisite in the previous tab. Assign processed materials before grounded objective generation, create at least one learning objective before the Blueprint, and generate at least one question before export. Refresh if the prerequisite was just saved but the tab has not updated.

## A search result is no longer available

Global Search only returns materials, questions, and learning objectives inside courses and quizzes owned by the signed-in instructor. CREATE verifies a result again when you select it. If its course, quiz, or material was deleted—or access changed after the search—the result is removed from the list instead of opening an invalid page. Search again or return to the course from the left sidebar. If an old browser tab already shows **Course not found** or **Quiz not found**, close that tab or navigate back; retrying generation on the invalid URL will continue to return 404 errors.

## AI generation does not start

Check that the account has a usable API key or permission to use the deployment key. Then confirm the learning object has assigned, processed materials where required and at least one learning objective. For a fixed Blueprint count, use a whole number between the number of objectives and 100.
Expand Down
Loading
Loading