Exercise editor

The Exercise editor lets you configure interactive exercises for learners. You choose the exercise type, set the task instruction, and define questions and answers. The AI provides a draft; you can adjust questions, fix correct answers, and reorder items in the widget before saving.

How to invoke it

Ask the AI to create an exercise in natural language:

“Add a multiple choice quiz on photosynthesis”
“Create a true/false exercise on common grammar myths”
“Make a fill-in-the-blanks exercise on the water cycle”
“Add a crossword about chemistry vocabulary”

The AI calls edit_exercise with a pre-filled draft and opens the widget. It chooses the exercise type automatically based on your description, or you can name the type explicitly.

Exercise types

Type	Best for
`multiple_choice`	Questions with distinct right/wrong answers
`true_false`	Factual recall and misconception correction
`sort_list`	Sequences, steps, processes (vertical ordering)
`sort_words`	Sentence reconstruction, grammar (horizontal ordering)
`fill_blanks`	Reading comprehension, vocabulary in context (typed input)
`fill_blanks_options`	Same as fill_blanks but with dropdown menus
`drag_words`	Gap-fill with draggable word tokens
`crossword`	Vocabulary review in puzzle form

Multiple choice

Questions each with selectable answer options. At least one answer per question must be marked correct.

Top-level fields:

Field	Type	Notes
`type`	`"multiple_choice"`
`task`	string	Instruction shown above all questions
`questions`	array	One or more questions

Each question:

Field	Type	Notes
`task`	string	The question text
`multiple`	boolean	`true` if the learner must select all correct options simultaneously
`answers`	array	Answer options — 4 by default, at least one marked correct

Each answer:

Field	Type	Notes
`value`	string	The answer text
`correct`	boolean	Whether this answer is correct
`score`	integer	`1` if correct, `0` otherwise

True/false

Statements the learner classifies as true or false. Effective for factual recall and correcting common misconceptions.

Top-level fields:

Field	Type	Notes
`type`	`"true_false"`
`task`	string	Instruction shown above the statements
`questions`	array	At least 3 statements recommended

Each question:

Field	Type	Notes
`task`	string	The statement to classify
`answers`	array	Exactly 2 answers: one with `value: "true"` and one with `value: "false"`. Set `correct: true` on the right one.

Sort list

Items the learner drags into the correct vertical order. Use for sequences, chronological events, or procedural steps.

Top-level fields:

Field	Type	Notes
`type`	`"sort_list"`
`task`	string	Instruction shown above the list
`questions`	array	Exactly one question entry

The single question:

Field	Type	Notes
`task`	string	Optional label for the list
`answers`	array	Items listed in the correct order — the widget shuffles them for the learner

Each answer:

Field	Type	Notes
`value`	string	The item text
`priority`	integer	Correct position, 0-based — defines the correct order
`correct`	`false`	Always `false` — position, not this flag, defines correctness
`score`	`0`	Always `0`

Sort words

Words the learner rearranges horizontally to form a correct sentence or phrase. Use for language and grammar exercises.

Top-level fields: same structure as Sort list — type: "sort_words", task, exactly one questions entry with answers in the correct word order.

Each answer’s priority defines the correct word position; the widget displays words shuffled.

Fill blanks

A paragraph with gaps the learner fills by typing. Blanks are marked with [#N#] tokens in the body text.

Top-level fields:

Field	Type	Notes
`type`	`"fill_blanks"`
`task`	string	Optional instruction
`body`	string	HTML text containing blank markers: `[#1#]`, `[#2#]`, etc.
`questions`	array	One entry per blank, in the order they appear in `body`

Body format example:

The capital of [#1#] is [#2#], founded in [#3#].

Each question (one per blank):

Field	Type	Notes
`placeholder`	string	The marker this entry corresponds to, e.g. `"[#1#]"`
`answers`	array	First answer is the correct one (`score: 1`); remaining are distractors (`score: 0`). Include at least one distractor.

Fill blanks with options

Identical to Fill blanks but gaps are filled with a dropdown menu instead of typed input. Use when answer precision matters but typing would be disruptive.

Set type: "fill_blanks_options" — all other fields are the same as fill_blanks.

Drag words

Identical to Fill blanks but gaps are filled by dragging word tokens from a pool. Use for language practice where the physical act of selecting reinforces learning.

Set type: "drag_words" — all other fields are the same as fill_blanks.

Crossword

A classic crossword puzzle with clues and single-word answers. Use for vocabulary review.

Top-level fields:

Field	Type	Notes
`type`	`"crossword"`
`task`	string	Optional instruction
`questions`	array	Clue–answer pairs. 5–10 words recommended for a well-formed crossword.

Each question:

Field	Type	Notes
`task`	string	The crossword clue
`answers`	array	Exactly one answer

The single answer:

Field	Type	Notes
`value`	string	The crossword word — UPPERCASE letters only, no spaces or punctuation
`correct`	`true`	Always `true`
`score`	`1`	Always `1`

Tips

Name the type when you know it. “Create a drag-words exercise” is more precise than “create a fill-in exercise” — the AI will use the right type without guessing.
The AI handles answer generation. For most types, you only need to describe the topic and the AI will generate questions, correct answers, and distractors.
Crossword words must be single uppercase words. Multi-word answers and punctuation are not supported — rewrite clues so that the answer is always one word.
Fill-blanks blanks must match questions exactly. The [#N#] markers in body must correspond one-to-one with placeholder values in questions, in the same order.
Sort exercises define correct order via priority. Items with lower priority values come first; the widget shuffles the display automatically.