Skip to main content

​
What Question Studio does

Question Studio is the generation surface for admins and contributors. It lets you choose a semester, class, and module, pull source material from the Content Library, generate questions with a LearnTerms model, review the output, and bulk insert selected questions into the chosen module. It is not a free-form chatbot. It is a structured generation workflow.

​
Destination comes first

The first part of the screen asks you to choose where generated questions should go:
  1. Semester
  2. Class
  3. Module
This ordering matters because LearnTerms generation is not useful without a destination. The app is designed to place generated questions directly into a module, not leave them floating in a detached draft space.

​
Source material selection

Question Studio works with selected text. In practice, that usually comes from chunks in the Content Library. The built-in document browser lets you:
  • browse cohort-scoped documents
  • search by title
  • open a document
  • select chunk text to feed into generation
The selected text is then summarized into character and word counts on the generation panel.

​
Generation controls

Current generation controls include:
  • product model selection
  • question count of 5, 10, or 15
  • optional custom prompt
The product-facing model options currently are:
  • swift-general
  • swift-optometry
  • swift-pharmacy
All of them currently resolve to the same underlying model with different focus settings. The product docs should explain the product meaning, not just the backend mapping.

​
Source-text quality

The generation panel explicitly tracks character count and gives rough quality feedback based on the amount of selected text. Current behavior treats:
  • empty text as unusable
  • very short text as low quality
  • very large text as over the preferred cap
The current soft cap shown in the component is 3500 characters. This is worth documenting because a lot of “bad generation” is actually a source-material problem.

​
Reviewing generated output

After generation, the user can:
  • inspect the generated questions
  • select which ones to keep
  • remove individual questions
  • discard the whole batch
  • regenerate from the same source
  • add only selected questions to the destination module
This review step is important. Generated output is meant to accelerate authoring, not replace editorial judgment.

​
Limits and errors

Question generation can fail for several operational reasons:
  • no destination module is selected
  • no source material is selected
  • usage limits are reached
  • module limits are reached when saving
The UI distinguishes between ordinary errors and daily-limit style errors, which is useful to explain in contributor docs.
  1. Clean the source chunks first
  2. Choose the exact destination module
  3. Pick the model that matches the subject matter
  4. Generate a smaller batch first
  5. Select only the questions worth keeping
  6. Edit weak questions after insertion rather than accepting them blindly