If set, required to unlock the detailed results table at the end.
Auto-submits when time is up.
If disabled, the exported results screen ends without a retry button.
Example: export 40 questions, but show only 10 random questions on each page reset or retake.
AI Studio
Teacher-side AI only. The builder stores these settings locally in your browser so you can generate questions and improve wording without affecting student-facing files.
Ollama note: if you use the free local Ollama provider, open this builder through `http://127.0.0.1:5500` or another local web server, not directly as a `file:///...` page.
Free local endpoint. Make sure Ollama is running on your machine.
Default local model is llama3.2. For stronger results on capable hardware, try qwen2.5:14b or gpt-oss:20b.
No API key needed for local Ollama. Example: ollama pull llama3.2 or ollama pull qwen2.5:14b
Supported: PDF, TXT, MD.
No AI-generated questions yet.
Question Bank
Your questions live here. Add, edit, reorder, then Preview or Export.
ID
Topic / Info
Question (EN)
Answer / Hash
Actions
Add New Question
Fill in the fields below, then Save Question to add it to your Question Bank.
Sets the display order. Must be unique.
These tools use the current topic and question fields and write directly back into the editor.
Leave blank when editing an existing challenge to keep its hash.
Drop file here or click to browse
Any file type — images, PDFs, code files, etc.
⚠ This question already has an attachment. Upload a new file to replace it, or leave empty to keep the existing one.
Select the radio button next to the correct answer. The correct answer will be securely hashed.
Select the correct answer first, then AI will add plausible wrong options.
Code answers are saved for the teacher to grade manually — they are not auto-scored.
Manual = you grade later. Output = compare the program's printed output. Tests = run assertions for partial credit.
Runs AFTER the student's code. Use assert(condition, 'message'). Score = passed / total assertions. (JavaScript & Python.)
The student gets partial credit based on how close their code and its output are to this reference (offline: token, edit-distance & gzip similarity). Great for 'almost right' answers.