promptukit 0.5.6

Utilities for building question banks and generating exam documents.

PromptuKit

Utilities for building and managing multiple-choice question banks and generating exam PDFs and slide decks.

Install from PyPI

pip install promptukit

The package is published at https://pypi.org/project/promptukit/. After installing you get the CLI entry points on your PATH:

add-question
extract-question --help
validate-question
question-bank --help
create-exam-md --help           # export question bank to editable Markdown (optional PDF via pandoc)
create-pptx --help             # generate a PPTX deck from a question bank
promptukit-gui                  # launch the browser-based authoring GUI
promptukit-claude-commands      # list/show/install bundled Claude Code slash commands (multi-subcommand CLI)
promptukit-claude-commands-install  # alias for `promptukit-claude-commands install`

You can also import the library in Python or a Jupyter notebook:

import promptukit as pk
# Top-level helpers: pk.load(path), pk.save(path, data), pk.pick(), pk.confirm()
# Launch the authoring GUI: pk.launch_gui()
# Subpackages are available as `pk.exams`, `pk.questions`, and `pk.utils`.

Quick Notebook Walkthrough

Here are short, copy-pasteable examples you can run inside a Jupyter notebook to load a question bank, validate it, and generate a PDF exam.

# 1) Import helpers
import promptukit as pk
from promptukit.questions import validate_question
from promptukit.exams import create_exam

# 2) Load a question bank (path relative to the repository root). If you're
# running outside the repository (for example from an installed package),
# fall back to the packaged sample dataset that ships with `promptukit`.
import os
bank_path = 'promptukit/data/question_banks/crb-water-management-sample.json'
if os.path.exists(bank_path):
  data = pk.load(bank_path)
else:
  # load packaged sample included with the installed package
  data = pk.load_resource('question_banks/crb-water-management-sample.json')

# 3) Inspect the file (section-based vs flat list)
if 'sections' in data:
   print('Sections:', [s.get('title') for s in data['sections']])
   print('First question:', data['sections'][0]['questions'][0])
elif 'questions' in data:
   print('Total questions:', len(data['questions']))
   print('First question:', data['questions'][0])
else:
   print('Unexpected file shape:', type(data))

# 4) Validate programmatically
errors, warnings = validate_question.validate(data)
if errors:
   print('Validation errors:', errors)
else:
   print('Bank valid — warnings:', warnings)

# 5) Generate a PDF exam from the same bank. We already have `data` loaded
# above as a dict, so pass it directly.
create_exam.build_exam_pdf(data, 'notebooks/output_exam.pdf')

Notes

• If you only want to run the library functions without Poetry activation, you can run modules with python -m promptukit.questions.extract_question or python -m promptukit.exams.create_exam as shown elsewhere in this README.
• PDF generation uses reportlab, which is installed automatically with promptukit.

Try the interactive Colab demo

If you'd like a runnable notebook that demonstrates the Quick Notebook Walkthrough, open the Colab demo:

https://colab.research.google.com/drive/1vzaUML_8nkWKhOfauv5MXPE-dQ5sXFF_?usp=sharing

Quick tips for Colab:

• To use the published package on PyPI:

!pip install promptukit

• To run the repository version (latest changes), clone and install from GitHub:

!git clone https://github.com/jrkasprzyk/promptukit.git
%cd promptukit
!pip install -e .

• The Colab notebook includes cells that use pk.load_resource(...) as a fallback when local content/ files aren't available.

Getting started (Poetry, for development)

1. Install Poetry (if you don't have it):

   pip install --user poetry
   

2. Use a compatible Python interpreter (>=3.12,<4.0) and create the virtual environment with dependencies:

   poetry install
   

3. Run the CLI tools via Poetry (console scripts / entry points):

   poetry run add-question
   poetry run extract-question --help
   poetry run validate-question
   poetry run question-bank --help
   poetry run create-pptx --help
   poetry run promptukit-gui --help
   poetry run promptukit-claude-commands --help
   

If you change dependencies or [tool.poetry.scripts] in pyproject.toml, run poetry install again so Poetry refreshes the local environment and console scripts.

Activating the virtualenv

If Poetry is configured to create an in-project virtualenv, it will be placed in a .venv folder at the repository root. Activate that environment from the project root using the command for your shell.

PowerShell (Windows):

.\.venv\Scripts\Activate.ps1

If script execution is blocked, temporarily allow it then activate:

Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass
.\.venv\Scripts\Activate.ps1

Command Prompt (cmd.exe):

.\.venv\Scripts\activate.bat

Git Bash / MSYS (Windows):

source .venv/Scripts/activate

macOS / Linux (POSIX):

source .venv/bin/activate

Alternatives (no manual activation required):

# Run a single command inside the virtualenv without activating it
poetry run <cmd>    # e.g. poetry run pytest

For development from a source checkout, prefer poetry run <cmd> unless you have activated the Poetry virtualenv. After activation, the bare console-script names work in that shell.

Poetry 2.x note:

• The poetry shell command (which previously spawned a new shell) is not installed by default in Poetry 2.0+. You can either use poetry env activate (then evaluate the printed activation command in your shell) or install the shell plugin to restore poetry shell.

Notes

• The package entry points are defined in pyproject.toml under [tool.poetry.scripts] and map console script names to the main() functions in the modules under the promptukit package.

Usage Examples

Quick (Poetry):

poetry run add-question path/to/mybank.json
poetry run extract-question --list-categories
poetry run validate-question
poetry run question-bank extract --help
poetry run create-pptx --help

Extracting data:

# List categories and available fields
poetry run extract-question --list-categories

# Print prompt and answer fields for the 'music' category
poetry run extract-question --file promptukit/data/question_banks/block-doku-sample.json --category music --fields prompt,answer

# Interactive picker
poetry run extract-question -i

Add questions:

# Interactive add; prompts for question type unless --type is provided
poetry run add-question path/to/mybank.json
poetry run add-question --type TrueFalse path/to/mybank.json

# Create a new bank while adding
poetry run add-question --create path/to/new-bank.json

# Batch mode
poetry run add-question --batch new_questions.json promptukit/data/question_banks/mybank.json

Validate a trivia file:

# Validate the default question bank
poetry run validate-question

# Validate a specific file
poetry run validate-question promptukit/data/question_banks/block-doku-sample.json

Manage files with question-bank (create/copy/extract):

# Create a new template JSON file
poetry run question-bank create --dest promptukit/data/question_banks/new.json --categories music,film-and-tv

# Copy an existing file
poetry run question-bank copy --src promptukit/data/question_banks/block-doku-sample.json --dest promptukit/data/question_banks/backup.json

# Extract a subset (easy music questions)
poetry run question-bank extract --src promptukit/data/question_banks/block-doku-sample.json --dest promptukit/data/question_banks/music_easy.json --categories music --difficulty easy

# Interactive extract
poetry run question-bank extract -i --src promptukit/data/question_banks/block-doku-sample.json --dest promptukit/data/question_banks/pick.json

Additional maintenance subcommands are available for schema and text cleanup:

poetry run question-bank migrate --help
poetry run question-bank audit-text --help
poetry run question-bank fix-text --help
poetry run question-bank render-audit --help

Alternative: run modules with python -m when not using Poetry:

python -m promptukit.questions.add_question path/to/mybank.json
python -m promptukit.questions.extract_question --help
python -m promptukit.questions.question_bank create --dest promptukit/data/question_banks/new.json

Authoring GUI (NiceGUI)

promptukit ships a lightweight browser GUI for authoring multiple-choice question banks without touching Python. It reads and writes the same JSON format used by the rest of the package (the one validated by validate-question), so you can open any existing bank in promptukit/data/question_banks/ (or your own) and edit it in place.

Launch from the shell after installing the package or activating the Poetry virtualenv:

promptukit-gui                                  # opens http://localhost:8080 in a browser tab
promptukit-gui -f my_bank.json                  # load (or create) this working file
promptukit-gui -p 9000 --no-browser             # custom port, don't auto-open a tab

From a source checkout without activation, prefix the same commands with poetry run, for example:

poetry run promptukit-gui -f my_bank.json

-f/--file points at the GUI's working file — if it exists it's loaded on startup; when you click Save all to file it gets overwritten with the current in-memory list. You can also change the working file from inside the GUI via the top-bar Open… button.

Or from Python:

from promptukit import launch_gui
launch_gui()                                     # defaults: ./promptukit_questions.json, port 8080
launch_gui(file_path="my_bank.json", port=9000, show=False)

The GUI is a single page with a resizable splitter: question list on the left, editor on the right. Each list row shows the full prompt (wrapping as needed) plus id, category, and a color-coded difficulty badge. The editor exposes every field in the schema:

• id (text)
• category (text with autocomplete from the file's categories list)
• difficulty (easy / medium / hard)
• prompt (autosizing textarea)
• choices (four inputs A–D, with a radio to pick which one is answer)
• quip_correct, quip_wrong (optional textareas — omitted from the saved file when blank, matching the existing banks' convention)

Top-bar buttons:

• Open… — switch the working file (loads it if it exists, or starts empty with that path queued for the next save).
• Reload from file — discard in-memory edits and re-read the working file.
• Save all to file — the only action that writes to the current working file, so you can discard a session.
• Save as… — write the in-memory bank to a new path and switch the working file to it (existing files are not overwritten unless you opt in).
• Copy all as JSON — full bank (including categories / _schema_notes).
• Copy selected as JSON — the single-question dict, ready to paste into another bank's questions array.

The editor's Apply button commits edits to the in-memory list (the top-bar Save writes them to disk).

On-disk format (unchanged from the rest of the package):

{
  "_schema_notes": ["optional free-form notes"],
  "categories": ["music", "motorsport"],
  "questions": [
    {
      "id": "music_001",
      "category": "music",
      "difficulty": "easy",
      "prompt": "Which instrument has a keyboard and strings?",
      "choices": ["Guitar", "Piano", "Violin", "Drums"],
      "answer": 1,
      "quip_correct": "Yep.",
      "quip_wrong": "Nope."
    }
  ]
}

Unknown top-level keys and unknown per-question keys are preserved verbatim on round-trip, so the GUI is safe to point at files with extra metadata it doesn't understand.

Requires nicegui (installed automatically as a dependency).

Create exam PDF

The create_exam.py script can generate a printable exam PDF. It accepts an external JSON question bank so you can build exams from your existing promptukit/data/question_banks/ files.

Usage (from the repository root with Poetry):

# Use the built-in hard-coded exam
poetry run python -m promptukit.exams.create_exam

# Load questions from a JSON bank and write a PDF
poetry run python -m promptukit.exams.create_exam -q promptukit/data/question_banks/block-doku-sample.json -o cven4333_from_json.pdf

# Save reproducible artifacts while creating the PDF
poetry run python -m promptukit.exams.create_exam \
  -q promptukit/data/question_banks/block-doku-sample.json \
  --save-questions exam_questions.json \
  --save-setup exam_setup.json \
  -o cven4333_from_json.pdf

You can also create the editable two-file artifact pair with the extraction tool, then render from those files:

poetry run question-bank extract \
  --src promptukit/data/question_banks/block-doku-sample.json \
  --dest exam_questions.json \
  --categories music \
  --setup-dest exam_setup.json \
  --artifact-kind exam \
  -f

poetry run python -m promptukit.exams.create_exam \
  -q exam_questions.json \
  -m exam_setup.json \
  -o exam.pdf

Supported JSON formats

• Top-level sections (preferred):

  {
     "sections": [
        {
           "title": "Section title",
           "questions": [ { "prompt": "...", "choices": ["...", "..."] }, ... ]
        }
     ]
  }
  

• categories is an alias for sections and is also accepted.

• Flat list of questions (top-level array) or top-level object with questions array:

  {
     "questions": [ { "prompt": "...", "choices": ["...", "..."], "category": "Section title" }, ... ]
  }
  

• Question objects support multiple common field names: prompt, q, question, or text for the question text; choices or answers for the answer options; optional category to group flat lists into sections.

• If choices are not already labeled (for example "Oceans" instead of "A) Oceans"), the script will prefix them with A), B), etc. Prompts without a leading number will be auto-numbered sequentially.

Create exam Markdown

The create_exam_md.py module exports a question bank to an editable Markdown file. Because the Markdown is the intermediate format, a professor can tweak questions, reorder sections, or adjust wording before producing the final PDF — without being constrained by what promptukit's PDF renderer directly supports.

Typical workflow:

JSON bank  →  Markdown (.md)  →  [edit]  →  PDF

Usage:

# Basic export
create-exam-md -q bank.json -o exam.md

# Include a numbered answer key at the end
create-exam-md -q bank.json -o exam.md --answers key

# Show answers inline after each question (instructor preview)
create-exam-md -q bank.json -o exam.md --answers inline

# Export Markdown and convert to PDF in one step (requires pandoc)
create-exam-md -q bank.json -o exam.md --to-pdf

# Specify a custom PDF output path
create-exam-md -q bank.json -o exam.md --to-pdf /path/to/final.pdf

# With a metadata/setup file
create-exam-md -q bank.json -m setup.json -o exam.md --to-pdf

# Using python -m (no Poetry activation required)
python -m promptukit.exams.create_exam_md -q bank.json -o exam.md --answers key

--to-pdf shells out to pandoc, which must be installed separately (pip install promptukit does not install pandoc). The .md file is always kept — it is the editable intermediate.

You can also call the public API directly:

from promptukit.exams.create_exam_md import build_exam_md

# Returns the Path to the written .md file
build_exam_md("bank.json", "exam.md", answers="key")

# With metadata dict
build_exam_md(data, "exam.md", metadata={"title": "Midterm 1"}, answers="none")

Supported answer modes:

--answers	Output
none (default)	No answers — student-facing exam
inline	Answer printed after each question
key	Numbered answer key appended at the end

All question types are supported: MultipleChoice, TrueFalse, ShortAnswer, FillInTheBlank, Matching, Calculation.

Student-facing Markdown uses scannable answer markers: open circles for choose-one questions and empty boxes for choose-multiple style records. The legacy default Multiple Choice Examination subtitle is not printed unless a metadata/setup file explicitly sets a different exam_type.

Create pub quiz PDF

The create_pub_quiz.py script generates a pub-quiz style group trivia PDF, with one printable sheet per round so a grader can split a stack of rounds and score them in parallel. Each sheet carries its own team-name / date / score header.

Usage (from the repository root with Poetry):

poetry run python -m promptukit.exams.create_pub_quiz \
  -q promptukit/data/question_banks/pub-quiz-sample.json \
  -o pub_quiz.pdf

# With custom metadata (title, host, instructions, labels)
poetry run python -m promptukit.exams.create_pub_quiz \
  -q my_quiz.json -m my_quiz_meta.json -o pub_quiz.pdf

# Extract an editable subset and setup file, then render from them
poetry run question-bank extract \
  --src promptukit/data/question_banks/pub-quiz-sample.json \
  --dest pub_quiz_questions.json \
  --setup-dest pub_quiz_setup.json \
  --artifact-kind pub_quiz \
  -f

poetry run python -m promptukit.exams.create_pub_quiz \
  -q pub_quiz_questions.json \
  -m pub_quiz_setup.json \
  -o pub_quiz.pdf

Input JSON layout — top-level rounds (aliases: sections, categories):

{
  "title": "JRB Industries Pub Quiz",
  "rounds": [
    {
      "title": "Motorsport",
      "theme": "Open wheel, closed wheel, and everything in between.",
      "questions": [
        {"prompt": "Which series uses the Dallara IR-18 as its spec chassis?"},
        {"prompt": "Which flag color signals an F1 race is stopped?",
         "choices": ["Yellow", "Blue", "White", "Red"]},
        {"prompt": "True or false: a full-course yellow bunches the field.",
         "question_type": "TrueFalse"}
      ]
    }
  ]
}

A flat questions list with a round or category key per item is also grouped into rounds. Questions without choices render as free-answer (blank line); questions with choices render as multiple-choice (team writes the letter); question_type: "TrueFalse" renders T / F.

See promptukit/data/question_banks/pub-quiz-sample.json for a 3-round example.

Create PPTX deck

The create-pptx console script generates a PowerPoint deck from a question bank. It accepts the same section-based and flat question layouts as the exam PDF generator.

Usage (from the repository root with Poetry):

# Generate one slide per question
poetry run create-pptx \
  -q promptukit/data/question_banks/block-doku-sample.json \
  -o questions.pptx

# Add answer reveal slides after each question
poetry run create-pptx \
  -q promptukit/data/question_banks/block-doku-sample.json \
  -o questions_with_answers.pptx \
  --answers after

# Save reproducible artifacts while creating the deck
poetry run create-pptx \
  -q promptukit/data/question_banks/block-doku-sample.json \
  --save-questions pptx_questions.json \
  --save-setup pptx_setup.json \
  -o questions.pptx

After installing the package or activating the Poetry virtualenv, the same commands can be run without the poetry run prefix.

Question types

Beyond multiple-choice (MultipleChoice), add-question (interactive or batch mode) and validate-question accept these non-MCQ types:

• TrueFalse - boolean answer
• ShortAnswer - free-text answer
• FillInTheBlank - prompt with [blank] placeholders + ordered answers
• Matching - ordered pairs as [left, right]
• Calculation - numeric answer with optional tolerance and unit

See promptukit/data/question_banks/mixed-types-sample.json for one of each. The OO model lives in promptukit.questions.question_models.

Bundled Claude Code slash commands

The package ships canonical add-trivia and audit-trivia slash-command prompts under promptukit.claude_commands. Use them via the promptukit-claude-commands CLI:

promptukit-claude-commands list                 # show available command names
promptukit-claude-commands show add-trivia      # print markdown to stdout
promptukit-claude-commands install              # copy into ./.claude/commands/
promptukit-claude-commands install --dest ~/.claude/commands  # user-level install

# Equivalent direct alias (no subcommand needed):
promptukit-claude-commands-install              # same as `... install`
promptukit-claude-commands-install --dest ~/.claude/commands

For local development, scripts/sync_claude_commands.py mirrors the same files into the repo's .claude/commands/ directory and supports --check for CI drift detection.

PATH note: pip install promptukit inside a Poetry-managed virtualenv puts the script on the venv's PATH only — the bare promptukit-claude-commands command will not resolve from a fresh shell. Pick one:

# 1. Prefix every call (no install changes)
poetry run promptukit-claude-commands install --dest ~/.claude/commands

# 2. Activate the venv first using one of the commands above
promptukit-claude-commands install --dest ~/.claude/commands

# 3. Install globally so the command is on PATH everywhere
pipx install promptukit            # recommended
pip install --user promptukit      # alternative

Example files

-- Example section-based bank: promptukit/data/question_banks/crb-water-management-sample.json -- Mixed question-type sample: promptukit/data/question_banks/mixed-types-sample.json -- JSON Schema describing accepted layouts: promptukit/data/question_banks/question_schema.json

Behavior notes

• If no -q/--questions file is provided to the exam generator, the script falls back to the built-in hard-coded 60-question exam and preserves its original 8-section breakdown.
• When you provide a section-based JSON file the PDF's section headings will be taken from each section's title (or name / label if present). When you provide a flat list with category fields, the loader will group questions by category to build sections automatically.

Running Tests

The test suite lives under dev/checks. The file dev/checks/test_question_tool.py contains unit tests that exercise the question-bank helpers and CLI-style interfaces.

Run the tests:

Using Poetry (recommended):

poetry install
poetry run pytest -q

Or run a single file directly:

poetry run pytest dev/checks/test_question_tool.py -q

Notes:

• Tests use pytest's tmp_path fixture and do not modify your repository files.