html2mcq 2.0.0

Convert any HTML tutorial page, PDF, or image into MCQ questions using AI

html2mcq

Convert any HTML tutorial page, PDF, or image into MCQ questions using AI.

---

Install

pip install html2mcq

Everything is included — HTML extraction, PDF support, OCR, all AI providers.

---

Quick Start

from html2mcq import MCQGenerator

gen = MCQGenerator(
    api_key="sk-or-v1-...",
    provider="openrouter",
    mcq_model="google/gemini-2.5-flash-lite",
)

# From an HTML tutorial page
mcq = gen.from_url("https://docs.python.org/3/tutorial/", n=15)

# From a PDF URL
mcq = gen.from_pdf_urls("https://example.com/tutorial.pdf", n=10)

# From a local PDF file
mcq = gen.from_pdf_paths("/path/to/notes.pdf", n=10)

# From image files (OCR → MCQ in one step)
mcq = gen.from_image_paths("screenshot.png", n=5)

# From image URLs (vision model direct)
mcq = gen.from_image_urls("https://example.com/diagram.png", n=5)

print(mcq.to_json())
print(mcq.to_pretty_str())

---

What it supports

Input	Method
HTML tutorial page URL	from_url(url)
Raw HTML string	from_html(html)
PDF via URL	from_pdf_urls(url)
Local PDF file	from_pdf_paths(path)
Image files (local)	from_image_paths(path)
Image URLs	from_image_urls(url)
Pre-built blocks	from_blocks(blocks)

Also accepts list[str] for batch — from_pdf_urls([url1, url2]), from_image_paths([img1, img2]).

---

Constructor

gen = MCQGenerator(
    provider="openrouter",                         # default, also: "anthropic" | "openai" | "ollama"
    api_key="sk-or-v1-...",                        # or set env var OPENROUTER_API_KEY
    api_key_override="sk-or-v1-...",               # override key for this instance
    mcq_model="google/gemini-2.5-flash-lite",      # model for ALL MCQ generation
    mcq_model_list=["model1", "model2"],           # fallback list for mcq_model="auto"
    ocr_model="pytesseract",                       # OCR for HTML images: "pytesseract" | "auto" | model ID
    ocr_models=["model1", "pytesseract"],          # priority list for ocr_model="auto"
    method="twostep",                              # "twostep" (OCR→MCQ) | "images2mcq" (vision direct)
    save_ocr_path="ocr_output.txt",                # save OCR text when method=twostep
    prompt_log_path="stdout",                      # dump prompts: file path | "stdout" | "-"
    batch_size=10,
    max_tokens=4096,
    custom_instructions="Make answers tricky",
)

New in v2

Parameter	What it does
mcq_model	Single source of truth for all MCQ generation (text + vision). Renamed from model.
mcq_model="auto"	Tries mcq_model_list in order until one succeeds.
mcq_model_list	Priority-ordered models for auto mode. Runtime-reloadable via HTML2MCQ_MCQ_MODELS env var.
ocr_model	OCR backend: "pytesseract" (default), "auto" (priority list), or any OpenRouter model ID.
ocr_model="auto"	Tries ocr_models priority list.
ocr_models	Priority list for auto OCR. Reloadable via HTML2MCQ_OCR_MODELS env var.
method	"twostep" (OCR images → text → MCQs) or "images2mcq" (vision model direct).
prompt_log_path	Dump full prompts to file or terminal. Use "stdout" or "-" for terminal.
api_key_override	Override key for this instance
save_ocr_path	Save OCR text to file when method=twostep

---

Two-Step Image Pipeline (method="twostep")

When method="twostep" (default), from_image_paths() and from_image_urls() automatically:

1. OCR — extract text from images using ocr_model
2. Generate — feed text into text-based MCQ generation

Optionally save the OCR text with save_ocr_path:

gen = MCQGenerator(method="twostep", ocr_model="google/gemini-2.5-flash-lite",
                   save_ocr_path="ocr_output.txt")

# OCR → save to file → generate MCQs
gen.from_image_paths("chart.png", n=5)

---

Per-Call Overrides

All 7 public methods accept api_key_override and prompt_log_path:

mcq = gen.from_url(
    "https://example.com/",
    n=10,
    api_key_override="sk-or-v1-...",       # different key for this call
    prompt_log_path="debug_prompt.txt",    # log prompts for this call only
)

---

Output Schema

{
  "total_exam_time": 20,
  "questions": [
    {
      "question_html": "Which of these are non-mutating array methods?",
      "options": ["push()", "map()", "filter()", "pop()"],
      "answers": [1, 2],
      "multi": true,
      "marks": 1,
      "negative_marks": 0,
      "difficulty": "medium",
      "explaination": "map() and filter() return new arrays without modifying the original."
    }
  ]
}

---

AI Providers

# OpenRouter (default) — 100+ models
gen = MCQGenerator(api_key="sk-or-...", provider="openrouter",
                   mcq_model="google/gemini-2.5-flash-lite")

# Anthropic Claude
gen = MCQGenerator(api_key="sk-ant-...", provider="anthropic")

# OpenAI
gen = MCQGenerator(api_key="sk-...", provider="openai", mcq_model="gpt-4o")

# Ollama (local, no API key needed)
gen = MCQGenerator(provider="ollama", mcq_model="qwen2.5:7b",
                   ollama_base_url="http://localhost:11434/v1")

---

OCR Priority (when ocr_model="auto")

The default priority list is:

1. google/gemini-2.5-flash-lite (fast, cheap)
2. google/gemma-3-27b-it (free)
3. google/gemma-3-12b-it (free)
4. openai/gpt-4o (paid)
5. pytesseract (local fallback)

Override via ocr_models parameter or HTML2MCQ_OCR_MODELS env var.

---

MCQ Model Priority (when mcq_model="auto")

Tries mcq_model_list in order. Override via HTML2MCQ_MCQ_MODELS env var (comma-separated):

export HTML2MCQ_MCQ_MODELS="model1,model2,model3"

---

Custom Instructions

# Apply to every call
gen = MCQGenerator(
    provider="openrouter",
    custom_instructions="Make answers very close and confusing."
)

# Or per individual call
mcq = gen.from_url("https://example.com/", n=10,
    custom_instructions="All questions must be code-based.")

---

PDF Backends

# Default — PyMuPDF, fast, works for most digital PDFs
gen = MCQGenerator(provider="openrouter")

---

CLI

# Basic
html2mcq https://example.com/tutorial --n 20

# All options
html2mcq https://example.com/tutorial \
    --n 20 \
    --provider openrouter \
    --mcq-model google/gemini-2.5-flash-lite \
    --ocr-model auto \
    --difficulty "40% easy, 40% medium, 20% hard" \
    --topics variables functions \
    --output quiz.json \
    --format json

---

API Reference

MCQGenerator

Parameter	Default	Description
provider	"openrouter"	"openrouter"
api_key	None	API key or env var
api_key_override	None	Override key for this instance
mcq_model	""	Model for all MCQ generation. "auto" tries mcq_model_list
mcq_model_list	None	Fallback models for auto mode
ocr_model	"pytesseract"	OCR backend: "pytesseract"
ocr_models	None	Priority list for auto OCR
method	"twostep"	"twostep" (OCR→MCQ)
prompt_log_path	None	Dump prompts to file/terminal
batch_size	10	Questions per API call
custom_instructions	None	Global custom instructions

Method	Description
from_url(url, n, ...)	HTML page
from_html(html, n, ...)	Raw HTML string
from_pdf_urls(urls, n, ...)	PDF via URL (str or list)
from_pdf_paths(paths, n, ...)	Local PDF file (str or list)
from_image_urls(urls, n, ...)	Image URLs → MCQ via vision
from_image_paths(paths, n, ...)	Local image files → MCQ
from_blocks(blocks, n, ...)	Pre-extracted ContentBlock list

All methods accept api_key_override, prompt_log_path, difficulty_mix, focus_topics, custom_instructions.

MCQSet

Property / Method	Description
.questions	List[MCQQuestion]
.total_exam_time	Minutes — auto-calculated as n × 2
.to_json()	Exam-ready JSON
.to_pretty_str()	Human-readable output
.filter_by_difficulty(d)	Filter by "easy" / "medium" / "hard"

---

Project Structure

html2mcq/
├── html2mcq/
│   ├── __init__.py
│   ├── extractor.py        # HTML parser
│   ├── generator.py        # MCQGenerator — main API
│   ├── image_ocr.py        # OCR (pytesseract + vision API)
│   ├── models.py           # ContentBlock, MCQQuestion, MCQSet
│   ├── pdf.py              # PDF extractor
│   ├── prompts.py          # System + user prompt builders
│   └── cli.py              # CLI entry point
├── tests/
│   └── test_html2mcq.py    # 67 unit tests (fully mocked)
├── pyproject.toml
├── README.md
└── CHANGELOG.md

---

License

MIT © 2025 html2mcq contributors