QAGeneratorLLM

A Python package for generating educational questions and answers using various LLM providers.

Features

• Support for multiple LLM providers (Anthropic, Ollama, OpenAI, XAI)
• Generate both Multiple Choice Questions (MCQ) and Open-Ended Questions
• Choose between structured dataclass or raw JSON output formats
• Batch processing support
• File-based context input

Installation

pip install qageneratorllm

Usage

from qageneratorllm import QuestionGenerator, LLMProviderType, QuestionType, OutputType

# Initialize with default settings (Ollama + Open-Ended Questions)
generator = QuestionGenerator()

# Generate open-ended questions from text
result = generator.invoke("Your context text here")

# Generate Multiple Choice Questions using OpenAI
generator = QuestionGenerator(
    provider_type=LLMProviderType.OPENAI,
    question_type=QuestionType.MCQ
)
result = generator.invoke("Your context text here")

# Generate from file
result = generator.invoke_from_file("path/to/your/file.txt")

# Generate with JSON output instead of dataclass
generator = QuestionGenerator(
    question_type=QuestionType.MCQ,
    output_type=OutputType.JSON
)
result = generator.invoke("Your context text here")
# result will be a dictionary instead of a Pydantic model

Command Line Usage

# Basic usage
python -m qageneratorllm.generator --input input.txt --output questions.json

# Generate MCQs in batch mode from a folder of text files
python -m qageneratorllm.generator --input texts_folder/ --output results/ --batch --questions 5 --output-type json

Gradio Interface

The package includes a Gradio web interface for interactive question generation:

# Launch the Gradio app
python -m qageneratorllm.gradio_app

With the Gradio interface, you can:

• Upload documents (.txt, .md, .pdf)
• View document chunks separated by headers
• Filter chunks by header level
• Select specific chunks for question generation
• Choose question type (MCQ or open-ended)
• Select LLM provider
• Generate questions interactively

Environment Variables

• ANTHROPIC_MODEL_NAME: Anthropic model name (default: claude-3-sonnet-20240229)
• OLLAMA_MODEL_NAME: Ollama model name (default: qwen2.5)
• OPENAI_MODEL_NAME: OpenAI model name (default: gpt-4o)
• XAI_MODEL_NAME: XAI model name (default: grok-beta)

License

MIT