llm-quiz-engine: An Intelligent Quiz Auto-Generation System Based on Large Language Models

llm-quiz-engine is an open-source AI-driven automatic quiz generation system. It supports extracting topics from text/PDF and generating structured quizzes. It uses Pydantic for output validation to ensure format compliance, integrates the Groq API for efficient inference, and is suitable for multiple scenarios such as education and corporate training, which can significantly improve quiz generation efficiency.

In today's education and knowledge assessment field, generating high-quality quiz questions quickly and accurately is a challenge. Traditional manual question creation is time-consuming and labor-intensive, while template-based generation lacks diversity and quality. With the maturity of LLM technology, AI automatic quiz generation has become a potential solution. llm-quiz-engine is an end-to-end open-source solution that includes user interface, input processing, intelligent generation, result validation, and quiz management, suitable for teachers, enterprises, etc., to improve efficiency.

The project uses a Python tech stack with a clear and modular architecture:

• Streamlit for building the web interface, enabling rapid development of interactive applications with pure Python;
• Groq API as the inference backend, providing millisecond-level response speed;
• Pydantic to define strict data models (Quiz, Question, etc.) to ensure structured LLM output;
• PromptManager for centralized management of prompt templates, facilitating adjustment, optimization, and multi-language support.

llm-quiz-engine has rich features:

• Dual-mode input: Topic mode allows direct input of topics; Document mode supports uploading TXT/PDF to automatically extract topics and generate quizzes;
• Intelligent topic extraction: Extract core topics from the first 2000 characters of the document;
• Multi-level difficulty control (Easy/Medium/Hard);
• Complete quiz experience: Timing, pagination, answer recording, and score statistics;
• Quiz history: Supports saving and reviewing quiz performance.

Multiple mechanisms ensure output quality:

• Structured output constraints: Pydantic models validate LLM output, and errors are reported if it does not comply;
• Context length management: Document input is truncated to the first 10000 characters to avoid exceeding the LLM context window;
• Error classification and logging: A dedicated module handles LLM call errors and generates structured logs;
• Workflow and fallback mechanism: Degraded processing or retries when the main process fails to improve reliability.

The project is suitable for multiple scenarios:

• Education and training: Teachers generate after-class quizzes based on course materials;
• Corporate training: HR generates employee assessment questions based on internal documents;
• Knowledge self-test: Learners upload notes to generate quizzes to consolidate memory;
• Content creation: Bloggers generate interactive quizzes based on content to increase audience engagement.

Low deployment threshold:

1. Clone the repository and install dependencies: pip install -r requirements.txt;
2. Configure the Groq API key in the .env file;
3. Run streamlit run src/app.py to start the application, which is accessible via the default port 8502. The simple deployment is suitable for non-technical users, and the modular code is easy for developers to customize.

llm-quiz-engine demonstrates a typical application model of LLM in the edtech field. It transforms AI capabilities into product functions through reasonable architecture, strict validation, and complete interaction. Its open-source nature supports community contributions. In the future, it can expand to more document formats, LLM providers, and quiz types (fill-in-the-blank, short answer, etc.), serving as a reference implementation for LLM educational applications.