Setup Instructions:
1. Add the plugin to your Bubble app.

2. Go to the plugin settings and enter your OpenAI API key.

3. Set the Default LLM Model. Recommended default: gpt-4o-mini.

4. Use one of the available server-side actions:

- Generate Questions from Topic
- Generate Questions from Text
- Generate Questions from PDF URL

5. Set the action inputs, including number of questions, question types, number of choices, grade level/audience, difficulty, explanations, rubrics, and optional model override.

6. The plugin returns structured JSON and Bubble-friendly list outputs that can be displayed in repeating groups.

7. For best reliability, generate 10–20 questions at a time. The plugin supports up to 25 questions per generation.

8. For larger question sets, especially 21–25 questions, use a higher-quality OpenAI model such as gpt-4o.

9. For PDF URL generation, use a public, direct PDF URL. The PDF should be text-based. Scanned or image-only PDFs may not extract properly unless OCR is added separately.

Important Notes:

- This plugin requires an OpenAI API key.
- OpenAI usage is billed separately by OpenAI based on the selected model and usage volume.
- The plugin does not include OCR for scanned PDFs.
- Very large exams should be generated in multiple batches.