Examplary's grading API brings the AI grading teachers use in Examplary to your own platform. You send in your students' answers, and get back a score for each answer, broken down per rubric criterion, with the reasoning behind it and feedback for the student. Open questions and essays are graded by AI; questions with a single correct answer are graded automatically.
How it works
Everything is organised around an exam, which holds what the AI needs to grade well:
- Questions and rubrics. Every question has a rubric that describes what earns points. This is the one thing every exam needs.
- Source materials. Optionally, the course materials the exam is based on, so answers are graded against what students were actually taught.
- Calibration answers. Optionally, answers your teachers have graded before, to show the AI how strictly to apply the rubric.
You then import your students' answers as assessment sessions, one per student. Examplary grades them in the background, and you collect the results by polling the API or through a webhook.
The AI's grades come back as suggestions. You can use them as they are, have a teacher review them first, or let teachers do that in Examplary itself.
Getting started
The developer quickstart walks you through the whole flow with requests you can run straight from the docs: creating an exam, adding questions, importing answers and getting the results.
If you'd rather let teachers review and adjust grades without building that interface yourself, the grade answer flow embeds Examplary's grading screen in your platform.
Billing
Grading is billed per answer graded. Each graded answer uses 1 credit, plus 1 extra credit for every additional 250 words, or for every additional page of a scanned answer.
| Answer | Credits |
|---|---|
| Minimum per graded answer | 1 |
| Every additional 250 words of text in the answer | 1 |
| Every additional scanned or PDF page of the answer | 1 |