Each tutor has a settings page where you can configure how it behaves, what tools it has access to, and how it interacts with students. Open any tutor and click Settings to access these options.
Name and prompt
The Name is what students see when they interact with the tutor. The Prompt is the system instruction that tells the tutor how to behave, what subject it covers, and what tone to use. A good prompt is specific about the subject, year level, and teaching approach.
AI model
Choose which AI model powers your tutor. Different models offer different trade-offs between speed and capability. The default model works well for most use cases.
Knowledge base
Upload documents to give your tutor custom knowledge beyond its training data. This is useful for course-specific material, textbook excerpts, or reference documents that students should be able to ask about.
Socratic mode
Controls how the tutor responds to student questions:
Off - The tutor answers questions directly
Guided - The tutor gives hints and asks follow-up questions
Adaptive - The tutor adjusts its approach based on how the student is progressing
Socratic - Full Socratic questioning, never giving direct answers
STEM mode
When enabled, the tutor gains access to computational tools for mathematics, physics, chemistry, and life sciences. It can solve equations, plot graphs, and perform calculations accurately rather than relying on the language model alone.
Suggested prompts
When enabled, the tutor suggests conversation starters and follow-up questions to help students engage. Suggestions appear as clickable chips in the chat interface.
Image generation
When enabled, the tutor can generate images inline during chat to illustrate concepts visually. Students can ask the tutor to draw, illustrate, or visualise topics and the tutor creates images directly in the conversation.
Testing your settings
After configuring your settings, switch to the Chat tab to test your tutor. Click Start Chat to begin a conversation and verify that Socratic mode, STEM tools, suggestions, and image generation all work as expected.