AI Chat
Orbit Classroom provides an AI-powered chat experience embedded directly into lessons. Students interact with the AI to ask questions about course materials, and responses are grounded in uploaded content through the RAG pipeline.
Lesson-Bound Conversations
AI chat is only available within published lessons -- it is not a free-floating chatbot. This ensures every conversation has a clear instructional context and that the AI draws from the materials associated with that lesson.
- Students can create multiple chats per lesson to organize different topics or study sessions.
- Each chat receives an auto-generated title based on the initial exchange, making it easy to find past conversations.
Chat is unavailable on draft lessons. Students will only see the chat option once a teacher publishes the lesson.
Privacy
Student chats are private by design. Teachers and admins cannot read individual student conversations. This encourages honest, exploratory questioning without fear of judgment.
Streaming Responses
Responses are delivered token-by-token via Server-Sent Events (SSE), giving students immediate visual feedback as the AI generates its answer.
| Capability | Description |
|---|---|
| Stop | Halt generation mid-stream |
| Continue | Resume a previously stopped response |
| Regenerate | Discard the last response and generate a new one |
Answer Feedback
Students can like or dislike any AI response. This feedback signal can help instructors understand where the AI is performing well and where materials may need improvement.
Citations & Grounding
When the AI draws on uploaded class materials, its responses include inline citations that reference specific source chunks. Students can trace an answer back to the original document.
If the AI cannot ground its answer in class materials, a "Not grounded" warning is displayed. This transparency helps students distinguish between material-backed answers and general model knowledge.
Hidden System Prompts
Teachers can configure lesson-level system prompts that guide AI behavior without being visible to students. Common uses include:
- Steering the AI toward Socratic questioning instead of direct answers
- Restricting responses to specific topics covered in the lesson
- Setting tone or complexity level appropriate for the audience
Hidden prompts are configured per lesson in the lesson settings. Students are unaware of these instructions, but the AI follows them throughout every chat in that lesson.