From your assignment meta. A boolean flag like graded on each exercise tells the bot to enter no-solve mode. Ungraded practice exercises can be configured to allow worked examples. The same bot handles both, scoped per exercise rather than per course, so practice and assessment can sit side by side.

Technically nothing, and any honest answer must acknowledge this. The bot is a teaching tool, not a copy-paste prevention system. Where it earns its keep is with the students who actually want to learn but reach a hint moment. For them, the Socratic flow is more useful than typing into a public chatbot, because the hint is grounded in the lesson they just watched.

Yes. The system prompt is editable, and many instructors set different policies for different cohorts. Some courses allow worked examples at the end if the student demonstrates two correct intermediate steps. Others enforce strict no-solution. The bot logs the policy in effect for each conversation, so the audit trail is clean.

Yes. The Socratic mode works just as well in writing courses (questioning argument structure), language courses (questioning grammar choices), and humanities (asking for the source supporting a claim). The system prompt adapts to the subject by referencing the active lesson, so the bot's prompts sound like a TA in that subject rather than a generic Socratic shell.

Yes. Conversation logs are visible to instructors and can be exported for any integrity review. A student's reliance on the bot is explicit data, which is healthier than the current situation where students use public chatbots invisibly. Some courses now require the in-course bot to be the only approved AI tool, with use logged and acknowledged.

Not by default. The bot is a study companion, not a grader. Some courses use a separate SleekAI configuration with a grading rubric in the system prompt to draft initial feedback for the instructor to review and finalize. Keeping help and grading as separate bots avoids conflating the supportive tone with the evaluative tone.

Transcripts are stored in your own WordPress database, tied to the student user ID. No data is sent to a SleekAI server. The model API of your choice (OpenAI, Anthropic, Google, OpenRouter) sees only the message content per turn, and you can configure your account with that provider for zero retention if needed.

Yes. The default opening prompt invites the student to describe what they have tried and what they expected. From there the bot can identify the misconception driving the confusion and address it. The pattern of "I don't know what to ask" is itself useful data for the instructor, because it usually points at a lesson that did not land.