AI tutor and chat¶
Chat¶
/new-chat always opens a fresh conversation. Past conversations are reopened from Search Chats (/history), grouped by date and type and searchable by title. + New chat sits in the chat header.
Mode chips sit above the composer. No chip means plain chat. Which chips appear is decided by the server from the user's capabilities.
| Mode | What it makes | Free plan |
|---|---|---|
| Chat | Answers | Unlimited |
| Presentation | A slide deck | From the generation pool |
| Research | A cited report | From the generation pool |
| Exam prep | A mock test | From the generation pool |
| Image | An educational diagram or illustration | From the generation pool |
Every generated artifact can be previewed in full. Downloads (PPTX, PDF, DOCX, Markdown) and copying from an artifact preview need Pro. Copying a normal chat answer is always allowed.
Image generation¶
With the Image chip on, the tutor writes an image prompt from the conversation and a media model renders it (Imagen, then Gemini, then DALL-E as fallbacks). The image is stored and shown inline, and it stays in chat history and in Artifacts.
My Space¶
/myspace is the learner's personal knowledge base. Upload notes, a syllabus or past papers, or connect Google Drive, and the tutor grounds answers in them and cites them.
It uses the same pipeline as the institutional knowledge base: OCR, chunking, embedding. Every Learner account shares one "Individual" workspace, so isolation is by owner: a document is only ever retrieved for the user who uploaded it. Nothing is shared.
Artifacts¶
/artifacts lists everything the learner has generated: lesson plans, presentations, practice tests, images and more. Practice tests reopen into a question-by-question review and can be retaken. Viewing is never metered.
Question banks and research reports aren't shown to Learner accounts. Research stays on the chat surface.
Test Prep¶
The first time a learner picks Practice, a short wizard runs instead of an empty chat:
- What the practice loop is
- Pick a subject
- Optionally upload study material (up to 4 files, 10 MB each)
- Live generation progress
- The mock test, inside the chat
It shows once. After that, Practice goes straight to generating a test.
Where it lives¶
Learner code lives in MySetu-AI/kwiloai-b2c; paths below are relative to its apps/web/src and apps/backend/src.
- Web:
pages/shared/AITutorPage/,pages/shared/HistoryPage/,pages/shared/KnowledgeBasePage/(as My Space) - Backend:
api/v1/ai_tutor.py,services/chatbot_client.py,services/b2c_quota.py - LLM work runs in the chatbot microservice