Use Tutor1on1 by role.

Course creation with AI

Turn PDF, PPT, or teaching notes into a Tutor1on1 course ZIP.

Use an AI agent to reformat existing teaching materials outside the app and package them as a ZIP. Tutor1on1 then validates, imports, uploads, and lets you refine one KP at a time.

1. Put the source PDF, PPT, Word files, worksheets, and answer notes into a source folder. If the material is scanned, run OCR first and check formulas manually.
2. Ask Codex to propose the contents.txt outline first. Approve the KP order, title, and granularity before asking it to write files.
3. Ask Codex to create the course files and package them as a ZIP with contents.txt at the ZIP root. Every id in contents.txt must have a matching <id>_lecture.txt. Optional question pools use <id>_easy.txt, <id>_medium.txt, and <id>_hard.txt.
4. As a teacher in Tutor1on1, choose Import course ZIP and select the generated ZIP directly; do not unzip it. If validation reports a missing file or bad id, paste that error back into Codex and ask it to repair and repackage the ZIP.
5. After the ZIP imports, use right-click AI edit content on a KP for smaller lesson or question-bank edits.

special_relativity_course.zip
  contents.txt
  context.txt
  1_lecture.txt
  1.1_lecture.txt
  1.1_easy.txt
  1.1_medium.txt
  1.1_hard.txt

Codex web

1. Create a new task named Generate Tutor1on1 course ZIP from teaching materials.
2. Upload the PDF/PPT/Word files.
3. Prompt Codex: Read the uploaded materials and propose a Tutor1on1 contents.txt outline. Use dotted ids like 1, 1.1, 1.2. Do not generate files yet. Wait for my approval.
4. After approving the outline, prompt Codex: Create a complete Tutor1on1 course ZIP with contents.txt at the ZIP root, one <id>_lecture.txt for every id, and optional easy/medium/hard question files. Run a self-check and list missing files.
5. Download the generated ZIP, then select it directly with Import course ZIP in Tutor1on1; do not unzip it.

Codex desktop or local GUI

1. Create source_materials/ for the original documents and an empty output folder such as special_relativity_course/.
2. Open the parent workspace in Codex.
3. Prompt Codex: Use source_materials/ as input. Create a complete Tutor1on1 course folder in special_relativity_course/. Do not modify source_materials/. First generate contents.txt, then generate all lecture and question files. Check that every contents.txt id has <id>_lecture.txt, then package the contents of special_relativity_course/ as special_relativity_course.zip with contents.txt at the ZIP root.
4. Let Codex write the files and generated ZIP directly, and ask it for a final self-check report. In Tutor1on1, choose Import course ZIP and select special_relativity_course.zip without unzipping it.

AI edit course inside Tutor1on1

1. After the ZIP imports, open the course skill tree as a teacher.
2. Right-click the target KP node or its title in the list, then choose AI edit content.
3. Choose Content when you want lesson text, or Question when you want question-bank text for the selected KP.
4. Chat with the AI until the latest reply is ready to save. Keep each request focused on the current KP.
5. Click Edit on the AI reply to open the diff window. Review the current text on the left and edit the proposed text on the right.
6. Choose Add to append or Replace to overwrite. For questions, also choose easy, medium, or hard.
7. Save the change. Upload the course bundle again when you want marketplace students to receive the updated content.

Personalized learning loop

API config fields

• Provider chooses the model service, such as OpenAI or SiliconFlow.
• Base URL is the provider endpoint used by Tutor1on1. The app shows the URL for the selected provider.
• API Key is saved locally and used for tutor, TTS, and STT requests when those features need it.
• Test API key calls {baseUrl}/models, verifies the key, and loads the available text, TTS, and STT model lists. It does not download models or auto-select a new model.
• Text model powers the tutor conversation. Thinking effort appears only for providers that support reasoning controls.
• TTS model controls speech output. STT model controls speech-to-text input. They can stay empty if you do not use voice features.
• Saved API configs store the provider URL, selected models, thinking effort, and a short key hash so you can reuse known-good setups.

Use ChatGPT OAuth instead of an API key

1. Open Settings, API Settings, and choose OpenAI Codex (ChatGPT OAuth) as the Provider.
2. Enable ChatGPT device login in your personal ChatGPT Security settings, or ask your workspace administrator to enable it in workspace permissions.
3. Click Login with ChatGPT. Tutor1on1 first shows a one-time code; then click Open ChatGPT and enter that code at auth.openai.com/codex/device only when you started this login. If the browser window is blocked, copy the displayed verification URL and open it in Chrome yourself.
4. After OpenAI approves the device, Tutor1on1 detects completion and saves the login automatically; click Refresh models, choose an available text model, and save the API config.
5. Start or reopen a tutor session and choose the saved Codex OAuth model from the model picker.

Because Chrome cannot call the ChatGPT Codex backend directly under browser CORS, Tutor1on1 sends the OpenAI access token and account ID transiently with model-list and tutor prompt/response traffic through fixed routes at api.tutor1on1.org. The relay does not persist or log those credentials or content. TTS and STT still use API-key providers when voice is needed.

Common problems

• If login or course loading fails, check that the app can reach api.tutor1on1.org.
• If a student cannot enter a course, confirm that enrollment was approved and the bundle was downloaded.
• If model lists look stale, run Test API key again for the active provider and key.
• If progress looks stale across devices, wait for sync to finish or run a manual refresh.