User guide

Moderator/Instructor

The moderator or instructor creates the session, directs the question-and-response cycle, communicates with participants, and decides what LLM-generated material is appropriate to share.

Before the session

Connect your account

Sign in with the account created by the server administrator. Connecting first opens the lobby so participants and enabled moderator assistants can connect and wait. Account and assistant management remain available on the Login tab; continue only when you are ready to configure the session.

Set up the session

Enter the course, section, and a unique session name. Choose where the moderator’s permanent local session record will be saved. Select whether participants may join without passwords, add an optional roster, and decide whether identifying-information review and anonymous response sharing should be enabled.

Configure the LLM

Select a provider and model, then use the selected model. Cloud providers require their own credentials; local Ollama keeps the inference process on the moderator’s computer. API keys are not saved in server preferences. Save preferences if you want your provider, model, appearance, privacy choices, and custom presets to follow your moderator account.

Conducting a live session

Continue to Live Session, then start the session. Write a question and send it to everyone. Responses are initially controlled separately so the moderator can decide when the floor opens, pause it, or stop it. A participant may revise a response; the latest-response view shows one current response per person, while the all-responses view shows the response history for that question.

Use a question that asks participants to explain their reasoning, propose an approach, or contribute an idea. “What is the definition?” produces less useful synthesis than “Explain why this principle applies here” or “What strategy would you try first, and what risk does it create?”

The moderator sees participant identities because responses must be routed, revised, and discussed. When anonymous sharing is enabled, participants see shared responses without participant names.

Messages and assistants

The Messages tab keeps private participant conversations separate from the question workspace. Select a participant before sending a private reply; public messages go to everyone. A raised-hand indicator signals a message needing attention. Moderator assistants can monitor questions, responses, and message history and can reply privately or publicly without gaining access to the moderator’s files, LLM configuration, or local history.

Synthesizing responses

Open the LLM Panel after responses arrive. Choose whether to use the current question or the entire session, then choose a built-in task: summarize responses, list points of agreement, or list different perspectives. The visible prompt can be edited before submission. A privacy-review screen shows the complete question-and-response material that will be sent; redact or revise it before continuing.

The LLM output is a draft. Review it for omissions, invented claims, misplaced consensus, or lost minority ideas. Edit the draft directly, then send it to participants only when it accurately represents the discussion. The edited version—not merely the original model output—is retained in the moderator’s local history.

Custom presets

The Presets tab lets you create reusable tasks. Give the preset a clear name, an optional moderator-facing prompt, and exact instructions for the LLM output. A preset can apply to every course or only the course specified in Session Setup. Save My Presets stores custom presets in your server preferences; Load My Presets refreshes the task list for the currently entered course.

Example: diagnosing understanding

An instructor asks, “Why can a profitable business still run out of cash?” Students respond independently. The instructor stops responses and asks the LLM to summarize. The draft shows that most students distinguish revenue from cash collection, but several overlook inventory purchases. The instructor edits the wording, shares the synthesis, and asks a follow-up focused on working capital. The value is not an automated grade; it is seeing the class’s current thinking soon enough to respond.

Saving, history, and ending

The moderator’s local record is updated during the session. Save when there is material worth preserving, finish the session when the activity is complete, and use History to preview or export completed sessions. Server recovery copies are encrypted and short-lived; they are not the permanent record. Quit disconnects participants while leaving the moderator free to review local history.