Lesson Summaries

What is a Summary?

A Summary is a short summary of a lesson, generated by a machine, that can be used to recap what happened during a lesson.

When does Lessonspace generate a lesson summary?

You can summarise a Session by:

selecting the option "Allow summary of sessions" under your Settings, or
via the Dashboard when you create a Space, you can toggle summaries on or off
by providing specific values via the Launch API, as we'll describe below.

Only eligible Sessions will be summarised. An eligible Session is any Session that:

lasts more than 300 seconds
has more than one user in it
and was also set to be transcribed

If a Session is not eligible, it will not be summarised, even if summarisation was requested.

Recording of AV must also be enabled in order for summaries to work. It can either be enabled in the same API call (if using the Launch API), toggled on when creating a Space on the dashboard, or flagged on by default for the whole organisation in your dashboard settings.

How long does Lessonspace keep a session summary?

Summaries are kept perpetually. They are not removed even if the original recording is removed.

Enabling Summaries

Automatic Summary

You can toggle summaries on the Addons section of your Settings dashboard to automatically summarise sessions.

Summary Override

You can override summarisation when creating a Space via the UI.

Launch API

Session summary is set for spaces made (or updated) via the Launch API by passing the summarise field as a boolean.

If the other values required for summarisation are incorrect (eg. if AV or transcription is turned off), the endpoint will return an error code and relevant error message.

Sample Request Body

{
  "id": "your-space-id",
  "record_av": true,
  "transcribe": true,
  "summarise": true,
  "user": {
    "name": "Teddy Transcriber"
  }
}

Example Response

{
  "client_url": "...",
  "api_base": "...",
  "room_id": "...",
  "secret": "..",
  "session_id": "..",
  "user_id": 3077485,
  "room_settings": {
    "record_av": true,
    "record_content": true,
    "waiting_room": false,
    "transcribe": true,
    "summarise": true
  }
}

You can use the room_settings object in the response to verify whether summarisation is enabled for a space.

Accessing Transcripts

You can access summaries in two ways, either via the Lessonspace Dashboard or via the Lessonspace API.

The Lessonspace Dashboard

Located on the Lessonspace dashboard, on the Sessions page, under the “Show Details” menu for each session.

The Lessonspace API

Summaries can be retrieved programmatically for an individual session by performing a GET request with the session UUID in the query parameter to the transcript endpoint. If no session was generated, the fields under the summary property will be null.

Webhooks

It is possible to subscribe to a webhook to be notified when a session’s summary is ready (and also to include a copy of the summary). You can read more about implementing webhooks generally in our documentation.

You define the webhook in the API call:

{
  "id": "your-space-id",
  "record_av": true,
  "transcribe": true,
  "summarise": true,
  "user": {
    "name": "Sally Summariser"
  },
  "webhooks": {
    "summary": {
      "finish": "https://your.url.here"
    }
  }
}

The webhook payload will be of the form:

{
  "room": { "id": string },
  "session": { "id": string },
  "summary": string,
  "nextSteps": []   // If no next steps were generated, this field will not be present
}

It is important to note that a session summary may potentially include private information.
As such, it is highly recommended to only use webhooks with the HTTPS scheme, to ensure they are not submitted in cleartext over the wire.

How does Lessonspace generate Lesson Summaries?

Our summaries are generated using an LLM (an "AI", colloquially). The LLM we use is provided by OpenAI. The LLM is provided with the full lesson transcript for each individual session as context, minus the metadata we add to the transcriptions ourselves (particularly names). This means the LLM is not explicitly given any personally-identifying information (PII) - however, if someone spoke a name or other PII during the lesson, that data would be included in the context.

In addition to the context, we provide a prompt that is designed to extract only educational content. The summary is limited to a few hundred words, and the LLM is explicitly instructed to avoid exposing PII, emotional judgments or evaluations, personal opinions, or off-topic information. Due to the nature of LLMs, this does not guarantee that the LLM will not still expose such data.

We explicitly do not permit OpenAI to use any of the data we submit to it to be used for training their models.

We use OpenAI's structured-data model to ensure a total separation between the LLM prompt and the LLM context, to prevent a user from deliberately or inadvertently manipulating the output by saying things that could be interpreted as commands to the LLM.

Overview

Guide

Resources