What Vidext reads

Vidext builds courses from material, not from what a model already knows. Vidext's agents — the AI agents that plan and write your courses in the chat — read what your organization gives them, add public research where your material leaves gaps, and record where each part of a course comes from. This page explains each of those.

Your organization's sourcesCopy Anchor Link

Two kinds of material belong to your organization:

• Documents. Every file and YouTube video attached in a conversation in the chat: documents, slides, spreadsheets, images, audio, video and more. See Adding your sources for the formats.
• Your company website. When your organization was created from a business email address, or when you gave your company's website in its guided start, Vidext read that site's public pages. The agents can read those pages, and they also give your modules their colours. See Branding.

These sources belong to the organization, not to a course or a conversation. Anyone who builds courses in the chat can draw on them, in any conversation.

What Vidext makes from a sourceCopy Anchor Link

When a source arrives, Vidext processes it in the background:

• Text, extracted from documents, read from pictures and scanned pages, and transcribed from audio and video, with speakers told apart.
• What videos show, from watching them, and a summary of each recording.
• Images pulled out of documents.
• A search index, so the agents can find the passage that covers a topic instead of reading whole documents.

The original file is never changed. Everything made from it is kept separately.

ResearchCopy Anchor Link

While planning a course, the agents also research the topic on the public web and in research papers, to cover what your material leaves out. They can read a web page you name in the conversation. Research fills in public knowledge only: it never stands in for your organization's own facts, such as a policy, a figure or a contact. When the course needs one of those and no source has it, the agents ask you for the document. You can tell them to use only your material.

A template you start fromCopy Anchor Link

When you start a course from one of Home's templates, the agents read the template's module to keep how it teaches. The facts of its example company are never presented as yours: what the agents do not know about your company, they ask you for or leave as a visible gap. Its pictures are drawn again from your content in the template's look, unless you choose to build with the template's pictures. See Plan a course.

What the agents rememberCopy Anchor Link

The agents also keep a short memory for each person between conversations: how they like to build courses, and facts about the organization they confirmed that no source states. It is kept with your organization's course files; the agents load only your own memory into your conversations. See Working in the chat.

Where you can see what was usedCopy Anchor Link

• The course brief — the short summary of the course you approve at the start — records each decision and where it came from: your answer, or the source that settled it.
• The course structure says which chapters or pages of which source each module covers.
• Each part of a module's script — the words and screens of the module — names the source it draws on, or says it was written without one or follows a decision you made.
• The conversation's Sources list the files, organization documents and web pages the agents used and read.

That is what makes a course traceable back to its material. See Review and traceability.

The learner tutorCopy Anchor Link

When learners ask the tutor a question, it answers from the course content and says so when the material does not cover it. If a learner explicitly asks about something beyond the course, it can search the public web and labels that answer as web research. See AI tutor.

What Vidext does not readCopy Anchor Link

The agents do not read services connected in Settings → Connectors (Google Drive, Microsoft Teams, Notion) yet. See Connectors.