The knowledge layer
Documents, wikis, and branding — how your raw material becomes usable learning context.
Vidext's core claim is that generation is grounded in your knowledge. The knowledge layer is how that happens: everything you bring in is processed into a substrate that planning, building, and the learner tutor can reliably use.
Sources
Knowledge enters Vidext three ways:
- Uploads — PDFs, Word, PowerPoint, Excel, text, Markdown, CSV, RTF, and audio files. See Adding your sources.
- Website import — point Vidext at a public site; it maps the site, selects the most relevant pages, and imports them as sources. See Website import.
- Connected sources — search and read content from Google Drive, Microsoft Teams, and Notion. See Connected sources.
What processing produces
When a source arrives, Vidext processes it in the background:
- Extracted text — a clean, canonical version of the document's content.
- Extracted images — usable figures and photos pulled from the document, available to the builder as visuals.
- Summaries — for long documents, layered summaries that keep them usable in planning.
- Concept wikis — an index of the key concepts in a module's sources, compiled per module and rolled up per organization. Planning, generation, and the in-module tutor read these instead of re-reading raw files.
Source files themselves are immutable: what you uploaded is what is stored. Everything derived from them can be rebuilt.
Branding
Vidext can extract your visual identity from your website — your palette and the art direction generated images should follow — so the images Vidext generates for your modules look like your company. See Branding your modules.
Why grounding matters
- The planner proposes outlines based on what your sources actually say.
- The builder reuses images you already have — from your documents, your organization's media, and your researched sources — and generates the rest on-brand.
- The learner tutor answers from module content and its concept wiki — and says so when the material does not cover a question, rather than guessing.
That grounding is also what makes content traceable: generated learning can be tracked back to the sources that informed it. See Review and traceability.
Last updated on