Nucleus OS Docs
    Create

    Importing sources

    Supported source formats and connections, how Nucleus OS uses your material, and tips for clean imports.

    Sources are the documents, decks, pages, and recordings a course is built from. They do two jobs: they give the authoring assistant something real to work from, and they ground the AI tutor so its answers cite your material instead of guessing.

    Supported formats

    TypeFormats
    DocumentsPDF, Word (.docx), Google Docs, Notion pages
    PresentationsPowerPoint (.pptx), Google Slides
    WebPublic web pages and wikis
    MediaAudio and video recordings (transcribed automatically)

    You can also start from a prompt with no source at all, which is useful for outlines and general skills, but grounded courses are more accurate and easier to keep up to date.

    Ways to add sources

    Attach files directly when you start a course. Uploaded files also appear under Files in the dashboard sidebar, so you can reuse them in other courses.

    How sources are used

    • Outline and drafts. The assistant proposes structure from your headings and generates lessons from the content.
    • Citations. Every generated claim links to the passage it came from, so reviewers can verify it quickly.
    • Tutor grounding. The AI tutor answers learners from the sources behind the courses they have access to.
    • Impact analysis. When a source changes, every course that cites it is flagged for update.

    Tips for clean imports

    • Prefer text over images of text. Scanned PDFs work, but text-based documents give better citations.
    • Remove what you do not want taught. Drafts, internal comments, and outdated appendices end up in the course if they are in the source.
    • Split very large documents into topic-sized files. A 300-page manual is easier to cite and update as chapters.
    • Use clear headings. They become the natural outline.

    Sensitive material

    Only upload material your organization is comfortable using for training. Content in your workspace is used to serve your workspace and is not used to train shared models. See Data and privacy.

    On this page