> ## Documentation Index
> Fetch the complete documentation index at: https://docs.northernlogic.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Source conversion

> Extract a source, optionally enhance it with a vault skill, and save the result.

You need write access to the destination vault and enough
[ingest credits](/administration/credits). Source conversion can send your
chosen source to extraction and AI providers. Only submit material you are
permitted to process this way.

## Convert a document or public URL

1. In **Knowledge**, choose **+ File → Convert**.
2. Choose a source type and provide a supported upload or public URL.
3. Select the destination and optionally enter a filename.
4. Select **Extract** and review **Preview**.
5. Optionally enable **AI enhancement**, select a skill from this vault, and run
   the enhancement step.
6. Save the preview as a Markdown file, then open the saved page to check it.

Supported source families include office documents, PDF, HTML, CSV, and public
webpages. Captioned YouTube URLs can be converted; a video without an available
transcript is refused. Audio and video file uploads are not supported.

## Use AI enhancement

Enhancement uses a Markdown file under the vault's `skills/` folder as
instructions. The separate Skills library is not automatically available here:
[copy a suitable skill into the vault](/library/skills) first.

Photos require a vision-capable enhancement step and a selected vault skill.
They do not use the document-extraction path. If the app says enhancement is not
configured, this needs product configuration rather than a different filename.

## Review before relying on the result

Extraction can lose layout, and AI rewriting can change meaning. Check names,
numbers, tables, quotations, and the source's qualifications. Keep the original
in your source system; the vault receives Markdown, not the uploaded binary.

Saved Markdown includes provenance such as source type and URL, plus skill/model
information when enhancement ran. After saving, normal editing, revisions,
search, export, and sync apply.

If extraction fails, confirm that the URL is publicly readable, the file is
supported, and credits remain. A new extraction or enhancement is a metered
action; avoid repeatedly rerunning it without checking the error.
