---
title: "Sources and documents"
description: "Adding a page for Blogged to read, uploading a private document, and the difference between the two."
updated: "2026-08-18"
verifiedAgainst:
  - packages/shared/src/knowledge.ts
---

A source is where a fact came from. There are exactly two kinds, and the
difference matters more than it looks.

<PropertyList>
  <Property name="web">
    A page anyone can open. A fact drawn from it can be cited in a published
    post, because a reader can follow the link.
  </Property>
  <Property name="upload">
    A document you handed over: a deck, a brief, an internal one pager. It has no
    public address, so a fact drawn from it is never emitted as a citation.
  </Property>
</PropertyList>

Records written before uploads existed are read as `web`.

## Adding a web source

Give the URL. Blogged reads the page, classifies what kind of page it is, and
extracts facts with the method the page supports: structured data if it is
there, the page's markup if not, and a model reading it as a last resort.

Reading is bounded. An initial ingest for an entity has a page budget, and a
refresh has a smaller one, so pointing Blogged at a large site does not turn
into an unbounded crawl.

## Uploading a document

Upload it in the Knowledge workspace or during onboarding. It becomes an
`upload` source. Facts extracted from it inform every draft and appear in none
of them as a link.

## Removing a source

Removing a source removes the facts that depend on it from future work. Posts
already published keep what they said.