Skip to content
Plain Docs

Architecture

Plain has two main pieces:

  • Plain, the native macOS SwiftUI app
  • PlainCore, the document pipeline and reusable policy code

Page Pipeline

Section titled “Page Pipeline”

The pipeline is:

  1. Normalize the address input.
  2. Reject unsupported or unsafe URLs.
  3. Fetch HTML with a privacy-oriented request policy.
  4. Sanitize active and unsafe markup.
  5. Extract a semantic DocumentModel.
  6. Optionally fetch non-SVG images through Plain’s image pipeline.
  7. Render the model with native SwiftUI/AppKit views.

Plain does not render pages through WKWebView.

URL Policy

Section titled “URL Policy”

URL normalization:

  • adds https:// for domain-like inputs
  • keeps supported relative URLs when there is a base URL
  • turns bare search text into a Mojeek query
  • strips common tracking parameters
  • rejects unsupported schemes

Safety validation blocks credential-bearing URLs, localhost, private ranges, reserved ranges, and local-domain targets.

Sanitization

Section titled “Sanitization”

Plain removes scripts, styles, forms, iframes, canvas, media embeds, hidden elements, unsafe URL attributes, event handlers, and obvious tracking pixels before building the document model.

Inert JSON-LD metadata may be preserved for extraction.

Rendering

Section titled “Rendering”

The app renders native document elements:

  • headings
  • paragraphs
  • inline links and emphasis
  • lists
  • block quotes
  • code blocks
  • simple tables
  • figures and captions
  • search result cards

The output intentionally does not preserve full CSS layout. Plain rebuilds readable content, not full pages.

Local App State

Section titled “Local App State”

Plain keeps:

  • recent pages
  • Later list, tags, and reading progress
  • Quotes, notes, and tags
  • reader settings
  • bounded image cache
  • DNS/IP resolution checks before page and image fetches

The image cache is capped at 50 MB and pruned after 30 days.

Plain News state is local too:

  • saved news sources
  • selected time window
  • interest text

Plain News Pipeline

Section titled “Plain News Pipeline”

Plain News is built on top of PlainCore.

The pipeline is:

  1. Load the user’s enabled sources.
  2. Fetch RSS feeds or source pages through Plain’s fetch policy.
  3. Parse feed items or extract candidate links from web sources.
  4. Apply the selected rolling or calendar time window.
  5. Deduplicate and shortlist articles locally.
  6. Assess selected articles with Apple Foundation Models when available, or a local heuristic fallback.
  7. Render the digest as native SwiftUI rows that open into Plain’s reader.

Plain News does not call remote AI services.