Fetch og:image from linked URLs when creating link posts

Add Pressa::OpenGraph, a best-effort scraper that pulls og:image (falling
back to twitter:image) from a linked page's HTML. bake new_link and bake
preview_link now use it to prefill the Image front-matter field for link
posts, unless the payload already supplies one. Network failures and
missing tags just resolve to nil so creating a post never blocks on a
slow or broken link; the fetch happens once at draft-creation time, not
on every build.

bake.rb

@@ -11,6 +11,7 @@ $LOAD_PATH.unshift(LIB_PATH) unless $LOAD_PATH.include?(LIB_PATH)
 
 require "pressa/drafts"
 require "pressa/link_post"
+require "pressa/open_graph"
 require "pressa/config/simple_toml"
 require "pressa/coverage"
 require "pressa/publish"
@@ -81,6 +82,7 @@ def new_link
     end
 
   author = payload["author"] || Pressa::Config::SimpleToml.load_file("site.toml")["author"]
+  image = payload["image"] || fetch_link_image(payload["link"])
   post =
     begin
       Pressa::LinkPost.build(
@@ -88,6 +90,7 @@ def new_link
         link: payload["link"],
         body: payload["body"],
         tags: payload["tags"],
+        image:,
         author:
       )
     rescue Pressa::LinkPost::Error => e
@@ -114,6 +117,7 @@ def preview_link
     end
 
   author = payload["author"] || Pressa::Config::SimpleToml.load_file("site.toml")["author"]
+  image = payload["image"] || fetch_link_image(payload["link"])
   post =
     begin
       Pressa::LinkPost.build(
@@ -121,6 +125,7 @@ def preview_link
         link: payload["link"],
         body: payload["body"],
         tags: payload["tags"],
+        image:,
         author:
       )
     rescue Pressa::LinkPost::Error => e
@@ -339,6 +344,14 @@ end
 
 private
 
+# Best-effort: a slow or broken link shouldn't block creating the post, it
+# just means the Image front-matter field is left for the author to fill in.
+def fetch_link_image(link)
+  return nil if link.to_s.strip.empty?
+
+  Pressa::OpenGraph.fetch(link)&.image
+end
+
 def run_test_suite(test_files)
   run_command("ruby", "-Ilib", "-Itest", "-e", "ARGV.each { |file| require File.expand_path(file) }", *test_files)
 end

lib/pressa/link_post.rb

@@ -9,7 +9,7 @@ module Pressa
 
     Result = Data.define(:filename, :target_path, :content)
 
-    def self.build(title:, link:, body: nil, tags: nil, author: Drafts.current_author, now: Time.now)
+    def self.build(title:, link:, body: nil, tags: nil, image: nil, author: Drafts.current_author, now: Time.now)
       title = title.to_s.strip
       raise Error, "title cannot be empty" if title.empty?
 
@@ -21,12 +21,12 @@ module Pressa
 
       filename = "#{slug}.md"
       target_path = "posts/#{now.strftime("%Y/%m")}/#{filename}"
-      content = render(title:, link:, body:, tags:, author:, now:)
+      content = render(title:, link:, body:, tags:, image:, author:, now:)
 
       Result.new(filename:, target_path:, content:)
     end
 
-    def self.render(title:, link:, body:, tags:, author:, now:)
+    def self.render(title:, link:, body:, tags:, image:, author:, now:)
       lines = [
         "---",
         "Title: #{yaml_quote(title)}",
@@ -37,6 +37,8 @@ module Pressa
       tag_list = normalize_tags(tags)
       lines << "Tags: #{tag_list.join(", ")}" unless tag_list.empty?
       lines << "Link: #{link}"
+      image = image.to_s.strip
+      lines << "Image: #{image}" unless image.empty?
       lines << "---"
 
       front_matter = lines.join("\n")

lib/pressa/open_graph.rb

@@ -0,0 +1,67 @@
+require "net/http"
+require "uri"
+
+module Pressa
+  # Best-effort scraper for OpenGraph metadata on a linked page, used to fill
+  # in an Image for link posts. Never raises: network failures, timeouts, and
+  # missing tags all just resolve to a nil image so post creation never blocks
+  # on a flaky or slow third-party site.
+  class OpenGraph
+    Result = Data.define(:image)
+
+    USER_AGENT = "samhuri.net-link-preview/1.0".freeze
+    MAX_REDIRECTS = 5
+
+    def self.fetch(url, http_get: method(:http_get))
+      html = http_get.call(url)
+      return nil if html.nil?
+
+      extract(html, base_url: url)
+    rescue
+      nil
+    end
+
+    def self.extract(html, base_url:)
+      image = meta_content(html, "og:image") || meta_content(html, "twitter:image")
+      return nil if image.nil?
+
+      Result.new(image: resolve(image, base_url:))
+    end
+
+    def self.meta_content(html, property)
+      pattern = /<meta\s+[^>]*(?:property|name)\s*=\s*["']#{Regexp.escape(property)}["'][^>]*>/i
+      tag = html[pattern]
+      return nil unless tag
+
+      content = tag[/content\s*=\s*["']([^"']*)["']/i, 1]
+      content&.strip&.then { |value| value.empty? ? nil : value }
+    end
+
+    def self.resolve(image, base_url:)
+      URI.join(base_url, image).to_s
+    rescue URI::InvalidURIError, URI::InvalidComponentError
+      image
+    end
+
+    def self.http_get(url, redirects_left: MAX_REDIRECTS)
+      return nil if redirects_left < 0
+
+      uri = URI.parse(url)
+      return nil unless uri.is_a?(URI::HTTP)
+
+      Net::HTTP.start(uri.host, uri.port, use_ssl: uri.scheme == "https", open_timeout: 5, read_timeout: 5) do |http|
+        response = http.get(uri.request_uri, "User-Agent" => USER_AGENT)
+
+        case response
+        when Net::HTTPSuccess
+          response.body
+        when Net::HTTPRedirection
+          location = response["location"]
+          return nil unless location
+
+          http_get(URI.join(url, location).to_s, redirects_left: redirects_left - 1)
+        end
+      end
+    end
+  end
+end

test/link_post_test.rb

@@ -71,4 +71,20 @@ class Pressa::LinkPostTest < Minitest::Test
     error = assert_raises(Pressa::LinkPost::Error) { build(link: "  ") }
     assert_match(/link/i, error.message)
   end
+
+  def test_image_is_included_in_front_matter_when_given
+    post = build(image: "https://example.net/preview.png")
+    meta = Pressa::Posts::PostMetadata.parse(post.content)
+    assert_equal("https://example.net/preview.png", meta.image)
+  end
+
+  def test_image_is_omitted_when_blank
+    post = build(image: "  ")
+    refute_includes(post.content, "Image:")
+  end
+
+  def test_image_is_omitted_when_not_given
+    post = build
+    refute_includes(post.content, "Image:")
+  end
 end

test/open_graph_test.rb

@@ -0,0 +1,61 @@
+require "test_helper"
+require "pressa/open_graph"
+
+class Pressa::OpenGraphTest < Minitest::Test
+  def test_extract_returns_og_image_resolved_against_base_url
+    html = <<~HTML
+      <html><head>
+        <meta property="og:image" content="/images/cover.png">
+      </head></html>
+    HTML
+
+    result = Pressa::OpenGraph.extract(html, base_url: "https://example.net/posts/cool-thing")
+    assert_equal("https://example.net/images/cover.png", result.image)
+  end
+
+  def test_extract_preserves_absolute_image_urls
+    html = %(<meta property="og:image" content="https://cdn.example.net/cover.png">)
+
+    result = Pressa::OpenGraph.extract(html, base_url: "https://example.net/posts/cool-thing")
+    assert_equal("https://cdn.example.net/cover.png", result.image)
+  end
+
+  def test_extract_falls_back_to_twitter_image
+    html = %(<meta name="twitter:image" content="https://cdn.example.net/tw.png">)
+
+    result = Pressa::OpenGraph.extract(html, base_url: "https://example.net/posts/cool-thing")
+    assert_equal("https://cdn.example.net/tw.png", result.image)
+  end
+
+  def test_extract_returns_nil_when_no_image_meta_present
+    html = "<html><head><title>No image here</title></head></html>"
+
+    refute(Pressa::OpenGraph.extract(html, base_url: "https://example.net"))
+  end
+
+  def test_extract_handles_single_quoted_attributes
+    html = %(<meta property='og:image' content='https://cdn.example.net/single.png'>)
+
+    result = Pressa::OpenGraph.extract(html, base_url: "https://example.net")
+    assert_equal("https://cdn.example.net/single.png", result.image)
+  end
+
+  def test_fetch_uses_injected_http_get_and_extracts_image
+    html = %(<meta property="og:image" content="https://cdn.example.net/x.png">)
+    result = Pressa::OpenGraph.fetch("https://example.net/post", http_get: ->(_url) { html })
+
+    assert_equal("https://cdn.example.net/x.png", result.image)
+  end
+
+  def test_fetch_returns_nil_when_http_get_returns_nil
+    result = Pressa::OpenGraph.fetch("https://example.net/post", http_get: ->(_url) {})
+    assert_nil(result)
+  end
+
+  def test_fetch_returns_nil_instead_of_raising_on_network_errors
+    failing_get = ->(_url) { raise Net::OpenTimeout, "timed out" }
+    result = Pressa::OpenGraph.fetch("https://example.net/post", http_get: failing_get)
+
+    assert_nil(result)
+  end
+end