From 54e9f712894f346ff4e5610c963bf40a9a901638 Mon Sep 17 00:00:00 2001
From: mkrastev <matey_krastev2@hotmail.com>
Date: Wed, 15 Jul 2026 13:18:03 +0300
Subject: [PATCH] add cps statement scraper

---
 datasets/amfv_datasets/scraping/__init__.py |  18 ++
 datasets/amfv_datasets/scraping/cli.py      |  38 ++-
 datasets/amfv_datasets/scraping/cps.py      | 285 ++++++++++++++++++++
 datasets/amfv_datasets/scraping/html.py     |   2 +
 datasets/test/test_scraping_cli.py          |  45 ++++
 datasets/test/test_scraping_cps.py          | 159 +++++++++++
 datasets/test/test_scraping_html.py         |   9 +
 7 files changed, 548 insertions(+), 8 deletions(-)
 create mode 100644 datasets/amfv_datasets/scraping/cps.py
 create mode 100644 datasets/test/test_scraping_cps.py

diff --git a/datasets/amfv_datasets/scraping/__init__.py b/datasets/amfv_datasets/scraping/__init__.py
index cd7cdb0..827edb3 100644
--- a/datasets/amfv_datasets/scraping/__init__.py
+++ b/datasets/amfv_datasets/scraping/__init__.py
@@ -9,6 +9,16 @@
     scrape_listing_documents,
 )
 from amfv_datasets.scraping.cli import OutputFormat, ScraperSource
+from amfv_datasets.scraping.cps import (
+    CpsFetchError,
+    CpsStatementRef,
+    build_statement_text,
+    list_statements,
+    listing_page_url,
+    scrape_cps,
+    scrape_statement,
+    statement_ref_from_url,
+)
 from amfv_datasets.scraping.html import (
     LinkMode,
     absolute_unique_urls,
@@ -31,6 +41,8 @@
 __all__ = [
     "GuidanceRef",
     "GuidanceListingPage",
+    "CpsFetchError",
+    "CpsStatementRef",
     "LinkMode",
     "NiceFetchError",
     "OutputFormat",
@@ -41,6 +53,7 @@
     "USER_AGENT",
     "absolute_unique_urls",
     "build_guideline_text",
+    "build_statement_text",
     "clean_text",
     "document_title",
     "default_client",
@@ -48,7 +61,12 @@
     "guidance_ref_from_url",
     "html_to_markdown",
     "list_published_guidance",
+    "list_statements",
+    "listing_page_url",
     "scrape_guideline",
     "scrape_listing_documents",
     "scrape_nice",
+    "scrape_cps",
+    "scrape_statement",
+    "statement_ref_from_url",
 ]
diff --git a/datasets/amfv_datasets/scraping/cli.py b/datasets/amfv_datasets/scraping/cli.py
index e8aece3..bb46d84 100644
--- a/datasets/amfv_datasets/scraping/cli.py
+++ b/datasets/amfv_datasets/scraping/cli.py
@@ -25,6 +25,7 @@
 )
 
 from amfv_datasets.scraping.base import ScrapedDocument, ScrapeRun
+from amfv_datasets.scraping.cps import scrape_cps
 from amfv_datasets.scraping.html import LinkMode
 from amfv_datasets.scraping.nice import scrape_nice
 
@@ -34,6 +35,7 @@ class ScraperSource(StrEnum):
 
     ALL = "all"
     NICE = "nice"
+    CPS = "cps"
 
 
 class OutputFormat(StrEnum):
@@ -68,13 +70,33 @@ def scrape_documents(
     if documents is not None and documents < 1:
         raise ValueError(f"documents must be at least 1; got {documents}")
 
-    for selected_source in _expand_source(source):
-        match selected_source:
-            case ScraperSource.NICE:
-                return scrape_nice(documents=documents, link_mode=link_mode, url=url)
-            case ScraperSource.ALL:
-                raise AssertionError("expanded source cannot be all")
-    raise AssertionError(f"unsupported source: {source}")
+    selected_sources = _expand_source(source)
+    if url is not None and len(selected_sources) != 1:
+        raise ValueError("--url requires one specific scraper source, not 'all'")
+    runs = tuple(
+        _scrape_source(selected_source, documents=documents, link_mode=link_mode, url=url)
+        for selected_source in selected_sources
+    )
+    if len(runs) == 1:
+        return runs[0]
+    total = sum(run.total for run in runs) if all(run.total is not None for run in runs) else None
+    return ScrapeRun(documents=(document for run in runs for document in run), total=total)
+
+
+def _scrape_source(
+    source: ScraperSource,
+    *,
+    documents: int | None,
+    link_mode: LinkMode,
+    url: str | None,
+) -> ScrapeRun:
+    match source:
+        case ScraperSource.NICE:
+            return scrape_nice(documents=documents, link_mode=link_mode, url=url)
+        case ScraperSource.CPS:
+            return scrape_cps(documents=documents, link_mode=link_mode, url=url)
+        case ScraperSource.ALL:
+            raise AssertionError("expanded source cannot be all")
 
 
 def write_jsonl(documents: Iterable[ScrapedDocument], output: TextIO) -> int:
@@ -125,7 +147,7 @@ def write_markdown_files(documents: Iterable[ScrapedDocument], output_path: Path
 
 def _expand_source(source: ScraperSource) -> tuple[ScraperSource, ...]:
     if source is ScraperSource.ALL:
-        return (ScraperSource.NICE,)
+        return (ScraperSource.NICE, ScraperSource.CPS)
     return (source,)
 
 
diff --git a/datasets/amfv_datasets/scraping/cps.py b/datasets/amfv_datasets/scraping/cps.py
new file mode 100644
index 0000000..4b9ba1e
--- /dev/null
+++ b/datasets/amfv_datasets/scraping/cps.py
@@ -0,0 +1,285 @@
+"""Scrape Canadian Paediatric Society statements into normalized markdown.
+
+CPS retains copyright in its position statements and practice points. This
+module supports a permission-gated ingestion workflow; obtain written CPS
+permission before running a corpus scrape or redistributing its output.
+"""
+
+from __future__ import annotations
+
+import re
+from collections.abc import Iterable
+from dataclasses import dataclass
+from urllib.parse import unquote, urljoin, urlparse
+
+import httpx
+from lxml import html as lxml_html
+from markdownify import MarkdownConverter
+
+from amfv_datasets.scraping.base import (
+    ScrapedDocument,
+    ScrapeError,
+    ScrapeRun,
+    default_client,
+    scrape_listing_documents,
+)
+from amfv_datasets.scraping.html import LinkMode, clean_text
+
+BASE_URL = "https://cps.ca"
+STATEMENTS_URL = f"{BASE_URL}/en/documents/statements-by-date"
+DOCUMENT_DELAY_SECONDS = 10.0
+LISTING_PAGE_SIZE = 10
+
+_POSITION_PATH_RE = re.compile(r"^/(?:en/)?documents/position/(?P<slug>[^/]+)/?$", re.IGNORECASE)
+_DATE_PATTERNS = {
+    "posted": re.compile(r"\bPosted:\s*([A-Za-z]+\s+\d{1,2},\s+\d{4})", re.IGNORECASE),
+    "reaffirmed": re.compile(r"\bReaffirmed:\s*([A-Za-z]+\s+\d{1,2},\s+\d{4})", re.IGNORECASE),
+    "updated": re.compile(r"\bUpdated:\s*([A-Za-z]+\s+\d{1,2},\s+\d{4})", re.IGNORECASE),
+}
+_BLANK_LINES_RE = re.compile(r"\n{3,}")
+_EMPTY_MARKDOWN_LINK_RE = re.compile(r"(?<!!)\[\]\([^)]*\)")
+_DECORATIVE_IMAGE_FILENAMES = {"file-pdf.svg"}
+
+
+class _CpsMarkdownConverter(MarkdownConverter):
+    """Preserve CPS superscripts that carry clinical meaning or citations."""
+
+    def convert_a(self, element, text: str, parent_tags: set[str]) -> str:  # noqa: ANN001
+        classes = element.get("class") or ()
+        href = element.get("href")
+        if "reference" in classes and href and "#ref" in href:
+            marker = element.get_text(strip=True).strip("[]")
+            if "a" in (self.options.get("strip") or ()):
+                return f"[{marker}]"
+            return f"[{marker}]({href})"
+        return super().convert_a(element, text, parent_tags)
+
+    def convert_sup(self, element, text: str, parent_tags: set[str]) -> str:  # noqa: ANN001
+        if not text.strip():
+            return ""
+        citation = element.find("a", href=lambda href: href and "#ref" in href)
+        if citation:
+            marker = element.get_text(strip=True).strip("[]")
+            if "a" in (self.options.get("strip") or ()):
+                return f"[{marker}]"
+            return f"[{marker}]({citation.get('href')})"
+        return f"<sup>{text}</sup>"
+
+
+_CPS_MARKDOWN_CONVERTERS = {
+    LinkMode.KEEP: _CpsMarkdownConverter(bullets="-", heading_style="ATX"),
+    LinkMode.STRIP: _CpsMarkdownConverter(bullets="-", heading_style="ATX", strip=("a",)),
+}
+
+
+class CpsFetchError(ScrapeError):
+    """Raised when a CPS statement cannot be discovered or parsed."""
+
+
+@dataclass(frozen=True)
+class CpsStatementRef:
+    """A CPS statement or practice point discovered from the date index."""
+
+    slug: str
+    title: str
+    page_url: str
+
+
+def statement_ref_from_url(url: str, *, title: str | None = None) -> CpsStatementRef:
+    """Parse a CPS position-statement URL into a canonical reference."""
+    parsed = urlparse(url.strip())
+    if parsed.scheme not in {"http", "https"} or parsed.netloc.lower() not in {"cps.ca", "www.cps.ca"}:
+        raise CpsFetchError(f"Enter a CPS statement URL from cps.ca; got {url!r}")
+
+    match = _POSITION_PATH_RE.match(unquote(parsed.path))
+    if not match:
+        raise CpsFetchError(f"Enter a URL like https://cps.ca/en/documents/position/example-statement; got {url!r}")
+    slug = match.group("slug")
+    return CpsStatementRef(
+        slug=slug,
+        title=title or slug.replace("-", " "),
+        page_url=f"{BASE_URL}/en/documents/position/{slug}",
+    )
+
+
+def listing_page_url(page: int) -> str:
+    """Return the CPS date-index URL for a one-based page number."""
+    if page < 1:
+        raise ValueError(f"page must be at least 1; got {page}")
+    if page == 1:
+        return STATEMENTS_URL
+    return f"{STATEMENTS_URL}/P{(page - 1) * LISTING_PAGE_SIZE}"
+
+
+def list_statements(client: httpx.Client, page: int) -> list[CpsStatementRef]:
+    """Return unique statement references from one CPS date-index page."""
+    response = client.get(listing_page_url(page))
+    response.raise_for_status()
+    doc = lxml_html.fromstring(response.text)
+    anchors = doc.xpath("//div[contains(concat(' ', normalize-space(@class), ' '), ' stmt-title ')]//a[@href]")
+
+    refs: list[CpsStatementRef] = []
+    seen_urls: set[str] = set()
+    for anchor in anchors:
+        href = anchor.get("href")
+        title = clean_text(anchor.text_content(), drop_numeric_citations=False)
+        if not href or not title:
+            continue
+        try:
+            ref = statement_ref_from_url(urljoin(BASE_URL, href), title=title)
+        except CpsFetchError:
+            continue
+        if ref.page_url in seen_urls:
+            continue
+        seen_urls.add(ref.page_url)
+        refs.append(ref)
+    return refs
+
+
+def _content_root(html_text: str) -> lxml_html.HtmlElement:
+    doc = lxml_html.fromstring(html_text)
+    selectors = (
+        "//div[contains(concat(' ', normalize-space(@class), ' '), ' statement-wrapper ')][1]",
+        "//main//article[1]",
+        "//*[@id='main-content']//article[1]",
+        "//*[@id='main-content'][1]",
+        "//main[1]",
+    )
+    for selector in selectors:
+        matches = doc.xpath(selector)
+        if matches:
+            return matches[0]
+    raise CpsFetchError("No readable statement content found on the CPS page")
+
+
+def _normalize_content(root: lxml_html.HtmlElement, *, base_url: str) -> None:
+    removable = root.xpath(
+        ".//script | .//style | .//nav | .//form | .//button | .//noscript | "
+        ".//*[contains(concat(' ', normalize-space(@class), ' '), ' breadcrumb ')] | "
+        ".//*[contains(@class, 'share')] | "
+        ".//*[contains(@class, 'social')] | "
+        ".//*[contains(concat(' ', normalize-space(@class), ' '), ' related-content ')] | "
+        ".//*[contains(concat(' ', normalize-space(@class), ' '), ' sidebar ')] | "
+        ".//*[contains(concat(' ', normalize-space(@class), ' '), ' hide-for-print ')] | "
+        ".//*[contains(concat(' ', normalize-space(@class), ' '), ' show-for-print ')] | "
+        ".//*[contains(@class, 'print:tw-hidden')] | "
+        ".//*[contains(concat(' ', normalize-space(@class), ' '), ' --podcast ')]"
+    )
+    for element in removable:
+        element.drop_tree()
+
+    for link in root.xpath(".//a[contains(@href, '/en/education/test-your-knowledge')]"):
+        link.drop_tree()
+
+    for image in root.xpath(".//img[@src]"):
+        filename = urlparse(image.get("src")).path.rsplit("/", 1)[-1].lower()
+        if filename in _DECORATIVE_IMAGE_FILENAMES:
+            image.drop_tree()
+
+    for link in root.xpath(".//a[@href]"):
+        link.set("href", urljoin(base_url, link.get("href")))
+    for image in root.xpath(".//img[@src]"):
+        image.set("src", urljoin(base_url, image.get("src")))
+
+    for nested in root.xpath(".//strong//strong | .//em//em"):
+        nested.drop_tag()
+    for title in root.xpath(".//h1"):
+        title.drop_tree()
+    for heading in root.xpath(".//*[self::h1 or self::h2 or self::h3 or self::h4 or self::h5 or self::h6]"):
+        for emphasis in heading.xpath(".//strong | .//em | .//b | .//i"):
+            emphasis.drop_tag()
+        if not clean_text(heading.text_content(), drop_numeric_citations=False) and not heading.xpath(".//img"):
+            heading.drop_tree()
+
+
+def _content_to_markdown(root: lxml_html.HtmlElement, *, link_mode: LinkMode) -> str:
+    source = lxml_html.tostring(root, encoding="unicode")
+    markdown = _CPS_MARKDOWN_CONVERTERS[link_mode].convert(source)
+    markdown = _EMPTY_MARKDOWN_LINK_RE.sub("", markdown)
+    lines = [line.rstrip() for line in markdown.splitlines()]
+    return _BLANK_LINES_RE.sub("\n\n", "\n".join(lines)).strip()
+
+
+def build_statement_text(
+    html_text: str,
+    *,
+    link_mode: LinkMode = LinkMode.KEEP,
+    base_url: str = BASE_URL,
+) -> tuple[str, int]:
+    """Extract one CPS statement as markdown and return its section count."""
+    root = _content_root(html_text)
+    _normalize_content(root, base_url=base_url)
+    section_count = max(1, len(root.xpath(".//*[self::h2 or self::h3 or self::h4 or self::h5 or self::h6]")))
+    content = _content_to_markdown(root, link_mode=link_mode)
+    if not content:
+        raise CpsFetchError("No readable statement content found on the CPS page")
+    return content, section_count
+
+
+def scrape_statement(
+    client: httpx.Client,
+    ref: CpsStatementRef,
+    *,
+    link_mode: LinkMode = LinkMode.KEEP,
+) -> ScrapedDocument:
+    """Scrape one CPS statement into the shared document schema."""
+    response = client.get(ref.page_url)
+    response.raise_for_status()
+    content, section_count = build_statement_text(response.text, link_mode=link_mode, base_url=ref.page_url)
+    raw_root = _content_root(response.text)
+    headings = [clean_text(value) for value in raw_root.xpath(".//h1[1]//text()")]
+    title = " ".join(value for value in headings if value) or ref.title
+    metadata: dict[str, str] = {"slug": ref.slug}
+    visible_text = clean_text(raw_root.text_content(), drop_numeric_citations=False)
+    for key, pattern in _DATE_PATTERNS.items():
+        if match := pattern.search(visible_text):
+            metadata[key] = match.group(1).strip()
+    return ScrapedDocument(
+        source="cps",
+        external_id=f"cps-{ref.slug.lower()}",
+        title=title,
+        url=ref.page_url,
+        content=content,
+        section_count=section_count,
+        metadata=metadata,
+    )
+
+
+def scrape_cps(
+    *,
+    documents: int | None,
+    link_mode: LinkMode = LinkMode.KEEP,
+    url: str | None = None,
+) -> ScrapeRun:
+    """Configure a CPS scrape from one URL or the current-statements index."""
+    if url is not None:
+
+        def scrape_url() -> Iterable[ScrapedDocument]:
+            with default_client() as client:
+                yield scrape_statement(client, statement_ref_from_url(url), link_mode=link_mode)
+
+        return ScrapeRun(documents=scrape_url(), total=1)
+
+    return ScrapeRun(
+        documents=scrape_listing_documents(
+            documents=documents,
+            client_factory=default_client,
+            list_page=list_statements,
+            scrape_item=lambda client, ref: scrape_statement(client, ref, link_mode=link_mode),
+            document_delay_seconds=DOCUMENT_DELAY_SECONDS,
+        )
+    )
+
+
+__all__ = [
+    "BASE_URL",
+    "CpsFetchError",
+    "CpsStatementRef",
+    "STATEMENTS_URL",
+    "build_statement_text",
+    "list_statements",
+    "listing_page_url",
+    "scrape_cps",
+    "scrape_statement",
+    "statement_ref_from_url",
+]
diff --git a/datasets/amfv_datasets/scraping/html.py b/datasets/amfv_datasets/scraping/html.py
index 2bd1c96..1580528 100644
--- a/datasets/amfv_datasets/scraping/html.py
+++ b/datasets/amfv_datasets/scraping/html.py
@@ -119,6 +119,8 @@ def _absolutize_links(html_text: str, *, base_url: str) -> str:
     root = lxml_html.fragment_fromstring(html_text, create_parent="div")
     for link in root.xpath(".//a[@href]"):
         link.set("href", urljoin(base_url, link.get("href")))
+    for image in root.xpath(".//img[@src]"):
+        image.set("src", urljoin(base_url, image.get("src")))
     return "".join(lxml_html.tostring(child, encoding="unicode") for child in root)
 
 
diff --git a/datasets/test/test_scraping_cli.py b/datasets/test/test_scraping_cli.py
index 44f714f..75239b6 100644
--- a/datasets/test/test_scraping_cli.py
+++ b/datasets/test/test_scraping_cli.py
@@ -11,6 +11,7 @@
 from amfv_datasets.scraping.cli import (
     ScraperSource,
     app,
+    scrape_documents,
     write_huggingface_dataset,
     write_jsonl,
     write_markdown_files,
@@ -219,6 +220,50 @@ def fake_scrape_documents(
     assert json.loads(result.stdout.splitlines()[0])["external_id"] == "nice-ng1"
 
 
+def test_scrape_documents_dispatches_cps(monkeypatch: pytest.MonkeyPatch) -> None:
+    """The CPS source delegates to its source-specific scraper."""
+    calls = 0
+
+    def fake_scrape_cps(*, documents: int | None, link_mode: LinkMode, url: str | None) -> ScrapeRun:
+        nonlocal calls
+        calls += 1
+        assert documents == 2
+        assert link_mode is LinkMode.KEEP
+        assert url == "https://cps.ca/en/documents/position/example"
+        return ScrapeRun([_document()], total=1)
+
+    monkeypatch.setattr("amfv_datasets.scraping.cli.scrape_cps", fake_scrape_cps)
+
+    run = scrape_documents(
+        ScraperSource.CPS,
+        documents=2,
+        link_mode=LinkMode.KEEP,
+        url="https://cps.ca/en/documents/position/example",
+    )
+
+    assert calls == 1
+    assert run.total == 1
+
+
+def test_scrape_documents_combines_all_sources(monkeypatch: pytest.MonkeyPatch) -> None:
+    """The all source streams NICE and CPS documents with a combined total."""
+    nice = _document()
+    cps = ScrapedDocument("cps", "cps-example", "CPS example", "https://cps.ca/example", "content")
+    monkeypatch.setattr(
+        "amfv_datasets.scraping.cli.scrape_nice",
+        lambda **_kwargs: ScrapeRun([nice], total=1),
+    )
+    monkeypatch.setattr(
+        "amfv_datasets.scraping.cli.scrape_cps",
+        lambda **_kwargs: ScrapeRun([cps], total=1),
+    )
+
+    run = scrape_documents(ScraperSource.ALL, documents=1, link_mode=LinkMode.KEEP)
+
+    assert run.total == 2
+    assert [document.source for document in run] == ["nice", "cps"]
+
+
 def _document() -> ScrapedDocument:
     return ScrapedDocument(
         source="nice",
diff --git a/datasets/test/test_scraping_cps.py b/datasets/test/test_scraping_cps.py
new file mode 100644
index 0000000..26d3bd9
--- /dev/null
+++ b/datasets/test/test_scraping_cps.py
@@ -0,0 +1,159 @@
+"""Tests for Canadian Paediatric Society statement scraping helpers."""
+
+import httpx
+import pytest
+
+from amfv_datasets.scraping.cps import (
+    BASE_URL,
+    STATEMENTS_URL,
+    CpsFetchError,
+    CpsStatementRef,
+    list_statements,
+    listing_page_url,
+    scrape_statement,
+    statement_ref_from_url,
+)
+
+
+def test_listing_page_url_uses_cps_offsets() -> None:
+    """One-based pages map to the CPS P30 offset convention."""
+    assert listing_page_url(1) == STATEMENTS_URL
+    assert listing_page_url(2) == f"{STATEMENTS_URL}/P10"
+    assert listing_page_url(4) == f"{STATEMENTS_URL}/P30"
+
+
+def test_list_statements_discovers_unique_position_pages() -> None:
+    """The index yields unique position statements and ignores unrelated links."""
+    html_text = """
+    <html><div class="cell main-body">
+      <div class="stmt-title"><a href="/en/documents/position/acute-asthma">Acute asthma</a></div>
+      <div class="stmt-title"><a href="https://cps.ca/documents/position/acute-asthma">Duplicate asthma</a></div>
+      <div class="stmt-title"><a href="/en/documents/position/newborn-glucose">Newborn glucose</a></div>
+      <a href="/en/documents/about-position-statements">About statements</a>
+    </div><footer><a href="/en/documents/position/footer-link">Footer</a></footer></html>
+    """
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        assert str(request.url) == STATEMENTS_URL
+        return httpx.Response(200, text=html_text)
+
+    with httpx.Client(transport=httpx.MockTransport(handler)) as client:
+        assert list_statements(client, 1) == [
+            CpsStatementRef("acute-asthma", "Acute asthma", f"{BASE_URL}/en/documents/position/acute-asthma"),
+            CpsStatementRef(
+                "newborn-glucose",
+                "Newborn glucose",
+                f"{BASE_URL}/en/documents/position/newborn-glucose",
+            ),
+        ]
+
+
+def test_statement_ref_from_url_normalizes_supported_routes() -> None:
+    """English and language-neutral CPS routes become canonical HTTPS URLs."""
+    assert statement_ref_from_url("http://www.cps.ca/documents/position/Acute-Asthma?print=1#dose") == (
+        CpsStatementRef(
+            "Acute-Asthma",
+            "Acute Asthma",
+            "https://cps.ca/en/documents/position/Acute-Asthma",
+        )
+    )
+
+
+def test_statement_ref_from_url_rejects_non_statement_pages() -> None:
+    """CPS navigation pages cannot be mistaken for clinical statements."""
+    with pytest.raises(CpsFetchError):
+        statement_ref_from_url("https://cps.ca/en/documents")
+
+
+def test_scrape_statement_preserves_clinical_structure_and_metadata() -> None:
+    """Statement content is normalized while navigation and sharing chrome are removed."""
+    html_text = """
+    <html>
+      <head><title>Fallback | Canadian Paediatric Society</title></head>
+      <body>
+        <header><h1>Site heading</h1></header>
+        <main id="main-content">
+          <div class="statement-wrapper">
+            <div class="breadcrumb">Home / Documents</div>
+            <p>Position statement</p>
+            <h1>Management of well-appearing febrile young infants</h1>
+            <p>Posted: Oct 27, 2023 | Reaffirmed: Jan 12, 2026 | Updated: May 27, 2026</p>
+            <h2>Recommendations</h2>
+            <ul><li>Assess the infant:<ul><li>Check vital signs.</li></ul></li></ul>
+            <p>Use the <a href="/en/tools/risk-calculator">risk calculator</a>.</p>
+            <p>Current citation<a class="reference" href="#ref1">[1]</a>.</p>
+            <p>Legacy citation<sup>[<a class="reference" href="#ref2">2</a>]</sup>.</p>
+            <p>Maternal fever above 38<sup>o</sup>C and counts of 10<sup>9</sup>/L require attention.</p>
+            <p><a href="/documents/full-statement.pdf">Full statement PDF</a>
+              <img src="/assets/img/file-pdf.svg" alt="PDF icon"></p>
+            <p><a href="/empty-target"></a></p>
+            <table><tr><th>Age</th><th>Action</th></tr><tr><td>0-28 days</td><td>Investigate</td></tr></table>
+            <h2><strong>Algorithm</strong></h2>
+            <p><img src="/uploads/flowcharts/febrile-infant.png" alt=""></p>
+            <h2>References</h2>
+            <div class="reference-print-style"><ol>
+              <li><a name="ref1"></a>Reference one.</li>
+              <li><a name="ref2"></a>Reference two.</li>
+            </ol></div>
+            <a href="/en/education/test-your-knowledge"><img src="/assets/img/test.png" alt="Quiz"></a>
+            <div class="share-tools">Share this statement</div>
+            <aside class="related-content">Related news</aside>
+            <script>window.track = true;</script>
+          </div>
+        </main>
+        <footer>Contact CPS</footer>
+      </body>
+    </html>
+    """
+    ref = CpsStatementRef(
+        "febrile-young-infants",
+        "Listing title",
+        f"{BASE_URL}/en/documents/position/febrile-young-infants",
+    )
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        assert str(request.url) == ref.page_url
+        return httpx.Response(200, text=html_text)
+
+    with httpx.Client(transport=httpx.MockTransport(handler)) as client:
+        document = scrape_statement(client, ref)
+
+    assert document.source == "cps"
+    assert document.external_id == "cps-febrile-young-infants"
+    assert document.title == "Management of well-appearing febrile young infants"
+    assert document.section_count == 3
+    assert document.metadata == {
+        "slug": "febrile-young-infants",
+        "posted": "Oct 27, 2023",
+        "reaffirmed": "Jan 12, 2026",
+        "updated": "May 27, 2026",
+    }
+    assert "- Assess the infant:\n  - Check vital signs." in document.content
+    assert "[risk calculator](https://cps.ca/en/tools/risk-calculator)" in document.content
+    assert (
+        "Current citation[1](https://cps.ca/en/documents/position/febrile-young-infants#ref1)."
+        in document.content
+    )
+    assert (
+        "Legacy citation[2](https://cps.ca/en/documents/position/febrile-young-infants#ref2)."
+        in document.content
+    )
+    assert "38<sup>o</sup>C" in document.content
+    assert "10<sup>9</sup>/L" in document.content
+    assert "# Management of well-appearing" not in document.content
+    assert "## Algorithm" in document.content
+    assert "## **Algorithm**" not in document.content
+    assert "| Age | Action |" in document.content
+    assert "![](https://cps.ca/uploads/flowcharts/febrile-infant.png)" in document.content
+    assert "## References" in document.content
+    assert "1. Reference one." in document.content
+    assert "2. Reference two." in document.content
+    assert "[Full statement PDF](https://cps.ca/documents/full-statement.pdf)" in document.content
+    assert "file-pdf.svg" not in document.content
+    assert "empty-target" not in document.content
+    assert "Share this statement" not in document.content
+    assert "Related news" not in document.content
+    assert "test-your-knowledge" not in document.content
+    assert "assets/img/test.png" not in document.content
+    assert "window.track" not in document.content
+    assert "Contact CPS" not in document.content
diff --git a/datasets/test/test_scraping_html.py b/datasets/test/test_scraping_html.py
index d34a8fa..6e82ae3 100644
--- a/datasets/test/test_scraping_html.py
+++ b/datasets/test/test_scraping_html.py
@@ -68,3 +68,12 @@ def test_html_to_markdown_can_strip_links() -> None:
     html_text = '<p>Offer <a href="https://example.org">treatment</a>.</p>'
 
     assert html_to_markdown(html_text, link_mode=LinkMode.STRIP) == "Offer treatment."
+
+
+def test_html_to_markdown_absolutizes_remote_image_references() -> None:
+    """Relative image references remain usable outside the source website."""
+    html_text = '<p><img src="/uploads/flowchart.png" alt="Treatment flowchart"></p>'
+
+    assert html_to_markdown(html_text, base_url="https://example.org") == (
+        "![Treatment flowchart](https://example.org/uploads/flowchart.png)"
+    )