fix: guess HTML content starting with script tag (#1673)

ceberam · web-flow · commit 984cb137f6a8 · 2025-06-02T08:43:24.000+02:00
Signed-off-by: Cesar Berrospi Ramis &lt;75900930+ceberam@users.noreply.github.com&gt;
diff --git a/docling/datamodel/document.py b/docling/datamodel/document.py
@@ -412,7 +412,11 @@ def _detect_html_xhtml(
             else:
                 return "application/xml"
 
-        if re.match(r"<!doctype\s+html|<html|<head|<body", content_str):
+        if re.match(
+            r"(<script.*?>.*?</script>\s*)?(<!doctype\s+html|<html|<head|<body)",
+            content_str,
+            re.DOTALL,
+        ):
             return "text/html"
 
         p = re.compile(
diff --git a/tests/test_input_doc.py b/tests/test_input_doc.py
@@ -132,6 +132,13 @@ def test_guess_format(tmp_path):
     doc_path = Path("./tests/data/html/wiki_duck.html")
     assert dci._guess_format(doc_path) == InputFormat.HTML
 
+    html_str = (  # HTML starting with a script
+        "<script>\nconsole.log('foo');\n</script>"
+        '<!doctype html>\n<html lang="en-us class="no-js"></html>'
+    )
+    stream = DocumentStream(name="lorem_ipsum", stream=BytesIO(f"{html_str}".encode()))
+    assert dci._guess_format(stream) == InputFormat.HTML
+
     # Valid MD
     buf = BytesIO(Path("./tests/data/md/wiki.md").open("rb").read())
     stream = DocumentStream(name="wiki.md", stream=buf)
