Nomad changes

crates/kreuzberg/tests/pdf_element_classification.rs

@@ -0,0 +1,49 @@
+use kreuzberg::types::{ElementType, ResultFormat};
+use kreuzberg::{ExtractionConfig, OutputFormat, extract_bytes_sync};
+use std::path::Path;
+
+/// Verifies that numbered chapter headings in an untagged ReportLab PDF are
+/// classified as Heading/Title, not ListItem (#961).
+#[test]
+fn numbered_chapters_in_untagged_pdf_become_headings() {
+    let path = Path::new("test_documents/pdf/multipage_marketing.pdf");
+    if !path.exists() {
+        eprintln!("skipping: test_documents/pdf/multipage_marketing.pdf not found");
+        return;
+    }
+
+    let bytes = std::fs::read(path).expect("failed to read PDF");
+    let config = ExtractionConfig {
+        output_format: OutputFormat::Plain,
+        result_format: ResultFormat::ElementBased,
+        ..Default::default()
+    };
+
+    let result = extract_bytes_sync(&bytes, "application/pdf", &config).expect("extraction failed");
+    let elements = result.elements.unwrap_or_default();
+
+    let chapter_list_items: Vec<_> = elements
+        .iter()
+        .filter(|e| {
+            e.element_type == ElementType::ListItem && e.text.chars().next().is_some_and(|c| c.is_ascii_digit())
+        })
+        .collect();
+
+    let numbered_headings: Vec<_> = elements
+        .iter()
+        .filter(|e| {
+            matches!(e.element_type, ElementType::Heading | ElementType::Title)
+                && e.text.chars().next().is_some_and(|c| c.is_ascii_digit())
+        })
+        .collect();
+
+    assert!(
+        chapter_list_items.is_empty(),
+        "numbered chapter headings must not be ListItem; got: {:?}",
+        chapter_list_items.iter().map(|e| &e.text).collect::<Vec<_>>()
+    );
+    assert!(
+        !numbered_headings.is_empty(),
+        "at least one numbered chapter heading must be promoted to Heading/Title"
+    );
+}