feat(extractor): impl HwpxExtractor for application/haansofthwpx

.gitattributes

@@ -0,0 +1,22 @@
+# Alef-generated binding and e2e files — collapsed in GitHub PR diffs
+packages/csharp/**        linguist-generated=true
+packages/dart/**          linguist-generated=true
+packages/elixir/**        linguist-generated=true
+packages/gleam/**         linguist-generated=true
+packages/go/**            linguist-generated=true
+packages/java/**          linguist-generated=true
+packages/kotlin/**        linguist-generated=true
+packages/php/**           linguist-generated=true
+packages/python/**        linguist-generated=true
+packages/r/**             linguist-generated=true
+packages/ruby/**          linguist-generated=true
+packages/swift/**         linguist-generated=true
+packages/typescript/**    linguist-generated=true
+packages/wasm/**          linguist-generated=true
+packages/zig/**           linguist-generated=true
+crates/kreuzberg-ffi/**   linguist-generated=true
+crates/kreuzberg-node/**  linguist-generated=true
+crates/kreuzberg-php/**   linguist-generated=true
+crates/kreuzberg-py/**    linguist-generated=true
+crates/kreuzberg-wasm/**  linguist-generated=true
+e2e/**                    linguist-generated=true

alef.toml

@@ -560,6 +560,7 @@ types = [
   "GzipExtractor",
   "HtmlExtractor",
   "HwpExtractor",
+  "HwpxExtractor",
   "ImageExtractor",
   "JatsExtractor",
   "JupyterExtractor",

crates/kreuzberg/Cargo.toml

@@ -67,6 +67,7 @@ office = [
     "html",
 ]
 hwp = ["dep:cfb", "dep:flate2"]
+hwpx = ["dep:unhwp"]
 iwork = ["dep:zip", "dep:snap"]
 email = ["dep:mail-parser", "dep:cfb", "dep:outlook-pst", "dep:tempfile", "dep:chrono"]
 html = ["dep:html-to-markdown-rs", "dep:v_htmlescape"]
@@ -208,6 +209,7 @@ formats = [
     "excel",
     "office",
     "hwp",
+    "hwpx",
     "iwork",
     "email",
     "html",
@@ -369,6 +371,7 @@ tower-http = { version = "0.6", features = [
 ], optional = true }
 tracing = { workspace = true }
 tracing-opentelemetry = { version = "0.32", optional = true }
+unhwp = { version = "0.2.4", default-features = false, features = ["hwpx"], optional = true }
 unicode-normalization = { version = "0.1.25", optional = true }
 urlencoding = "2"
 utoipa = { version = "5.4", features = ["axum_extras"], optional = true }

crates/kreuzberg/src/core/mime.rs

@@ -41,6 +41,7 @@ pub(crate) const SOURCE_CODE_MIME_TYPE: &str = "text/x-source-code";
 
 pub(crate) const EXCEL_MIME_TYPE: &str = "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet";
 
+pub(crate) const HWPX_MIME_TYPE: &str = "application/haansofthwpx";
 pub(crate) const IWORK_PAGES_MIME_TYPE: &str = "application/x-iwork-pages-sffpages";
 pub(crate) const IWORK_NUMBERS_MIME_TYPE: &str = "application/x-iwork-numbers-sffnumbers";
 pub(crate) const IWORK_KEYNOTE_MIME_TYPE: &str = "application/x-iwork-keynote-sffkey";
@@ -783,6 +784,12 @@ fn detect_office_format_from_zip(content: &[u8]) -> Option<&'static str> {
     const NUMBERS_MARKER: &[u8] = b"Index/CalculationEngine.iwa";
     const KEYNOTE_MARKER: &[u8] = b"Index/Presentation.iwa";
 
+    // HWPX: ZIP of OWPML XML, contains Contents/content.hpf manifest
+    const HWPX_MARKER: &[u8] = b"Contents/content.hpf";
+    if contains_subsequence(content, HWPX_MARKER) {
+        return Some(HWPX_MIME_TYPE);
+    }
+
     // Check iWork first (before generic Office) since iWork ZIPs also contain XML
     if contains_subsequence(content, PAGES_MARKER) {
         return Some(IWORK_PAGES_MIME_TYPE);

crates/kreuzberg/src/extractors/hwp.rs

@@ -88,7 +88,7 @@ impl DocumentExtractor for HwpExtractor {
     }
 
     fn supported_mime_types(&self) -> &[&str] {
-        &["application/x-hwp", "application/haansofthwpx"]
+        &["application/x-hwp"]
     }
 
     fn priority(&self) -> i32 {
@@ -106,10 +106,7 @@ mod tests {
         assert_eq!(extractor.name(), "hwp-extractor");
         assert_eq!(extractor.version(), env!("CARGO_PKG_VERSION"));
         assert_eq!(extractor.priority(), 50);
-        assert_eq!(
-            extractor.supported_mime_types(),
-            &["application/x-hwp", "application/haansofthwpx"]
-        );
+        assert_eq!(extractor.supported_mime_types(), &["application/x-hwp"]);
     }
 
     #[test]
@@ -118,4 +115,20 @@ mod tests {
         assert!(extractor.initialize().is_ok());
         assert!(extractor.shutdown().is_ok());
     }
+
+    #[test]
+    fn test_hwpx_mime_not_routed_to_hwp_extractor() {
+        use crate::KreuzbergError;
+        use crate::plugins::registry::DocumentExtractorRegistry;
+        use std::sync::Arc;
+
+        let mut registry = DocumentExtractorRegistry::new();
+        registry.register(Arc::new(HwpExtractor::new())).unwrap();
+
+        let result = registry.get("application/haansofthwpx");
+        assert!(
+            matches!(result, Err(KreuzbergError::UnsupportedFormat(_))),
+            "application/haansofthwpx must not be routed to HwpExtractor"
+        );
+    }
 }