Nomad changes

e2e/rust/tests/smoke_test.rs

@@ -0,0 +1,272 @@
+// This file is auto-generated by alef — DO NOT EDIT.
+// alef:hash:4e15143f4af1ae8bafbdb1506ef057da924484c66a19483966333558ad437e75
+// To regenerate: alef generate
+// To verify freshness: alef verify --exit-code
+// Issues & docs: https://github.com/kreuzberg-dev/alef
+//! E2e tests for category: smoke
+
+use kreuzberg::{extract_bytes, extract_file};
+
+#[tokio::test]
+async fn test_ocr_image_png() {
+    // OCR: PNG image extraction with OCR enabled. In WASM this exercises the Uint8Array bridge parameter and Promise await in the generated OcrBackend bridge.
+    let content = std::fs::read(concat!(
+        env!("CARGO_MANIFEST_DIR"),
+        "/../../test_documents/images/test_hello_world.png"
+    ))
+    .expect("test_documents/images/test_hello_world.png must exist");
+    let mime_type = r#"image/png"#;
+    let config_json: serde_json::Value = serde_json::from_str(r#"{}"#).unwrap();
+    let config = serde_json::from_value(config_json).unwrap();
+    let result = extract_bytes(&content, mime_type, &config)
+        .await
+        .expect("should succeed");
+    assert_eq!(
+        result.mime_type.to_string().as_str().trim(),
+        r#"image/png"#,
+        "equals assertion failed"
+    );
+    assert!(
+        result.content.len() >= 1,
+        "expected length >= 1, got {}",
+        result.content.len()
+    );
+    assert!(
+        result.content.contains(r#"Hello"#)
+            || result.content.contains(r#"World"#)
+            || result.content.contains(r#"hello"#)
+            || result.content.contains(r#"world"#),
+        "expected to contain at least one of the specified values"
+    );
+}
+
+#[tokio::test]
+async fn test_smoke_docx_basic() {
+    // Smoke test: DOCX with formatted text
+    let path: &str = concat!(env!("CARGO_MANIFEST_DIR"), "/../../test_documents/", "docx/fake.docx");
+    let mime_type = Some(r#"application/vnd.openxmlformats-officedocument.wordprocessingml.document"#);
+    let config_json: serde_json::Value = serde_json::from_str(r#"{}"#).unwrap();
+    let config = serde_json::from_value(config_json).unwrap();
+    let result = extract_file(path, mime_type.as_deref(), &config)
+        .await
+        .expect("should succeed");
+    assert_eq!(
+        result.mime_type.to_string().as_str().trim(),
+        r#"application/vnd.openxmlformats-officedocument.wordprocessingml.document"#,
+        "equals assertion failed"
+    );
+    assert!(
+        result.content.len() >= 20,
+        "expected length >= 20, got {}",
+        result.content.len()
+    );
+    assert!(
+        result.content.contains(r#"Lorem"#)
+            || result.content.contains(r#"ipsum"#)
+            || result.content.contains(r#"document"#)
+            || result.content.contains(r#"text"#),
+        "expected to contain at least one of the specified values"
+    );
+}
+
+#[tokio::test]
+async fn test_smoke_html_basic() {
+    // Smoke test: HTML table extraction
+    let path: &str = concat!(
+        env!("CARGO_MANIFEST_DIR"),
+        "/../../test_documents/",
+        "html/simple_table.html"
+    );
+    let mime_type = Some(r#"text/html"#);
+    let config_json: serde_json::Value = serde_json::from_str(r#"{}"#).unwrap();
+    let config = serde_json::from_value(config_json).unwrap();
+    let result = extract_file(path, mime_type.as_deref(), &config)
+        .await
+        .expect("should succeed");
+    assert_eq!(
+        result.mime_type.to_string().as_str().trim(),
+        r#"text/html"#,
+        "equals assertion failed"
+    );
+    assert!(
+        result.content.len() >= 10,
+        "expected length >= 10, got {}",
+        result.content.len()
+    );
+    assert!(
+        result.content.contains(r#"Sample Data Table"#)
+            || result.content.contains(r#"Laptop"#)
+            || result.content.contains(r#"Electronics"#)
+            || result.content.contains(r#"Product"#),
+        "expected to contain at least one of the specified values"
+    );
+}
+
+#[tokio::test]
+async fn test_smoke_image_png() {
+    // Smoke test: PNG image (without OCR, metadata only)
+    let path: &str = concat!(
+        env!("CARGO_MANIFEST_DIR"),
+        "/../../test_documents/",
+        "images/sample.png"
+    );
+    let mime_type: Option<String> = None;
+    let config_json: serde_json::Value = serde_json::from_str(r#"{"disable_ocr":true}"#).unwrap();
+    let config = serde_json::from_value(config_json).unwrap();
+    let result = extract_file(path, mime_type.as_deref(), &config)
+        .await
+        .expect("should succeed");
+    assert_eq!(
+        result.mime_type.to_string().as_str().trim(),
+        r#"image/png"#,
+        "equals assertion failed"
+    );
+}
+
+#[tokio::test]
+async fn test_smoke_json_basic() {
+    // Smoke test: JSON file extraction
+    let path: &str = concat!(env!("CARGO_MANIFEST_DIR"), "/../../test_documents/", "json/simple.json");
+    let mime_type = Some(r#"application/json"#);
+    let config_json: serde_json::Value = serde_json::from_str(r#"{}"#).unwrap();
+    let config = serde_json::from_value(config_json).unwrap();
+    let result = extract_file(path, mime_type.as_deref(), &config)
+        .await
+        .expect("should succeed");
+    assert_eq!(
+        result.mime_type.to_string().as_str().trim(),
+        r#"application/json"#,
+        "equals assertion failed"
+    );
+    assert!(
+        result.content.len() >= 5,
+        "expected length >= 5, got {}",
+        result.content.len()
+    );
+}
+
+#[tokio::test]
+async fn test_smoke_pdf_basic() {
+    // Smoke test: PDF with simple text extraction
+    let path: &str = concat!(
+        env!("CARGO_MANIFEST_DIR"),
+        "/../../test_documents/",
+        "pdf/fake_memo.pdf"
+    );
+    let mime_type = Some(r#"application/pdf"#);
+    let config_json: serde_json::Value = serde_json::from_str(r#"{}"#).unwrap();
+    let config = serde_json::from_value(config_json).unwrap();
+    let result = extract_file(path, mime_type.as_deref(), &config)
+        .await
+        .expect("should succeed");
+    assert_eq!(
+        result.mime_type.to_string().as_str().trim(),
+        r#"application/pdf"#,
+        "equals assertion failed"
+    );
+    assert!(
+        result.content.len() >= 50,
+        "expected length >= 50, got {}",
+        result.content.len()
+    );
+    assert!(
+        result.content.contains(r#"May 5, 2023"#) || result.content.contains(r#"To Whom it May Concern"#),
+        "expected to contain at least one of the specified values"
+    );
+}
+
+#[tokio::test]
+async fn test_smoke_txt_basic() {
+    // Smoke test: Plain text file
+    let path: &str = concat!(env!("CARGO_MANIFEST_DIR"), "/../../test_documents/", "text/report.txt");
+    let mime_type = Some(r#"text/plain"#);
+    let config_json: serde_json::Value = serde_json::from_str(r#"{}"#).unwrap();
+    let config = serde_json::from_value(config_json).unwrap();
+    let result = extract_file(path, mime_type.as_deref(), &config)
+        .await
+        .expect("should succeed");
+    assert_eq!(
+        result.mime_type.to_string().as_str().trim(),
+        r#"text/plain"#,
+        "equals assertion failed"
+    );
+    assert!(
+        result.content.len() >= 5,
+        "expected length >= 5, got {}",
+        result.content.len()
+    );
+}
+
+#[tokio::test]
+async fn test_smoke_xlsx_basic() {
+    // Smoke test: XLSX with basic spreadsheet data including tables
+    let path: &str = concat!(
+        env!("CARGO_MANIFEST_DIR"),
+        "/../../test_documents/",
+        "xlsx/stanley_cups.xlsx"
+    );
+    let mime_type = Some(r#"application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"#);
+    let config_json: serde_json::Value = serde_json::from_str(r#"{}"#).unwrap();
+    let config = serde_json::from_value(config_json).unwrap();
+    let result = extract_file(path, mime_type.as_deref(), &config)
+        .await
+        .expect("should succeed");
+    assert_eq!(
+        result.mime_type.to_string().as_str().trim(),
+        r#"application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"#,
+        "equals assertion failed"
+    );
+    assert!(
+        result.content.len() >= 100,
+        "expected length >= 100, got {}",
+        result.content.len()
+    );
+    assert!(
+        format!("{:?}", result.content).contains(r#"Team"#),
+        "expected to contain: {}",
+        r#"Team"#
+    );
+    assert!(
+        format!("{:?}", result.content).contains(r#"Location"#),
+        "expected to contain: {}",
+        r#"Location"#
+    );
+    assert!(
+        format!("{:?}", result.content).contains(r#"Stanley Cups"#),
+        "expected to contain: {}",
+        r#"Stanley Cups"#
+    );
+    assert!(
+        format!("{:?}", result.content).contains(r#"Blues"#),
+        "expected to contain: {}",
+        r#"Blues"#
+    );
+    assert!(
+        format!("{:?}", result.content).contains(r#"Flyers"#),
+        "expected to contain: {}",
+        r#"Flyers"#
+    );
+    assert!(
+        format!("{:?}", result.content).contains(r#"Maple Leafs"#),
+        "expected to contain: {}",
+        r#"Maple Leafs"#
+    );
+    assert!(
+        format!("{:?}", result.content).contains(r#"STL"#),
+        "expected to contain: {}",
+        r#"STL"#
+    );
+    assert!(
+        format!("{:?}", result.content).contains(r#"PHI"#),
+        "expected to contain: {}",
+        r#"PHI"#
+    );
+    assert!(
+        format!("{:?}", result.content).contains(r#"TOR"#),
+        "expected to contain: {}",
+        r#"TOR"#
+    );
+    // skipped: field 'tables' not available on result type
+    // skipped: field 'metadata.format.excel.sheet_count' not available on result type
+    // skipped: field 'metadata.format.excel.sheet_names' not available on result type
+}