kreuzberg 4.2.6 → 4.2.7

data/vendor/kreuzberg/src/ocr/tesseract_backend.rs

@@ -8,7 +8,9 @@ use crate::core::config::OcrConfig;
 use crate::ocr::processor::OcrProcessor;
 use crate::plugins::{OcrBackend, OcrBackendType, Plugin};
 use crate::types::ExtractionResult;
+use ahash::AHashMap;
 use async_trait::async_trait;
+use std::borrow::Cow;
 use std::path::Path;
 use std::sync::{Arc, OnceLock};
 
@@ -196,9 +198,23 @@ impl OcrBackend for TesseractBackend {
             source: Some(Box::new(e)),
         })?;
 
+        // Use resolved language from OCR result metadata (handles "all"/"*" resolution)
+        let resolved_language = ocr_result
+            .metadata
+            .get("language")
+            .and_then(|v| v.as_str())
+            .unwrap_or(&tess_config.language)
+            .to_string();
+
+        // Convert HashMap<String, Value> to AHashMap<Cow<'static, str>, Value>
+        let mut additional = AHashMap::new();
+        for (key, value) in ocr_result.metadata {
+            additional.insert(Cow::Owned(key), value);
+        }
+
         let metadata = crate::types::Metadata {
             format: Some(crate::types::FormatMetadata::Ocr(crate::types::OcrMetadata {
-                language: tess_config.language.clone(),
+                language: resolved_language,
                 psm: tess_config.psm as i32,
                 output_format: tess_config.output_format.clone(),
                 table_count: ocr_result.tables.len(),
@@ -208,13 +224,13 @@ impl OcrBackend for TesseractBackend {
                     .first()
                     .and_then(|t| t.cells.first().map(|row| row.len())),
             })),
-            additional: ocr_result.metadata,
+            additional,
             ..Default::default()
         };
 
         Ok(ExtractionResult {
             content: ocr_result.content,
-            mime_type: ocr_result.mime_type,
+            mime_type: ocr_result.mime_type.into(),
             metadata,
             pages: None,
             tables: ocr_result
@@ -256,9 +272,23 @@ impl OcrBackend for TesseractBackend {
             source: Some(Box::new(e)),
         })?;
 
+        // Use resolved language from OCR result metadata (handles "all"/"*" resolution)
+        let resolved_language = ocr_result
+            .metadata
+            .get("language")
+            .and_then(|v| v.as_str())
+            .unwrap_or(&tess_config.language)
+            .to_string();
+
+        // Convert HashMap<String, Value> to AHashMap<Cow<'static, str>, Value>
+        let mut additional = AHashMap::new();
+        for (key, value) in ocr_result.metadata {
+            additional.insert(Cow::Owned(key), value);
+        }
+
         let metadata = crate::types::Metadata {
             format: Some(crate::types::FormatMetadata::Ocr(crate::types::OcrMetadata {
-                language: tess_config.language.clone(),
+                language: resolved_language,
                 psm: tess_config.psm as i32,
                 output_format: tess_config.output_format.clone(),
                 table_count: ocr_result.tables.len(),
@@ -268,13 +298,13 @@ impl OcrBackend for TesseractBackend {
                     .first()
                     .and_then(|t| t.cells.first().map(|row| row.len())),
             })),
-            additional: ocr_result.metadata,
+            additional,
             ..Default::default()
         };
 
         Ok(ExtractionResult {
             content: ocr_result.content,
-            mime_type: ocr_result.mime_type,
+            mime_type: ocr_result.mime_type.into(),
             metadata,
             pages: None,
             tables: ocr_result

data/vendor/kreuzberg/src/ocr/types.rs

@@ -1,5 +1,4 @@
 use serde::{Deserialize, Serialize};
-use std::collections::HashMap;
 
 pub use crate::types::ImagePreprocessingConfig;
 
@@ -154,7 +153,7 @@ impl From<&crate::types::TesseractConfig> for TesseractConfig {
 pub struct ExtractionResult {
     pub content: String,
     pub mime_type: String,
-    pub metadata: HashMap<String, serde_json::Value>,
+    pub metadata: std::collections::HashMap<String, serde_json::Value>,
     pub tables: Vec<Table>,
 }
 
@@ -260,7 +259,7 @@ mod tests {
 
     #[test]
     fn test_extraction_result_creation() {
-        let mut metadata = HashMap::new();
+        let mut metadata = std::collections::HashMap::new();
         metadata.insert("key".to_string(), serde_json::json!("value"));
 
         let table = Table {
@@ -308,7 +307,7 @@ mod tests {
         let result = crate::types::OcrExtractionResult {
             content: "content".to_string(),
             mime_type: "text/plain".to_string(),
-            metadata: HashMap::new(),
+            metadata: std::collections::HashMap::new(),
             tables: vec![],
         };
 

data/vendor/kreuzberg/src/ocr/validation.rs

@@ -131,6 +131,12 @@ lazy_static::lazy_static! {
 }
 
 pub fn validate_language_code(lang_code: &str) -> Result<(), OcrError> {
+    // Accept "all" and "*" as special values to auto-detect installed languages
+    let lower = lang_code.to_ascii_lowercase();
+    if lower == "all" || lower == "*" {
+        return Ok(());
+    }
+
     for code in lang_code.split('+') {
         if !TESSERACT_SUPPORTED_LANGUAGE_CODES.contains(code) {
             return Err(OcrError::InvalidLanguageCode(format!(
@@ -156,6 +162,14 @@ pub fn validate_tesseract_version(version: u32) -> Result<(), OcrError> {
 mod tests {
     use super::*;
 
+    #[test]
+    fn test_validate_language_code_all_keyword() {
+        assert!(validate_language_code("all").is_ok());
+        assert!(validate_language_code("*").is_ok());
+        assert!(validate_language_code("ALL").is_ok());
+        assert!(validate_language_code("All").is_ok());
+    }
+
     #[test]
     fn test_validate_language_code_valid() {
         assert!(validate_language_code("eng").is_ok());

data/vendor/kreuzberg/src/pdf/metadata.rs

@@ -10,6 +10,7 @@ use serde::{Deserialize, Serialize};
 /// `Metadata` structure. Common fields like title, authors, keywords, and dates
 /// are now at the `Metadata` level.
 #[derive(Debug, Clone, Serialize, Deserialize, Default)]
+#[cfg_attr(feature = "api", derive(utoipa::ToSchema))]
 pub struct PdfMetadata {
     /// PDF version (e.g., "1.7", "2.0")
     #[serde(skip_serializing_if = "Option::is_none")]

data/vendor/kreuzberg/src/plugins/extractor/mod.rs

@@ -19,6 +19,7 @@ mod tests {
     use crate::plugins::Plugin;
     use crate::types::ExtractionResult;
     use async_trait::async_trait;
+    use std::borrow::Cow;
 
     struct MockExtractor {
         mime_types: Vec<&'static str>,
@@ -53,7 +54,7 @@ mod tests {
         ) -> Result<ExtractionResult> {
             Ok(ExtractionResult {
                 content: String::from_utf8_lossy(content).to_string(),
-                mime_type: mime_type.to_string(),
+                mime_type: mime_type.to_string().into(),
                 metadata: crate::types::Metadata::default(),
                 tables: vec![],
                 detected_languages: None,
@@ -228,7 +229,7 @@ mod tests {
             ) -> Result<ExtractionResult> {
                 Ok(ExtractionResult {
                     content: String::new(),
-                    mime_type: String::new(),
+                    mime_type: Cow::Borrowed(""),
                     metadata: crate::types::Metadata::default(),
                     tables: vec![],
                     detected_languages: None,

data/vendor/kreuzberg/src/plugins/extractor/registry.rs

@@ -50,7 +50,7 @@ use std::sync::Arc;
 ///         -> Result<ExtractionResult> {
 ///         Ok(ExtractionResult {
 ///             content: String::from_utf8_lossy(content).to_string(),
-///             mime_type: mime_type.to_string(),
+///             mime_type: mime_type.to_string().into(),
 ///             metadata: Metadata::default(),
 ///             tables: vec![],
 ///             detected_languages: None,
@@ -189,6 +189,7 @@ mod tests {
     use crate::types::ExtractionResult;
     use async_trait::async_trait;
     use serial_test::serial;
+    use std::borrow::Cow;
 
     struct MockExtractor {
         mime_types: Vec<&'static str>,
@@ -223,7 +224,7 @@ mod tests {
         ) -> Result<ExtractionResult> {
             Ok(ExtractionResult {
                 content: String::from_utf8_lossy(content).to_string(),
-                mime_type: mime_type.to_string(),
+                mime_type: mime_type.to_string().into(),
                 metadata: crate::types::Metadata::default(),
                 tables: vec![],
                 detected_languages: None,
@@ -362,7 +363,7 @@ mod tests {
             async fn extract_bytes(&self, _: &[u8], _: &str, _: &ExtractionConfig) -> Result<ExtractionResult> {
                 Ok(ExtractionResult {
                     content: String::new(),
-                    mime_type: String::new(),
+                    mime_type: Cow::Borrowed(""),
                     metadata: crate::types::Metadata::default(),
                     tables: vec![],
                     detected_languages: None,
@@ -410,7 +411,7 @@ mod tests {
             async fn extract_bytes(&self, _: &[u8], _: &str, _: &ExtractionConfig) -> Result<ExtractionResult> {
                 Ok(ExtractionResult {
                     content: String::new(),
-                    mime_type: String::new(),
+                    mime_type: Cow::Borrowed(""),
                     metadata: crate::types::Metadata::default(),
                     tables: vec![],
                     detected_languages: None,

data/vendor/kreuzberg/src/plugins/ocr.rs

@@ -61,7 +61,7 @@ pub enum OcrBackendType {
 ///         // Implement OCR logic here
 ///         Ok(ExtractionResult {
 ///             content: "Extracted text".to_string(),
-///             mime_type: "text/plain".to_string(),
+///             mime_type: Cow::Borrowed("text/plain"),
 ///             metadata: Metadata::default(),
 ///             tables: vec![],
 ///             detected_languages: None,
@@ -142,7 +142,7 @@ pub trait OcrBackend: Plugin {
     ///
     ///     Ok(ExtractionResult {
     ///         content: text,
-    ///         mime_type: "text/plain".to_string(),
+    ///         mime_type: Cow::Borrowed("text/plain"),
     ///         metadata: Metadata::default(),
     ///         tables: vec![],
     ///         detected_languages: None,
@@ -315,7 +315,7 @@ pub trait OcrBackend: Plugin {
 ///     async fn process_image(&self, _: &[u8], _: &OcrConfig) -> Result<ExtractionResult> {
 ///         Ok(ExtractionResult {
 ///             content: "text".to_string(),
-///             mime_type: "text/plain".to_string(),
+///             mime_type: Cow::Borrowed("text/plain"),
 ///             metadata: Metadata::default(),
 ///             tables: vec![],
 ///             detected_languages: None,
@@ -450,6 +450,7 @@ pub fn clear_ocr_backends() -> crate::Result<()> {
 #[cfg(test)]
 mod tests {
     use super::*;
+    use std::borrow::Cow;
 
     struct MockOcrBackend {
         languages: Vec<String>,
@@ -478,7 +479,7 @@ mod tests {
         async fn process_image(&self, _image_bytes: &[u8], _config: &OcrConfig) -> Result<ExtractionResult> {
             Ok(ExtractionResult {
                 content: "Mocked OCR text".to_string(),
-                mime_type: "text/plain".to_string(),
+                mime_type: Cow::Borrowed("text/plain"),
                 metadata: crate::types::Metadata::default(),
                 tables: vec![],
                 detected_languages: None,

data/vendor/kreuzberg/src/plugins/processor/mod.rs

@@ -18,8 +18,9 @@ mod tests {
     use crate::core::config::ExtractionConfig;
     use crate::plugins::Plugin;
     use crate::types::ExtractionResult;
+    use ahash::AHashMap;
     use async_trait::async_trait;
-    use std::collections::HashMap;
+    use std::borrow::Cow;
 
     struct MockPostProcessor {
         stage: ProcessingStage,
@@ -49,7 +50,7 @@ mod tests {
             result
                 .metadata
                 .additional
-                .insert("processed_by".to_string(), serde_json::json!(self.name()));
+                .insert(Cow::Borrowed("processed_by"), serde_json::json!(self.name()));
             Ok(())
         }
 
@@ -66,7 +67,7 @@ mod tests {
 
         let mut result = ExtractionResult {
             content: "test content".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -118,7 +119,7 @@ mod tests {
 
         let result = ExtractionResult {
             content: "test".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -187,7 +188,7 @@ mod tests {
 
         let mut result = ExtractionResult {
             content: String::new(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -211,12 +212,12 @@ mod tests {
             stage: ProcessingStage::Early,
         };
 
-        let mut additional = HashMap::new();
-        additional.insert("existing_key".to_string(), serde_json::json!("existing_value"));
+        let mut additional = AHashMap::new();
+        additional.insert(Cow::Borrowed("existing_key"), serde_json::json!("existing_value"));
 
         let mut result = ExtractionResult {
             content: "test".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata {
                 additional,
                 ..Default::default()
@@ -248,7 +249,7 @@ mod tests {
 
         let result = ExtractionResult {
             content: "test".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -301,7 +302,7 @@ mod tests {
 
         let pdf_result = ExtractionResult {
             content: "test".to_string(),
-            mime_type: "application/pdf".to_string(),
+            mime_type: Cow::Borrowed("application/pdf"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -314,7 +315,7 @@ mod tests {
 
         let txt_result = ExtractionResult {
             content: "test".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -345,7 +346,7 @@ mod tests {
 
         let mut result = ExtractionResult {
             content: "test".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![table],
             detected_languages: None,

data/vendor/kreuzberg/src/plugins/registry/extractor.rs

@@ -225,6 +225,7 @@ mod tests {
     use crate::plugins::Plugin;
     use crate::types::ExtractionResult;
     use async_trait::async_trait;
+    use std::borrow::Cow;
 
     struct MockExtractor {
         name: String,
@@ -252,7 +253,7 @@ mod tests {
         async fn extract_bytes(&self, _: &[u8], _: &str, _: &ExtractionConfig) -> Result<ExtractionResult> {
             Ok(ExtractionResult {
                 content: "test".to_string(),
-                mime_type: "text/plain".to_string(),
+                mime_type: Cow::Borrowed("text/plain"),
                 metadata: crate::types::Metadata::default(),
                 tables: vec![],
                 detected_languages: None,
@@ -494,7 +495,7 @@ mod tests {
         async fn extract_bytes(&self, _: &[u8], _: &str, _: &ExtractionConfig) -> Result<ExtractionResult> {
             Ok(ExtractionResult {
                 content: "test".to_string(),
-                mime_type: "text/plain".to_string(),
+                mime_type: Cow::Borrowed("text/plain"),
                 metadata: crate::types::Metadata::default(),
                 tables: vec![],
                 detected_languages: None,

data/vendor/kreuzberg/src/plugins/registry/ocr.rs

@@ -191,6 +191,7 @@ mod tests {
     use crate::plugins::{OcrBackend, Plugin};
     use crate::types::ExtractionResult;
     use async_trait::async_trait;
+    use std::borrow::Cow;
 
     struct MockOcrBackend {
         name: String,
@@ -217,7 +218,7 @@ mod tests {
         async fn process_image(&self, _: &[u8], _: &OcrConfig) -> Result<ExtractionResult> {
             Ok(ExtractionResult {
                 content: "test".to_string(),
-                mime_type: "text/plain".to_string(),
+                mime_type: Cow::Borrowed("text/plain"),
                 metadata: crate::types::Metadata::default(),
                 tables: vec![],
                 detected_languages: None,
@@ -344,7 +345,7 @@ mod tests {
         async fn process_image(&self, _: &[u8], _: &OcrConfig) -> Result<ExtractionResult> {
             Ok(ExtractionResult {
                 content: "test".to_string(),
-                mime_type: "text/plain".to_string(),
+                mime_type: Cow::Borrowed("text/plain"),
                 metadata: crate::types::Metadata::default(),
                 tables: vec![],
                 detected_languages: None,

data/vendor/kreuzberg/src/plugins/validator/mod.rs

@@ -19,8 +19,9 @@ mod tests {
     use crate::core::config::ExtractionConfig;
     use crate::plugins::Plugin;
     use crate::types::ExtractionResult;
+    use ahash::AHashMap;
     use async_trait::async_trait;
-    use std::collections::HashMap;
+    use std::borrow::Cow;
 
     struct MockValidator {
         should_fail: bool,
@@ -61,7 +62,7 @@ mod tests {
 
         let result = ExtractionResult {
             content: "test content".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -82,7 +83,7 @@ mod tests {
 
         let result = ExtractionResult {
             content: "test content".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -105,7 +106,7 @@ mod tests {
 
         let result = ExtractionResult {
             content: "test".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -143,7 +144,7 @@ mod tests {
 
         let result = ExtractionResult {
             content: String::new(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -193,7 +194,7 @@ mod tests {
 
         let pdf_result = ExtractionResult {
             content: "test".to_string(),
-            mime_type: "application/pdf".to_string(),
+            mime_type: Cow::Borrowed("application/pdf"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -206,7 +207,7 @@ mod tests {
 
         let txt_result = ExtractionResult {
             content: "test".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -292,7 +293,7 @@ mod tests {
 
         let result = ExtractionResult {
             content: "test".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -318,12 +319,12 @@ mod tests {
     async fn test_validator_with_metadata() {
         let validator = MockValidator { should_fail: false };
 
-        let mut additional = HashMap::new();
-        additional.insert("quality_score".to_string(), serde_json::json!(0.95));
+        let mut additional = AHashMap::new();
+        additional.insert(Cow::Borrowed("quality_score"), serde_json::json!(0.95));
 
         let result = ExtractionResult {
             content: "test".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata {
                 additional,
                 ..Default::default()
@@ -355,7 +356,7 @@ mod tests {
 
         let result = ExtractionResult {
             content: "test".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![table],
             detected_languages: None,
@@ -386,7 +387,7 @@ mod tests {
         for mime_type in mime_types {
             let result = ExtractionResult {
                 content: "test".to_string(),
-                mime_type: mime_type.to_string(),
+                mime_type: Cow::Borrowed(mime_type),
                 metadata: crate::types::Metadata::default(),
                 tables: vec![],
                 detected_languages: None,
@@ -407,7 +408,7 @@ mod tests {
 
         let result = ExtractionResult {
             content: "test content ".repeat(10000),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: crate::types::Metadata::default(),
             tables: vec![],
             detected_languages: None,

data/vendor/kreuzberg/src/text/quality.rs

@@ -1,7 +1,7 @@
+use ahash::AHashMap;
 use once_cell::sync::Lazy;
 use regex::Regex;
 use std::borrow::Cow;
-use std::collections::HashMap;
 
 use crate::utils::quality::{collapse_scattered_ascii, normalize_whitespace_ascii};
 
@@ -123,7 +123,7 @@ where
     }
 }
 
-pub fn calculate_quality_score(text: &str, metadata: Option<&HashMap<String, serde_json::Value>>) -> f64 {
+pub fn calculate_quality_score(text: &str, metadata: Option<&AHashMap<Cow<'static, str>, serde_json::Value>>) -> f64 {
     if text.is_empty() || text.trim().is_empty() {
         return 0.0;
     }
@@ -266,7 +266,7 @@ fn calculate_structure_bonus(text: &str) -> f64 {
 }
 
 #[inline]
-fn calculate_metadata_bonus(metadata: &HashMap<String, serde_json::Value>) -> f64 {
+fn calculate_metadata_bonus(metadata: &AHashMap<Cow<'static, str>, serde_json::Value>) -> f64 {
     const IMPORTANT_FIELDS: &[&str] = &["title", "author", "subject", "description", "keywords"];
 
     let present_fields = IMPORTANT_FIELDS
@@ -491,9 +491,9 @@ mod tests {
     #[test]
     fn test_calculate_quality_score_with_metadata() {
         let text = "This is a normal text with proper structure.";
-        let mut metadata = HashMap::new();
-        metadata.insert("title".to_string(), serde_json::json!("Test Title"));
-        metadata.insert("author".to_string(), serde_json::json!("Test Author"));
+        let mut metadata: AHashMap<Cow<'static, str>, serde_json::Value> = AHashMap::new();
+        metadata.insert(Cow::Borrowed("title"), serde_json::json!("Test Title"));
+        metadata.insert(Cow::Borrowed("author"), serde_json::json!("Test Author"));
 
         let score = calculate_quality_score(text, Some(&metadata));
         assert!(score > 0.0);
@@ -558,19 +558,19 @@ mod tests {
 
     #[test]
     fn test_calculate_metadata_bonus_empty() {
-        let metadata = HashMap::new();
+        let metadata: AHashMap<Cow<'static, str>, serde_json::Value> = AHashMap::new();
         let bonus = calculate_metadata_bonus(&metadata);
         assert_eq!(bonus, 0.0);
     }
 
     #[test]
     fn test_calculate_metadata_bonus_full() {
-        let mut metadata = HashMap::new();
-        metadata.insert("title".to_string(), serde_json::json!("Title"));
-        metadata.insert("author".to_string(), serde_json::json!("Author"));
-        metadata.insert("subject".to_string(), serde_json::json!("Subject"));
-        metadata.insert("description".to_string(), serde_json::json!("Description"));
-        metadata.insert("keywords".to_string(), serde_json::json!("Keywords"));
+        let mut metadata: AHashMap<Cow<'static, str>, serde_json::Value> = AHashMap::new();
+        metadata.insert(Cow::Borrowed("title"), serde_json::json!("Title"));
+        metadata.insert(Cow::Borrowed("author"), serde_json::json!("Author"));
+        metadata.insert(Cow::Borrowed("subject"), serde_json::json!("Subject"));
+        metadata.insert(Cow::Borrowed("description"), serde_json::json!("Description"));
+        metadata.insert(Cow::Borrowed("keywords"), serde_json::json!("Keywords"));
 
         let bonus = calculate_metadata_bonus(&metadata);
         assert_eq!(bonus, 1.0);

data/vendor/kreuzberg/src/text/quality_processor.rs

@@ -15,6 +15,7 @@
 use crate::plugins::{Plugin, PostProcessor, ProcessingStage};
 use crate::{ExtractionConfig, ExtractionResult, Result};
 use async_trait::async_trait;
+use std::borrow::Cow;
 
 /// Post-processor that calculates quality score and cleans text.
 ///
@@ -65,7 +66,7 @@ impl PostProcessor for QualityProcessor {
         };
 
         result.metadata.additional.insert(
-            "quality_score".to_string(),
+            Cow::Borrowed("quality_score"),
             serde_json::Value::Number(
                 serde_json::Number::from_f64(quality_score).unwrap_or(serde_json::Number::from(0)),
             ),
@@ -116,7 +117,7 @@ mod tests {
 
         let mut result = ExtractionResult {
 	            content: "This is a well-written paragraph with proper structure. It contains multiple sentences. The quality should be good.".to_string(),
-	            mime_type: "text/plain".to_string(),
+	            mime_type: Cow::Borrowed("text/plain"),
 	            metadata: Metadata::default(),
 	            tables: vec![],
 	            detected_languages: None,
@@ -144,7 +145,7 @@ mod tests {
 
         let mut result = ExtractionResult {
             content: "Some text".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -179,7 +180,7 @@ mod tests {
 
         let result = ExtractionResult {
             content: "Sample text".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -209,7 +210,7 @@ mod tests {
 
         let short_result = ExtractionResult {
             content: "Short".to_string(),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: Metadata::default(),
             tables: vec![],
             detected_languages: None,
@@ -222,7 +223,7 @@ mod tests {
 
         let long_result = ExtractionResult {
             content: "a".repeat(1000000),
-            mime_type: "text/plain".to_string(),
+            mime_type: Cow::Borrowed("text/plain"),
             metadata: Metadata::default(),
             tables: vec![],
             detected_languages: None,