kreuzberg 4.2.6 → 4.2.7

data/vendor/kreuzberg/src/core/pipeline/tests.rs

@@ -4,6 +4,7 @@ use super::*;
 use crate::core::config::OutputFormat;
 use crate::types::Metadata;
 use lazy_static::lazy_static;
+use std::borrow::Cow;
 
 const VALIDATION_MARKER_KEY: &str = "registry_validation_marker";
 #[cfg(feature = "quality")]
@@ -19,7 +20,7 @@ lazy_static! {
 async fn test_run_pipeline_basic() {
     let mut result = ExtractionResult {
         content: "test".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -30,7 +31,7 @@ async fn test_run_pipeline_basic() {
         elements: None,
     };
     result.metadata.additional.insert(
-        VALIDATION_MARKER_KEY.to_string(),
+        Cow::Borrowed(VALIDATION_MARKER_KEY),
         serde_json::json!(ORDER_VALIDATION_MARKER),
     );
     let config = ExtractionConfig::default();
@@ -44,7 +45,7 @@ async fn test_run_pipeline_basic() {
 async fn test_pipeline_with_quality_processing() {
     let result = ExtractionResult {
         content: "This is a test document with some meaningful content.".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -67,7 +68,7 @@ async fn test_pipeline_with_quality_processing() {
 async fn test_pipeline_without_quality_processing() {
     let result = ExtractionResult {
         content: "test".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -91,7 +92,7 @@ async fn test_pipeline_without_quality_processing() {
 async fn test_pipeline_with_chunking() {
     let result = ExtractionResult {
         content: "This is a long text that should be chunked. ".repeat(100),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -103,8 +104,10 @@ async fn test_pipeline_with_chunking() {
     };
     let config = ExtractionConfig {
         chunking: Some(crate::ChunkingConfig {
-            max_chars: 500,
-            max_overlap: 50,
+            max_characters: 500,
+            overlap: 50,
+            trim: true,
+            chunker_type: crate::ChunkerType::Text,
             embedding: None,
             preset: None,
         }),
@@ -121,7 +124,7 @@ async fn test_pipeline_with_chunking() {
 async fn test_pipeline_without_chunking() {
     let result = ExtractionResult {
         content: "test".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -142,14 +145,14 @@ async fn test_pipeline_without_chunking() {
 
 #[tokio::test]
 async fn test_pipeline_preserves_metadata() {
-    use std::collections::HashMap;
-    let mut additional = HashMap::new();
-    additional.insert("source".to_string(), serde_json::json!("test"));
-    additional.insert("page".to_string(), serde_json::json!(1));
+    use ahash::AHashMap;
+    let mut additional = AHashMap::new();
+    additional.insert(Cow::Borrowed("source"), serde_json::json!("test"));
+    additional.insert(Cow::Borrowed("page"), serde_json::json!(1));
 
     let result = ExtractionResult {
         content: "test".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata {
             additional,
             ..Default::default()
@@ -187,7 +190,7 @@ async fn test_pipeline_preserves_tables() {
 
     let result = ExtractionResult {
         content: "test".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![table],
         detected_languages: None,
@@ -219,7 +222,7 @@ async fn test_pipeline_empty_content() {
 
     let result = ExtractionResult {
         content: String::new(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -242,7 +245,7 @@ async fn test_pipeline_empty_content() {
 async fn test_pipeline_with_all_features() {
     let result = ExtractionResult {
         content: "This is a comprehensive test document. ".repeat(50),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -255,8 +258,10 @@ async fn test_pipeline_with_all_features() {
     let config = ExtractionConfig {
         enable_quality_processing: true,
         chunking: Some(crate::ChunkingConfig {
-            max_chars: 500,
-            max_overlap: 50,
+            max_characters: 500,
+            overlap: 50,
+            trim: true,
+            chunker_type: crate::ChunkerType::Text,
             embedding: None,
             preset: None,
         }),
@@ -295,7 +300,7 @@ machine learning that uses neural networks with multiple layers.
 Natural language processing enables computers to understand human language.
             "#
         .to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -342,7 +347,7 @@ async fn test_pipeline_without_keyword_config() {
     }
     let result = ExtractionResult {
         content: "Machine learning and artificial intelligence.".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -380,7 +385,7 @@ async fn test_pipeline_keyword_extraction_short_content() {
 
     let result = ExtractionResult {
         content: "Short text".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -437,7 +442,7 @@ async fn test_postprocessor_runs_before_validator() {
             result
                 .metadata
                 .additional
-                .insert("processed".to_string(), serde_json::json!(true));
+                .insert(Cow::Borrowed("processed"), serde_json::json!(true));
             Ok(())
         }
 
@@ -517,7 +522,7 @@ async fn test_postprocessor_runs_before_validator() {
 
     let mut result = ExtractionResult {
         content: "test".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -528,7 +533,7 @@ async fn test_postprocessor_runs_before_validator() {
         elements: None,
     };
     result.metadata.additional.insert(
-        VALIDATION_MARKER_KEY.to_string(),
+        Cow::Borrowed(VALIDATION_MARKER_KEY),
         serde_json::json!(POSTPROCESSOR_VALIDATION_MARKER),
     );
 
@@ -614,7 +619,7 @@ async fn test_quality_processing_runs_before_validator() {
 
     let mut result = ExtractionResult {
         content: "This is meaningful test content for quality scoring.".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -625,7 +630,7 @@ async fn test_quality_processing_runs_before_validator() {
         elements: None,
     };
     result.metadata.additional.insert(
-        VALIDATION_MARKER_KEY.to_string(),
+        Cow::Borrowed(VALIDATION_MARKER_KEY),
         serde_json::json!(QUALITY_VALIDATION_MARKER),
     );
 
@@ -682,7 +687,7 @@ async fn test_multiple_postprocessors_run_before_validator() {
             result
                 .metadata
                 .additional
-                .insert("execution_order".to_string(), serde_json::json!(order));
+                .insert(Cow::Borrowed("execution_order"), serde_json::json!(order));
             Ok(())
         }
 
@@ -721,7 +726,7 @@ async fn test_multiple_postprocessors_run_before_validator() {
             result
                 .metadata
                 .additional
-                .insert("execution_order".to_string(), serde_json::json!(order));
+                .insert(Cow::Borrowed("execution_order"), serde_json::json!(order));
             Ok(())
         }
 
@@ -812,7 +817,7 @@ async fn test_multiple_postprocessors_run_before_validator() {
 
     let result = ExtractionResult {
         content: "test".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -839,7 +844,7 @@ async fn test_multiple_postprocessors_run_before_validator() {
 async fn test_run_pipeline_with_output_format_plain() {
     let result = ExtractionResult {
         content: "test content".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -865,7 +870,7 @@ async fn test_run_pipeline_with_output_format_djot() {
 
     let result = ExtractionResult {
         content: "test content".to_string(),
-        mime_type: "text/djot".to_string(),
+        mime_type: Cow::Borrowed("text/djot"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -894,7 +899,7 @@ async fn test_run_pipeline_with_output_format_djot() {
             images: vec![],
             links: vec![],
             footnotes: vec![],
-            attributes: std::collections::HashMap::new(),
+            attributes: Vec::new(),
         }),
     };
 
@@ -912,7 +917,7 @@ async fn test_run_pipeline_with_output_format_djot() {
 async fn test_run_pipeline_with_output_format_html() {
     let result = ExtractionResult {
         content: "test content".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -942,7 +947,7 @@ async fn test_run_pipeline_applies_output_format_last() {
 
     let result = ExtractionResult {
         content: "test".to_string(),
-        mime_type: "text/plain".to_string(),
+        mime_type: Cow::Borrowed("text/plain"),
         metadata: Metadata::default(),
         tables: vec![],
         detected_languages: None,
@@ -958,7 +963,7 @@ async fn test_run_pipeline_applies_output_format_last() {
             images: vec![],
             links: vec![],
             footnotes: vec![],
-            attributes: std::collections::HashMap::new(),
+            attributes: Vec::new(),
         }),
     };
 

data/vendor/kreuzberg/src/extraction/email.rs

@@ -24,6 +24,8 @@
 //! # Ok(())
 //! # }
 //! ```
+use bytes::Bytes;
+
 use crate::error::{KreuzbergError, Result};
 use crate::types::{EmailAttachment, EmailExtractionResult};
 use mail_parser::MimeHeaders;
@@ -101,7 +103,7 @@ pub fn parse_eml_content(data: &[u8]) -> Result<EmailExtractionResult> {
 
     let html_content = message.body_html(0).map(|s| s.to_string());
 
-    let cleaned_text = if let Some(plain) = &plain_text {
+    let cleaned_text = if let Some(ref plain) = plain_text {
         plain.clone()
     } else if let Some(html) = &html_content {
         clean_html_content(html)
@@ -132,7 +134,7 @@ pub fn parse_eml_content(data: &[u8]) -> Result<EmailExtractionResult> {
             mime_type: Some(mime_type),
             size: Some(size),
             is_image,
-            data: Some(data.to_vec()),
+            data: Some(Bytes::copy_from_slice(data)),
         });
     }
 
@@ -174,39 +176,49 @@ pub fn parse_msg_content(data: &[u8]) -> Result<EmailExtractionResult> {
     let to_emails = outlook
         .to
         .iter()
-        .map(|p| p.email.clone())
-        .filter(|e| !e.is_empty())
+        .filter_map(|p| {
+            if p.email.is_empty() {
+                None
+            } else {
+                Some(p.email.clone())
+            }
+        })
         .collect::<Vec<String>>();
 
     let cc_emails = outlook
         .cc
         .iter()
-        .map(|p| p.email.clone())
-        .filter(|e| !e.is_empty())
+        .filter_map(|p| {
+            if p.email.is_empty() {
+                None
+            } else {
+                Some(p.email.clone())
+            }
+        })
         .collect::<Vec<String>>();
 
-    let bcc_emails = if !outlook.bcc.is_empty() {
-        vec![outlook.bcc.clone()]
-    } else {
+    let bcc_emails = if outlook.bcc.is_empty() {
         vec![]
+    } else {
+        vec![outlook.bcc.clone()]
     };
 
-    let date = if !outlook.headers.date.is_empty() {
-        Some(outlook.headers.date.clone())
-    } else {
+    let date = if outlook.headers.date.is_empty() {
         None
+    } else {
+        Some(outlook.headers.date.clone())
     };
 
-    let message_id = if !outlook.headers.message_id.is_empty() {
-        Some(outlook.headers.message_id.clone())
-    } else {
+    let message_id = if outlook.headers.message_id.is_empty() {
         None
+    } else {
+        Some(outlook.headers.message_id.clone())
     };
 
-    let plain_text = if !outlook.body.is_empty() {
-        Some(outlook.body.clone())
-    } else {
+    let plain_text = if outlook.body.is_empty() {
         None
+    } else {
+        Some(outlook.body.clone())
     };
 
     let html_content = None;
@@ -231,7 +243,7 @@ pub fn parse_msg_content(data: &[u8]) -> Result<EmailExtractionResult> {
             };
 
             let data = if !att.payload.is_empty() {
-                hex::decode(&att.payload).ok()
+                hex::decode(&att.payload).ok().map(Bytes::from)
             } else {
                 None
             };

data/vendor/kreuzberg/src/extraction/excel.rs

@@ -448,13 +448,13 @@ fn generate_markdown_and_cells(sheet_name: &str, range: &Range<Data>, capacity:
             markdown.push_str(" | ");
         }
         let cell_str = format_cell_to_string(cell);
-        header_cells.push(cell_str.clone());
 
         if cell_str.contains('|') || cell_str.contains('\\') {
             escape_markdown_into(&mut markdown, &cell_str);
         } else {
             markdown.push_str(&cell_str);
         }
+        header_cells.push(cell_str);
     }
     markdown.push_str(" |\n");
     cells.push(header_cells);
@@ -475,18 +475,19 @@ fn generate_markdown_and_cells(sheet_name: &str, range: &Range<Data>, capacity:
             if i > 0 {
                 markdown.push_str(" | ");
             }
-            if let Some(cell) = row.get(i) {
+            let cell_str = if let Some(cell) = row.get(i) {
                 let cell_str = format_cell_to_string(cell);
-                row_cells.push(cell_str.clone());
 
                 if cell_str.contains('|') || cell_str.contains('\\') {
                     escape_markdown_into(&mut markdown, &cell_str);
                 } else {
                     markdown.push_str(&cell_str);
                 }
+                cell_str
             } else {
-                row_cells.push(String::new());
-            }
+                String::new()
+            };
+            row_cells.push(cell_str);
         }
         markdown.push_str(" |\n");
         cells.push(row_cells);

data/vendor/kreuzberg/src/extraction/html/image_handling.rs

@@ -1,5 +1,7 @@
 //! Image handling and conversion functionality for HTML extraction.
 
+use bytes::Bytes;
+
 use super::types::ExtractedInlineImage;
 use html_to_markdown_rs::{InlineImage, InlineImageFormat};
 
@@ -49,13 +51,16 @@ pub fn inline_image_format_to_str(format: &InlineImageFormat) -> String {
     }
 }
 
+// Note: This function returns String because ExtractedInlineImage.format is String (internal to HTML extraction).
+// For external ExtractedImage, use detect_image_format from pptx which returns Cow<'static, str>.
+
 /// Convert a library InlineImage to an ExtractedInlineImage.
 ///
 /// Maps the library's image representation to the extraction API's format,
 /// converting the format enum to a string representation.
 pub fn inline_image_to_extracted(image: InlineImage) -> ExtractedInlineImage {
     ExtractedInlineImage {
-        data: image.data,
+        data: Bytes::from(image.data),
         format: inline_image_format_to_str(&image.format),
         filename: image.filename,
         description: image.description,

data/vendor/kreuzberg/src/extraction/html/types.rs

@@ -1,7 +1,7 @@
 //! Type definitions for HTML extraction.
 
+use bytes::Bytes;
 use serde::{Deserialize, Serialize};
-use std::collections::HashMap;
 
 pub use html_to_markdown_rs::{
     CodeBlockStyle, HeadingStyle, HighlightStyle, ListIndentType, NewlineStyle, PreprocessingOptions,
@@ -19,10 +19,11 @@ pub struct HtmlExtractionResult {
 /// Extracted inline image with metadata.
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct ExtractedInlineImage {
-    pub data: Vec<u8>,
+    /// Uses `bytes::Bytes` for cheap cloning of large buffers.
+    pub data: Bytes,
     pub format: String,
     pub filename: Option<String>,
     pub description: Option<String>,
     pub dimensions: Option<(u32, u32)>,
-    pub attributes: HashMap<String, String>,
+    pub attributes: Vec<(String, String)>,
 }

data/vendor/kreuzberg/src/extraction/libreoffice.rs

@@ -45,6 +45,7 @@
 
 use crate::error::{KreuzbergError, Result};
 use crate::types::LibreOfficeConversionResult;
+use std::borrow::Cow;
 use std::collections::HashSet;
 use std::env;
 use std::fs as std_fs;
@@ -326,9 +327,9 @@ pub async fn convert_doc_to_docx(doc_bytes: &[u8]) -> Result<LibreOfficeConversi
 
     Ok(LibreOfficeConversionResult {
         converted_bytes,
-        original_format: "doc".to_string(),
-        target_format: "docx".to_string(),
-        target_mime: crate::core::mime::DOCX_MIME_TYPE.to_string(),
+        original_format: Cow::Borrowed("doc"),
+        target_format: Cow::Borrowed("docx"),
+        target_mime: Cow::Borrowed(crate::core::mime::DOCX_MIME_TYPE),
     })
 }
 
@@ -350,9 +351,9 @@ pub async fn convert_ppt_to_pptx(ppt_bytes: &[u8]) -> Result<LibreOfficeConversi
 
     Ok(LibreOfficeConversionResult {
         converted_bytes,
-        original_format: "ppt".to_string(),
-        target_format: "pptx".to_string(),
-        target_mime: crate::core::mime::POWER_POINT_MIME_TYPE.to_string(),
+        original_format: Cow::Borrowed("ppt"),
+        target_format: Cow::Borrowed("pptx"),
+        target_mime: Cow::Borrowed(crate::core::mime::POWER_POINT_MIME_TYPE),
     })
 }
 
@@ -505,9 +506,9 @@ mod tests {
     async fn test_conversion_result_structure() {
         let result = LibreOfficeConversionResult {
             converted_bytes: vec![1, 2, 3],
-            original_format: "doc".to_string(),
-            target_format: "docx".to_string(),
-            target_mime: crate::core::mime::DOCX_MIME_TYPE.to_string(),
+            original_format: Cow::Borrowed("doc"),
+            target_format: Cow::Borrowed("docx"),
+            target_mime: Cow::Borrowed(crate::core::mime::DOCX_MIME_TYPE),
         };
 
         assert_eq!(result.original_format, "doc");

data/vendor/kreuzberg/src/extraction/pptx/image_handling.rs

@@ -3,6 +3,8 @@
 //! This module handles image-related parsing from slide XML and
 //! detection of image formats from file data.
 
+use std::borrow::Cow;
+
 pub(super) fn html_escape(text: &str) -> String {
     text.replace('&', "&")
         .replace('<', "&lt;")
@@ -11,21 +13,21 @@ pub(super) fn html_escape(text: &str) -> String {
         .replace('\'', "&#x27;")
 }
 
-pub(super) fn detect_image_format(data: &[u8]) -> String {
+pub(super) fn detect_image_format(data: &[u8]) -> Cow<'static, str> {
     if data.starts_with(&[0xFF, 0xD8, 0xFF]) {
-        "jpeg".to_string()
+        Cow::Borrowed("jpeg")
     } else if data.starts_with(&[0x89, 0x50, 0x4E, 0x47]) {
-        "png".to_string()
+        Cow::Borrowed("png")
     } else if data.starts_with(b"GIF") {
-        "gif".to_string()
+        Cow::Borrowed("gif")
     } else if data.starts_with(b"BM") {
-        "bmp".to_string()
+        Cow::Borrowed("bmp")
     } else if data.starts_with(b"<svg") || data.starts_with(b"<?xml") {
-        "svg".to_string()
+        Cow::Borrowed("svg")
     } else if data.starts_with(b"II\x2A\x00") || data.starts_with(b"MM\x00\x2A") {
-        "tiff".to_string()
+        Cow::Borrowed("tiff")
     } else {
-        "unknown".to_string()
+        Cow::Borrowed("unknown")
     }
 }
 

data/vendor/kreuzberg/src/extraction/pptx/mod.rs

@@ -45,6 +45,8 @@ mod image_handling;
 mod metadata;
 mod parser;
 
+use bytes::Bytes;
+
 use crate::error::Result;
 use crate::types::{ExtractedImage, PptxExtractionResult};
 
@@ -117,8 +119,8 @@ pub fn extract_pptx_from_path(
                 let image_index = extracted_images.len();
 
                 extracted_images.push(ExtractedImage {
-                    data,
-                    format,
+                    data: Bytes::from(data),
+                    format, // Already a Cow<'static, str> from detect_image_format
                     image_index,
                     page_number: Some(slide.slide_number as usize),
                     width: None,
@@ -333,11 +335,13 @@ mod tests {
 <Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships">"#,
             );
             for (i, _) in slides.iter().enumerate() {
-                rels_xml.push_str(&format!(
+                use std::fmt::Write;
+                let _ = write!(
+                    rels_xml,
                     r#"<Relationship Id="rId{}" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/slide" Target="slides/slide{}.xml"/>"#,
                     i + 1,
                     i + 1
-                ));
+                );
             }
             rels_xml.push_str("</Relationships>");
             zip.start_file("ppt/_rels/presentation.xml.rels", options).unwrap();

data/vendor/kreuzberg/src/extraction/structured.rs

@@ -33,12 +33,13 @@
 use crate::error::{KreuzbergError, Result};
 use crate::text::utf8_validation;
 use serde::{Deserialize, Serialize};
+use std::borrow::Cow;
 use std::collections::HashMap;
 
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct StructuredDataResult {
     pub content: String,
-    pub format: String,
+    pub format: Cow<'static, str>,
     pub metadata: HashMap<String, String>,
     pub text_fields: Vec<String>,
 }
@@ -97,7 +98,7 @@ pub fn parse_json(data: &[u8], config: Option<JsonExtractionConfig>) -> Result<S
 
     Ok(StructuredDataResult {
         content,
-        format: "json".to_string(),
+        format: Cow::Borrowed("json"),
         metadata,
         text_fields,
     })
@@ -254,7 +255,7 @@ pub fn parse_yaml(data: &[u8]) -> Result<StructuredDataResult> {
 
     Ok(StructuredDataResult {
         content,
-        format: "yaml".to_string(),
+        format: Cow::Borrowed("yaml"),
         metadata,
         text_fields,
     })
@@ -326,7 +327,7 @@ pub fn parse_toml(data: &[u8]) -> Result<StructuredDataResult> {
 
     Ok(StructuredDataResult {
         content,
-        format: "toml".to_string(),
+        format: Cow::Borrowed("toml"),
         metadata,
         text_fields,
     })

data/vendor/kreuzberg/src/extraction/transform/content.rs

@@ -167,7 +167,7 @@ pub(super) fn process_images(
                 element_index: Some(elements.len()),
                 additional: {
                     let mut m = HashMap::new();
-                    m.insert("format".to_string(), image.format.clone());
+                    m.insert("format".to_string(), image.format.to_string());
                     if let Some(width) = image.width {
                         m.insert("width".to_string(), width.to_string());
                     }