kreuzberg 4.3.5-aarch64-linux

data/sig/kreuzberg.rbs

@@ -0,0 +1,1337 @@
+# Type signatures for Kreuzberg document intelligence framework
+
+module Kreuzberg
+  VERSION: String
+
+  # Error code constants
+  ERROR_CODE_SUCCESS: Integer
+  ERROR_CODE_GENERIC: Integer
+  ERROR_CODE_PANIC: Integer
+  ERROR_CODE_INVALID_ARGUMENT: Integer
+  ERROR_CODE_IO: Integer
+  ERROR_CODE_PARSING: Integer
+  ERROR_CODE_OCR: Integer
+  ERROR_CODE_MISSING_DEPENDENCY: Integer
+
+  # Semantic element type classification (T.type_alias)
+  type element_type = 'title' | 'narrative_text' | 'heading' | 'list_item' | 'table' | 'image' | 'page_break' | 'code_block' | 'block_quote' | 'footer' | 'header'
+
+  # Bounding box coordinates for element positioning (T::Struct from types.rb)
+  class BoundingBox     attr_reader x0: Float
+    attr_reader y0: Float
+    attr_reader x1: Float
+    attr_reader y1: Float
+
+    def initialize: (x0: Float, y0: Float, x1: Float, y1: Float) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # Metadata for a semantic element (T::Struct from types.rb)
+  class ElementMetadata     attr_reader page_number: Integer?
+    attr_reader filename: String?
+    attr_reader coordinates: BoundingBox?
+    attr_reader element_index: Integer?
+    attr_reader additional: Hash[String, String]
+
+    def initialize: (page_number: Integer?, filename: String?, coordinates: BoundingBox?, element_index: Integer?, additional: Hash[String, String]) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # Semantic element extracted from document (T::Struct from types.rb)
+  class Element     attr_reader element_id: String
+    attr_reader element_type: String
+    attr_reader text: String
+    attr_reader metadata: ElementMetadata
+
+    def initialize: (element_id: String, element_type: String, text: String, metadata: ElementMetadata) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # Header/Heading metadata (T::Struct from types.rb)
+  class HeaderMetadata     attr_reader level: Integer
+    attr_reader text: String
+    attr_reader id: String?
+    attr_reader depth: Integer
+    attr_reader html_offset: Integer
+
+    def initialize: (level: Integer, text: String, id: String?, depth: Integer, html_offset: Integer) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # Link metadata (T::Struct from types.rb)
+  class LinkMetadata     attr_reader href: String
+    attr_reader text: String
+    attr_reader title: String?
+    attr_reader link_type: String
+    attr_reader rel: Array[String]
+    attr_reader attributes: Hash[String, String]
+
+    def initialize: (href: String, text: String, title: String?, link_type: String, rel: Array[String], attributes: Hash[String, String]) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # Image metadata (T::Struct from types.rb)
+  class ImageMetadata     attr_reader src: String
+    attr_reader alt: String?
+    attr_reader title: String?
+    attr_reader dimensions: Array[Integer]?
+    attr_reader image_type: String
+    attr_reader attributes: Hash[String, String]
+
+    def initialize: (src: String, alt: String?, title: String?, dimensions: Array[Integer]?, image_type: String, attributes: Hash[String, String]) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # Structured data metadata (T::Struct from types.rb)
+  class StructuredData     attr_reader data_type: String
+    attr_reader raw_json: String
+    attr_reader schema_type: String?
+
+    def initialize: (data_type: String, raw_json: String, schema_type: String?) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # HTML metadata (T::Struct from types.rb)
+  class HtmlMetadata     attr_reader title: String?
+    attr_reader description: String?
+    attr_reader author: String?
+    attr_reader copyright: String?
+    attr_reader keywords: Array[String]
+    attr_reader canonical_url: String?
+    attr_reader language: String?
+    attr_reader text_direction: String?
+    attr_reader mime_type: String?
+    attr_reader charset: String?
+    attr_reader generator: String?
+    attr_reader viewport: String?
+    attr_reader theme_color: String?
+    attr_reader application_name: String?
+    attr_reader robots: String?
+    attr_reader open_graph: Hash[String, String]
+    attr_reader twitter_card: Hash[String, String]
+    attr_reader meta_tags: Hash[String, String]
+    attr_reader headers: Array[HeaderMetadata]
+    attr_reader links: Array[LinkMetadata]
+    attr_reader images: Array[ImageMetadata]
+    attr_reader structured_data: Array[StructuredData]
+
+    def initialize: (
+      title: String?,
+      description: String?,
+      author: String?,
+      copyright: String?,
+      keywords: Array[String],
+      canonical_url: String?,
+      language: String?,
+      text_direction: String?,
+      mime_type: String?,
+      charset: String?,
+      generator: String?,
+      viewport: String?,
+      theme_color: String?,
+      application_name: String?,
+      robots: String?,
+      open_graph: Hash[String, String],
+      twitter_card: Hash[String, String],
+      meta_tags: Hash[String, String],
+      headers: Array[HeaderMetadata],
+      links: Array[LinkMetadata],
+      images: Array[ImageMetadata],
+      structured_data: Array[StructuredData]
+    ) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # Extracted keyword with relevance metadata (T::Struct from types.rb)
+  class ExtractedKeyword     attr_reader text: String
+    attr_reader score: Float
+    attr_reader algorithm: String
+    attr_reader positions: Array[Integer]?
+
+    def initialize: (text: String, score: Float, algorithm: String, ?positions: Array[Integer]?) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # Processing warning from a pipeline stage (T::Struct from types.rb)
+  class ProcessingWarning     attr_reader source: String
+    attr_reader message: String
+
+    def initialize: (source: String, message: String) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # Bounding box for document node positioning (T::Struct from types.rb)
+  class DocumentBoundingBox     attr_reader x0: Float
+    attr_reader y0: Float
+    attr_reader x1: Float
+    attr_reader y1: Float
+
+    def initialize: (x0: Float, y0: Float, x1: Float, y1: Float) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # Annotation for a document node (T::Struct from types.rb)
+  class DocumentAnnotation     attr_reader key: String
+    attr_reader value: String
+
+    def initialize: (key: String, value: String) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # Single node in the document structure tree (T::Struct from types.rb)
+  class DocumentNode     attr_reader id: String
+    attr_reader content: String
+    attr_reader parent: Integer?
+    attr_reader children: Array[Integer]
+    attr_reader content_layer: String
+    attr_reader page: Integer?
+    attr_reader page_end: Integer?
+    attr_reader bbox: DocumentBoundingBox?
+    attr_reader annotations: Array[DocumentAnnotation]
+
+    def initialize: (
+      id: String,
+      content: String,
+      parent: Integer?,
+      children: Array[Integer],
+      content_layer: String,
+      page: Integer?,
+      page_end: Integer?,
+      bbox: DocumentBoundingBox?,
+      annotations: Array[DocumentAnnotation]
+    ) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # Structured document representation (T::Struct from types.rb)
+  class DocumentStructure     attr_reader nodes: Array[DocumentNode]
+
+    def initialize: (nodes: Array[DocumentNode]) -> void
+    def serialize: () -> Hash[Symbol, untyped]
+  end
+
+  # Config namespace (defined in lib/kreuzberg/config.rb)
+  module Config
+    class OCR
+      attr_reader backend: String
+      attr_reader language: String
+      attr_reader tesseract_config: Tesseract?
+      attr_reader paddle_ocr_config: PaddleOcr?
+      attr_reader element_config: OcrElementConfig?
+
+      def initialize: (?backend: String, ?language: String, ?tesseract_config: (Tesseract | Hash[Symbol, untyped])?, ?paddle_ocr_config: (PaddleOcr | Hash[Symbol, untyped])?, ?element_config: (OcrElementConfig | Hash[Symbol, untyped])?) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class Tesseract
+      attr_reader options: Hash[Symbol, untyped]
+
+      def initialize: (**untyped options) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class PaddleOcr
+      attr_reader language: String?
+      attr_reader cache_dir: String?
+      attr_reader use_angle_cls: bool?
+      attr_reader enable_table_detection: bool?
+      attr_reader det_db_thresh: Float?
+      attr_reader det_db_box_thresh: Float?
+      attr_reader det_db_unclip_ratio: Float?
+      attr_reader det_limit_side_len: Integer?
+      attr_reader rec_batch_num: Integer?
+      def initialize: (?language: String?, ?cache_dir: String?, ?use_angle_cls: bool?, ?enable_table_detection: bool?, ?det_db_thresh: Float?, ?det_db_box_thresh: Float?, ?det_db_unclip_ratio: Float?, ?det_limit_side_len: Integer?, ?rec_batch_num: Integer?) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class OcrElementConfig
+      attr_reader include_elements: bool
+      attr_reader min_level: String?
+      attr_reader min_confidence: Float?
+      attr_reader build_hierarchy: bool
+      def initialize: (?include_elements: bool, ?min_level: String?, ?min_confidence: Float?, ?build_hierarchy: bool) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class Chunking
+      attr_reader max_chars: Integer
+      attr_reader max_overlap: Integer
+      attr_reader preset: String?
+      attr_reader embedding: Embedding?
+      attr_reader enabled: bool?
+
+      def initialize: (
+        ?max_chars: Integer?,
+        ?max_overlap: Integer?,
+        ?preset: String?,
+        ?embedding: (Embedding | Hash[Symbol, untyped])?,
+        ?chunk_size: Integer?,
+        ?chunk_overlap: Integer?,
+        ?enabled: bool
+      ) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class Embedding
+      attr_reader model: Hash[Symbol, untyped]
+      attr_reader normalize: bool?
+      attr_reader batch_size: Integer?
+      attr_reader show_download_progress: bool?
+      attr_reader cache_dir: String?
+
+      def initialize: (
+        ?model: Hash[Symbol, untyped],
+        ?normalize: bool?,
+        ?batch_size: Integer?,
+        ?show_download_progress: bool?,
+        ?cache_dir: String?
+      ) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class LanguageDetection
+      attr_reader enabled: bool
+      attr_reader min_confidence: Float
+      attr_reader detect_multiple: bool
+
+      def initialize: (?enabled: bool, ?min_confidence: Float, ?detect_multiple: bool) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class FontConfig
+      attr_accessor enabled: bool
+      attr_accessor custom_font_dirs: Array[String]?
+
+      def initialize: (?enabled: bool, ?custom_font_dirs: Array[String]?) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class Hierarchy
+      attr_reader enabled: bool
+      attr_reader k_clusters: Integer
+      attr_reader include_bbox: bool
+      attr_reader ocr_coverage_threshold: Float?
+
+      def initialize: (?enabled: bool, ?k_clusters: Integer, ?include_bbox: bool, ?ocr_coverage_threshold: Float?) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+      def self.from_h: (Hash[Symbol, untyped]?) -> Hierarchy?
+    end
+
+    class PDF
+      attr_reader extract_images: bool
+      attr_reader passwords: Array[String]?
+      attr_reader extract_metadata: bool
+      attr_reader font_config: FontConfig?
+      attr_reader hierarchy: Hierarchy?
+
+      def initialize: (?extract_images: bool, ?passwords: (Array[String] | String)?, ?extract_metadata: bool, ?font_config: (FontConfig | Hash[Symbol, untyped])?, ?hierarchy: (Hierarchy | Hash[Symbol, untyped])?) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class ImageExtraction
+      attr_reader extract_images: bool
+      attr_reader target_dpi: Integer
+      attr_reader max_image_dimension: Integer
+      attr_reader auto_adjust_dpi: bool
+      attr_reader min_dpi: Integer
+      attr_reader max_dpi: Integer
+
+      def initialize: (
+        ?extract_images: bool,
+        ?target_dpi: Integer,
+        ?max_image_dimension: Integer,
+        ?auto_adjust_dpi: bool,
+        ?min_dpi: Integer,
+        ?max_dpi: Integer
+      ) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class ImagePreprocessing
+      attr_reader target_dpi: Integer
+      attr_reader auto_rotate: bool
+      attr_reader deskew: bool
+      attr_reader denoise: bool
+      attr_reader contrast_enhance: bool
+      attr_reader binarization_method: String
+      attr_reader invert_colors: bool
+
+      def initialize: (
+        ?target_dpi: Integer,
+        ?auto_rotate: bool,
+        ?deskew: bool,
+        ?denoise: bool,
+        ?contrast_enhance: bool,
+        ?binarization_method: String,
+        ?invert_colors: bool
+      ) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class TokenReduction
+      attr_reader mode: String
+      attr_reader preserve_important_words: bool
+
+      def initialize: (?mode: String, ?preserve_important_words: bool) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class PostProcessor
+      attr_reader enabled: bool
+      attr_reader enabled_processors: Array[String]?
+      attr_reader disabled_processors: Array[String]?
+
+      def initialize: (?enabled: bool, ?enabled_processors: Array[String]?, ?disabled_processors: Array[String]?) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class HtmlPreprocessing
+      attr_reader enabled: bool?
+      attr_reader preset: Symbol?
+      attr_reader remove_navigation: bool?
+      attr_reader remove_forms: bool?
+
+      def initialize: (?enabled: bool?, ?preset: Symbol?, ?remove_navigation: bool?, ?remove_forms: bool?) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class HtmlOptions
+      attr_reader options: Hash[Symbol, untyped]
+
+      def initialize: (**untyped options) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class KeywordYakeParams
+      attr_reader window_size: Integer
+
+      def initialize: (?window_size: Integer) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class KeywordRakeParams
+      attr_reader min_word_length: Integer
+      attr_reader max_words_per_phrase: Integer
+
+      def initialize: (?min_word_length: Integer, ?max_words_per_phrase: Integer) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class Keywords
+      attr_reader algorithm: String?
+      attr_reader max_keywords: Integer?
+      attr_reader min_score: Float?
+      attr_reader ngram_range: Array[Integer]?
+      attr_reader language: String?
+      attr_reader yake_params: KeywordYakeParams?
+      attr_reader rake_params: KeywordRakeParams?
+
+      def initialize: (
+        ?algorithm: (Symbol | String)?,
+        ?max_keywords: Integer?,
+        ?min_score: Float?,
+        ?ngram_range: Array[Integer]?,
+        ?language: (Symbol | String)?,
+        ?yake_params: (KeywordYakeParams | Hash[Symbol, untyped])?,
+        ?rake_params: (KeywordRakeParams | Hash[Symbol, untyped])?
+      ) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class PageConfig
+      attr_reader extract_pages: bool
+      attr_reader insert_page_markers: bool
+      attr_reader marker_format: String
+
+      def initialize: (?extract_pages: bool, ?insert_page_markers: bool, ?marker_format: String) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    class Extraction
+      attr_reader use_cache: bool
+      attr_reader enable_quality_processing: bool
+      attr_reader force_ocr: bool
+      attr_reader include_document_structure: bool
+      attr_reader ocr: OCR?
+      attr_reader chunking: Chunking?
+      attr_reader language_detection: LanguageDetection?
+      attr_reader pdf_options: PDF?
+      attr_reader images: ImageExtraction?
+      attr_reader postprocessor: PostProcessor?
+      attr_reader token_reduction: TokenReduction?
+      attr_reader keywords: Keywords?
+      attr_reader html_options: HtmlOptions?
+      attr_reader pages: PageConfig?
+      attr_reader max_concurrent_extractions: Integer?
+      attr_reader output_format: String?
+      attr_reader result_format: String?
+      attr_reader security_limits: Hash[String, Integer]?
+
+      alias image_extraction images
+
+      def self.from_file: (String path) -> Extraction
+      def self.discover: () -> Extraction?
+      def initialize: (
+        ?use_cache: bool,
+        ?enable_quality_processing: bool,
+        ?force_ocr: bool,
+        ?include_document_structure: bool,
+        ?ocr: (OCR | Hash[Symbol, untyped])?,
+        ?chunking: (Chunking | Hash[Symbol, untyped])?,
+        ?language_detection: (LanguageDetection | Hash[Symbol, untyped])?,
+        ?pdf_options: (PDF | Hash[Symbol, untyped])?,
+        ?image_extraction: (ImageExtraction | Hash[Symbol, untyped])?,
+        ?postprocessor: (PostProcessor | Hash[Symbol, untyped])?,
+        ?token_reduction: (TokenReduction | Hash[Symbol, untyped])?,
+        ?keywords: (Keywords | Hash[Symbol, untyped])?,
+        ?html_options: (HtmlOptions | Hash[Symbol, untyped])?,
+        ?pages: (PageConfig | Hash[Symbol, untyped])?,
+        ?max_concurrent_extractions: Integer?,
+        ?output_format: String?,
+        ?result_format: String?
+      ) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+      def to_json: (*untyped) -> String
+      def get_field: (String | Symbol field_name) -> untyped
+      def merge: (Extraction | Hash[Symbol, untyped] other) -> Extraction
+      def merge!: (Extraction | Hash[Symbol, untyped] other) -> self
+      def []: (Symbol | String key) -> untyped
+      def []=: (Symbol | String key, untyped value) -> untyped
+      def output_format=: (String? value) -> String?
+      def result_format=: (String? value) -> String?
+
+      private
+
+      def normalize_config: [T] (T | Hash[Symbol, untyped] | nil value, Class klass) -> T?
+      def extract_from_hash: (Hash[Symbol, untyped]? hash, Hash[Symbol, untyped] defaults) -> Hash[Symbol, untyped]
+      def assign_attributes: (Hash[Symbol, untyped] params) -> void
+      def validate_output_format: (untyped value) -> String?
+      def validate_result_format: (untyped value) -> String?
+      def update_from_merged: (Extraction merged) -> void
+    end
+
+  end
+
+  # Alias for Config::Extraction (for API consistency with other language bindings)
+  ExtractionConfig: singleton(Config::Extraction)
+
+  # Alias for Config::PageConfig (for API consistency with other language bindings)
+  PageConfig: singleton(Config::PageConfig)
+
+  # Keyword algorithm constants
+  module KeywordAlgorithm
+    YAKE: Symbol
+    RAKE: Symbol
+  end
+
+  # Extraction result type
+  type extraction_result_hash = {
+    content: String,
+    mime_type: String,
+    metadata_json: String,
+    metadata: Hash[String, untyped],
+    tables: Array[table_hash]?,
+    detected_languages: Array[String]?,
+    chunks: Array[chunk_hash]?,
+    images: Array[image_hash]?,
+    pages: Array[page_content_hash]?,
+    elements: Array[element_hash]?,
+    ocr_elements: Array[ocr_element_hash]?,
+    djot_content: djot_content_hash?,
+    document: document_structure_hash?,
+    extracted_keywords: Array[extracted_keyword_hash]?,
+    quality_score: Float?,
+    processing_warnings: Array[processing_warning_hash]?
+  }
+
+  type extracted_keyword_hash = {
+    text: String,
+    score: Float,
+    algorithm: String,
+    positions: Array[Integer]?
+  }
+
+  type processing_warning_hash = {
+    source: String,
+    message: String
+  }
+
+  type page_content_hash = {
+    page_number: Integer,
+    content: String,
+    tables: Array[table_hash],
+    images: Array[image_hash],
+    is_blank: bool?
+  }
+
+  type djot_content_hash = {
+    plain_text: String,
+    blocks: Array[formatted_block_hash],
+    metadata_json: String,
+    tables: Array[table_hash],
+    images: Array[djot_image_hash],
+    links: Array[djot_link_hash],
+    footnotes: Array[footnote_hash],
+    attributes: Hash[String, attributes_hash]?
+  }
+
+  type formatted_block_hash = {
+    block_type: String,
+    level: Integer?,
+    content: String?,
+    children: Array[formatted_block_hash]?,
+    attributes: attributes_hash?
+  }
+
+  type djot_image_hash = {
+    url: String,
+    alt: String?,
+    title: String?,
+    attributes: attributes_hash?
+  }
+
+  type djot_link_hash = {
+    url: String,
+    text: String,
+    title: String?,
+    link_type: String?
+  }
+
+  type footnote_hash = {
+    label: String,
+    content: String
+  }
+
+  type attributes_hash = Hash[String, String | Integer | bool | Array[String] | nil]
+
+  type document_bounding_box_hash = {
+    x0: Float,
+    y0: Float,
+    x1: Float,
+    y1: Float
+  }
+
+  type document_annotation_hash = {
+    start: Integer,
+    end: Integer,
+    kind: {
+      annotation_type: String,
+      url: String?,
+      title: String?
+    }
+  }
+
+  type document_node_hash = {
+    id: String,
+    content: Hash[untyped, untyped],
+    parent: Integer?,
+    children: Array[Integer],
+    content_layer: String,
+    page: Integer?,
+    page_end: Integer?,
+    bbox: document_bounding_box_hash?,
+    annotations: Array[document_annotation_hash]
+  }
+
+  type document_structure_hash = {
+    nodes: Array[document_node_hash]
+  }
+
+  type element_hash = {
+    element_id: String,
+    element_type: String,
+    text: String,
+    metadata: Hash[String, untyped]?
+  }
+
+  type ocr_confidence_hash = {
+    detection: Float?,
+    recognition: Float?
+  }
+
+  type ocr_rotation_hash = {
+    angle_degrees: Float?,
+    confidence: Float?
+  }
+
+  type ocr_bounding_geometry_hash = {
+    type: String,
+    left: Float?,
+    top: Float?,
+    width: Float?,
+    height: Float?,
+    points: Array[Array[Float]]?
+  }
+
+  type ocr_element_hash = {
+    text: String,
+    geometry: ocr_bounding_geometry_hash?,
+    confidence: ocr_confidence_hash?,
+    level: String?,
+    rotation: ocr_rotation_hash?,
+    page_number: Integer?,
+    parent_id: String?,
+    backend_metadata: Hash[String, untyped]?
+  }
+
+  type table_hash = {
+    cells: Array[Array[String]],
+    markdown: String,
+    page_number: Integer,
+    bounding_box: { x0: Float, y0: Float, x1: Float, y1: Float }?
+  }
+
+  type chunk_hash = {
+    content: String,
+    byte_start: Integer,
+    byte_end: Integer,
+    token_count: Integer?,
+    chunk_index: Integer,
+    total_chunks: Integer,
+    first_page: Integer?,
+    last_page: Integer?,
+    embedding: Array[Float]?
+  }
+
+  type image_hash = {
+    data: String,
+    format: String,
+    image_index: Integer,
+    page_number: Integer?,
+    width: Integer?,
+    height: Integer?,
+    colorspace: String?,
+    bits_per_component: Integer?,
+    is_mask: bool,
+    description: String?,
+    bounding_box: { x0: Float, y0: Float, x1: Float, y1: Float }?,
+    ocr_result: extraction_result_hash?
+  }
+
+  # Metadata hash type - represents the parsed JSON metadata from extraction results.
+  # Fields correspond to the Rust Metadata struct (serialized as JSON and parsed back).
+  #
+  # Common fields:
+  #   "title"                  => String?      - Document title
+  #   "subject"                => String?      - Document subject or description
+  #   "authors"                => Array[String]? - Primary author(s)
+  #   "keywords"               => Array[String]? - Keywords/tags
+  #   "language"               => String?      - Primary language (ISO 639 code)
+  #   "created_at"             => String?      - Creation timestamp (ISO 8601)
+  #   "modified_at"            => String?      - Last modification timestamp (ISO 8601)
+  #   "created_by"             => String?      - User who created the document
+  #   "modified_by"            => String?      - User who last modified the document
+  #   "pages"                  => Hash?        - Page/slide/sheet structure with boundaries
+  #   "format_type"            => String?      - Format discriminator (pdf, docx, excel, etc.)
+  #   "image_preprocessing"    => Hash?        - Image preprocessing metadata
+  #   "json_schema"            => untyped?     - JSON schema for structured data extraction
+  #   "error"                  => Hash?        - Error metadata (for batch operations)
+  #   "extraction_duration_ms" => Integer?     - Extraction duration in milliseconds
+  #   "category"               => String?      - Document category
+  #   "tags"                   => Array[String]? - Document tags
+  #   "document_version"       => String?      - Document version string
+  #   "abstract_text"          => String?      - Abstract or summary text
+  #   "output_format"          => String?      - Output format identifier
+  type metadata_hash = Hash[String, untyped]
+
+  type config_hash = Hash[Symbol, untyped]
+  type config_input = config_hash | _ToH
+
+  interface _ToH
+    def to_h: () -> config_hash
+  end
+
+  # Extraction result wrapper
+  class Result
+    # Table structure (Struct from result.rb)
+    class Table
+      attr_reader cells: Array[Array[String]]
+      attr_reader markdown: String
+      attr_reader page_number: Integer
+      attr_reader bounding_box: BoundingBox?
+
+      def initialize: (cells: Array[Array[String]], markdown: String, page_number: Integer, bounding_box: BoundingBox?) -> void
+      def to_h: () -> table_hash
+    end
+
+    # Text chunk (Struct from result.rb)
+    class Chunk
+      attr_reader content: String
+      attr_reader byte_start: Integer
+      attr_reader byte_end: Integer
+      attr_reader token_count: Integer?
+      attr_reader chunk_index: Integer
+      attr_reader total_chunks: Integer
+      attr_reader first_page: Integer?
+      attr_reader last_page: Integer?
+      attr_reader embedding: Array[Float]?
+
+      def initialize: (
+        content: String,
+        byte_start: Integer,
+        byte_end: Integer,
+        token_count: Integer?,
+        chunk_index: Integer,
+        total_chunks: Integer,
+        first_page: Integer?,
+        last_page: Integer?,
+        embedding: Array[Float]?
+      ) -> void
+      def to_h: () -> chunk_hash
+    end
+
+    # Extracted image (Struct from result.rb)
+    class Image
+      attr_reader data: String
+      attr_reader format: String
+      attr_reader image_index: Integer
+      attr_reader page_number: Integer?
+      attr_reader width: Integer?
+      attr_reader height: Integer?
+      attr_reader colorspace: String?
+      attr_reader bits_per_component: Integer?
+      attr_reader is_mask: bool
+      attr_reader description: String?
+      attr_reader bounding_box: BoundingBox?
+      attr_reader ocr_result: Result?
+
+      def initialize: (
+        data: String,
+        format: String,
+        image_index: Integer,
+        page_number: Integer?,
+        width: Integer?,
+        height: Integer?,
+        colorspace: String?,
+        bits_per_component: Integer?,
+        is_mask: bool,
+        description: String?,
+        bounding_box: BoundingBox?,
+        ocr_result: Result?
+      ) -> void
+      def to_h: () -> image_hash
+    end
+
+    # Page content with text and extracted elements (Struct from result.rb)
+    class PageContent
+      attr_reader page_number: Integer
+      attr_reader content: String
+      attr_reader tables: Array[Table]
+      attr_reader images: Array[Image]?
+      attr_reader hierarchy: PageHierarchy?
+      attr_reader is_blank: bool?
+
+      def initialize: (page_number: Integer, content: String, tables: Array[Table], images: Array[Image]?, hierarchy: PageHierarchy?, is_blank: bool?) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    # Hierarchical block element for page hierarchy (Struct from result.rb)
+    class HierarchicalBlock
+      attr_reader text: String
+      attr_reader font_size: Float?
+      attr_reader level: String?
+      attr_reader bbox: Array[Float]?
+
+      def initialize: (text: String, font_size: Float?, level: String?, bbox: Array[Float]?) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    # Page hierarchy information (Struct from result.rb)
+    class PageHierarchy
+      attr_reader block_count: Integer
+      attr_reader blocks: Array[HierarchicalBlock]
+
+      def initialize: (block_count: Integer, blocks: Array[HierarchicalBlock]) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    # Element bounding box coordinates (Struct from result.rb)
+    class ElementBoundingBox
+      attr_reader x0: Float
+      attr_reader y0: Float
+      attr_reader x1: Float
+      attr_reader y1: Float
+
+      def initialize: (x0: Float, y0: Float, x1: Float, y1: Float) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    # Element metadata (Struct from result.rb)
+    class ElementMetadataStruct
+      attr_reader page_number: Integer?
+      attr_reader filename: String?
+      attr_reader coordinates: ElementBoundingBox?
+      attr_reader element_index: Integer?
+      attr_reader additional: Hash[String, String]
+
+      def initialize: (page_number: Integer?, filename: String?, coordinates: ElementBoundingBox?, element_index: Integer?, additional: Hash[String, String]) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    # Structured document element (Struct from result.rb)
+    class ElementStruct
+      attr_reader element_id: String
+      attr_reader element_type: String
+      attr_reader text: String
+      attr_reader metadata: ElementMetadataStruct
+
+      def initialize: (element_id: String, element_type: String, text: String, metadata: ElementMetadataStruct) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+    end
+
+    # OCR bounding geometry (class from result.rb)
+    class OcrBoundingGeometry
+      attr_reader type: String
+      attr_reader left: Float?
+      attr_reader top: Float?
+      attr_reader width: Float?
+      attr_reader height: Float?
+      attr_reader points: Array[Array[Float]]?
+      def initialize: (type: String, ?left: Float?, ?top: Float?, ?width: Float?, ?height: Float?, ?points: Array[Array[Float]]?) -> void
+      def to_h: () -> ocr_bounding_geometry_hash
+    end
+
+    # OCR confidence scores (class from result.rb)
+    class OcrConfidence
+      attr_reader detection: Float?
+      attr_reader recognition: Float?
+      def initialize: (?detection: Float?, ?recognition: Float?) -> void
+      def to_h: () -> ocr_confidence_hash
+    end
+
+    # OCR rotation information (class from result.rb)
+    class OcrRotation
+      attr_reader angle_degrees: Float?
+      attr_reader confidence: Float?
+      def initialize: (?angle_degrees: Float?, ?confidence: Float?) -> void
+      def to_h: () -> ocr_rotation_hash
+    end
+
+    # OCR text element with geometry and metadata (class from result.rb)
+    class OcrElement
+      attr_reader text: String
+      attr_reader geometry: OcrBoundingGeometry?
+      attr_reader confidence: OcrConfidence?
+      attr_reader level: String?
+      attr_reader rotation: OcrRotation?
+      attr_reader page_number: Integer?
+      attr_reader parent_id: String?
+      attr_reader backend_metadata: Hash[String, untyped]?
+      def initialize: (text: String, ?geometry: OcrBoundingGeometry?, ?confidence: OcrConfidence?, ?level: String?, ?rotation: OcrRotation?, ?page_number: Integer?, ?parent_id: String?, ?backend_metadata: Hash[String, untyped]?) -> void
+      def to_h: () -> ocr_element_hash
+    end
+
+    # Structured Djot document representation (class from djot_content.rb)
+    class DjotContent
+      attr_reader plain_text: String
+      attr_reader blocks: Array[DjotContent::FormattedBlock]
+      attr_reader metadata_json: String
+      attr_reader tables: Array[untyped]
+      attr_reader images: Array[DjotContent::DjotImage]
+      attr_reader links: Array[DjotContent::DjotLink]
+      attr_reader footnotes: Array[DjotContent::Footnote]
+      attr_reader attributes: Hash[String, untyped]?
+
+      def initialize: (untyped hash) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+      def metadata: () -> Hash[untyped, untyped]
+
+      private
+
+      def parse_metadata: (String metadata_json) -> Hash[untyped, untyped]
+      def parse_blocks: (Array[untyped] blocks_data) -> Array[FormattedBlock]
+      def parse_images: (Array[untyped] images_data) -> Array[DjotImage]
+      def parse_links: (Array[untyped] links_data) -> Array[DjotLink]
+      def parse_footnotes: (Array[untyped] footnotes_data) -> Array[Footnote]
+
+      class FormattedBlock
+        attr_reader block_type: String
+        attr_reader level: Integer?
+        attr_reader content: String?
+        attr_reader children: Array[FormattedBlock]?
+        attr_reader attributes: Hash[String, untyped]?
+
+        def initialize: (?untyped hash_or_type, ?children: untyped, ?attributes: untyped, ?content: untyped, ?level: untyped, ?block_type: untyped) -> void
+        def to_h: () -> Hash[Symbol, untyped]
+      end
+
+      class DjotImage
+        attr_reader url: String
+        attr_reader alt: String?
+        attr_reader title: String?
+        attr_reader width: Integer?
+        attr_reader height: Integer?
+
+        def initialize: (?untyped hash_or_url, ?alt: untyped, ?title: untyped, ?width: untyped, ?height: untyped, ?url: untyped, ?src: untyped) -> void
+        def src: () -> String
+        def to_h: () -> Hash[Symbol, untyped]
+      end
+
+      class DjotLink
+        attr_reader url: String
+        attr_reader text: String?
+        attr_reader title: String?
+        attr_reader link_type: String?
+
+        def initialize: (?untyped hash_or_url, ?text: untyped, ?title: untyped, ?url: untyped, ?href: untyped, ?link_type: untyped) -> void
+        def href: () -> String
+        def to_h: () -> Hash[Symbol, untyped]
+      end
+
+      class Footnote
+        attr_reader label: String
+        attr_reader content: String
+
+        def initialize: (label: String, content: String) -> void
+        def to_h: () -> footnote_hash
+      end
+    end
+
+    # Structured document representation (class from document_structure.rb)
+    # Note: This is Result::DocumentStructure, distinct from Kreuzberg::DocumentStructure (T::Struct)
+    class DocumentStructure
+      attr_reader nodes: Array[DocumentNode]
+
+      def initialize: (Hash[String | Symbol, untyped] hash) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+
+      private
+
+      def parse_nodes: (Array[untyped]? nodes_data) -> Array[DocumentNode]
+    end
+
+    # Single node in the document structure tree (class from document_structure.rb)
+    # Note: This is Result::DocumentNode, distinct from Kreuzberg::DocumentNode (T::Struct)
+    class DocumentNode
+      attr_reader id: String
+      attr_reader content: Hash[untyped, untyped]
+      attr_reader parent: Integer?
+      attr_reader children: Array[Integer]
+      attr_reader content_layer: String
+      attr_reader page: Integer?
+      attr_reader page_end: Integer?
+      attr_reader bbox: DocumentBoundingBox?
+      attr_reader annotations: Array[DocumentAnnotation]
+
+      def initialize: (Hash[String | Symbol, untyped] hash) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+
+      private
+
+      def assign_core_fields: (Hash[String | Symbol, untyped] hash) -> void
+      def assign_tree_fields: (Hash[String | Symbol, untyped] hash) -> void
+      def assign_metadata_fields: (Hash[String | Symbol, untyped] hash) -> void
+      def parse_children: (Array[untyped]? children_data) -> Array[Integer]
+      def extract_child_index: (untyped child) -> Integer
+      def parse_bbox: (Hash[String | Symbol, untyped]? bbox_data) -> DocumentBoundingBox?
+      def parse_annotations: (Array[untyped]? annotations_data) -> Array[DocumentAnnotation]
+    end
+
+    # Bounding box for document node positioning (class from document_structure.rb)
+    # Note: This is Result::DocumentBoundingBox, distinct from Kreuzberg::DocumentBoundingBox (T::Struct)
+    class DocumentBoundingBox
+      attr_reader x0: Float?
+      attr_reader y0: Float?
+      attr_reader x1: Float?
+      attr_reader y1: Float?
+
+      def initialize: (Hash[String | Symbol, untyped] hash) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+
+      private
+
+      def extract_float: (Hash[String | Symbol, untyped] hash, String key) -> Float?
+    end
+
+    # Annotation for a document node (class from document_structure.rb)
+    # Note: This is Result::DocumentAnnotation, distinct from Kreuzberg::DocumentAnnotation (T::Struct)
+    class DocumentAnnotation
+      attr_reader start: Integer
+      attr_reader end_offset: Integer
+      attr_reader annotation_type: String
+      attr_reader url: String?
+      attr_reader title: String?
+
+      def initialize: (Hash[String | Symbol, untyped] hash) -> void
+      def to_h: () -> Hash[Symbol, untyped]
+
+      private
+
+      def parse_kind: (Hash[String | Symbol, untyped]? kind_hash) -> void
+    end
+
+    attr_reader content: String
+    attr_reader mime_type: String
+    attr_reader metadata: metadata_hash
+    attr_reader metadata_json: String
+    attr_reader tables: Array[Table]
+    attr_reader detected_languages: Array[String]?
+    attr_reader chunks: Array[Chunk]?
+    attr_reader images: Array[Image]?
+    attr_reader pages: Array[PageContent]?
+    attr_reader elements: Array[ElementStruct]?
+    attr_reader ocr_elements: Array[OcrElement]?
+    attr_reader djot_content: DjotContent?
+    attr_reader document: DocumentStructure?
+    attr_reader extracted_keywords: Array[ExtractedKeyword]?
+    attr_reader quality_score: Float?
+    attr_reader processing_warnings: Array[ProcessingWarning]?
+
+    def initialize: (extraction_result_hash hash) -> void
+    def to_h: () -> Hash[Symbol, untyped]
+    def to_json: (*untyped) -> String
+
+    def page_count: () -> Integer
+    def chunk_count: () -> Integer
+    def detected_language: () -> String?
+    def metadata_field: (String | Symbol name) -> untyped
+
+    private
+
+    def parse_metadata: (String metadata_json) -> metadata_hash
+    def parse_tables: (Array[table_hash]? tables_data) -> Array[Table]
+    def parse_detected_languages: (Array[String]? langs_data) -> Array[String]?
+    def parse_chunks: (Array[chunk_hash]? chunks_data) -> Array[Chunk]?
+    def parse_images: (Array[image_hash]? images_data) -> Array[Image]?
+    def parse_pages: (Array[page_content_hash]? pages_data) -> Array[PageContent]?
+    def parse_elements: (Array[untyped]? elements_data) -> Array[ElementStruct]?
+    def parse_element: (Hash[String, untyped] element_hash) -> ElementStruct
+    def parse_element_coordinates: (Hash[String, untyped]? coordinates_data) -> ElementBoundingBox?
+    def parse_ocr_elements: (Array[ocr_element_hash]? ocr_elements_data) -> Array[OcrElement]?
+    def parse_page_hierarchy: (Hash[String, untyped]? hierarchy_data) -> PageHierarchy?
+    def parse_djot_content: (Hash[String, untyped]? djot_data) -> DjotContent?
+    def parse_document_structure: (Hash[String, untyped]? document_data) -> DocumentStructure?
+    def parse_extracted_keywords: (Array[extracted_keyword_hash]? keywords_data) -> Array[ExtractedKeyword]?
+    def parse_processing_warnings: (Array[processing_warning_hash]? warnings_data) -> Array[ProcessingWarning]
+    def get_value: (Hash[String | Symbol, untyped] hash, String key, ?untyped default) -> untyped
+    def serialize_tables: () -> Array[table_hash]
+    def serialize_chunks: () -> Array[chunk_hash]?
+    def serialize_images: () -> Array[image_hash]?
+    def serialize_pages: () -> Array[Hash[Symbol, untyped]]?
+    def serialize_elements: () -> Array[Hash[Symbol, untyped]]?
+    def serialize_ocr_elements: () -> Array[ocr_element_hash]?
+    def parse_ocr_geometry: (Hash[String, untyped]? data) -> OcrBoundingGeometry?
+    def parse_ocr_confidence: (Hash[String, untyped]? data) -> OcrConfidence?
+    def parse_ocr_rotation: (Hash[String, untyped]? data) -> OcrRotation?
+  end
+
+  # Module methods (extraction API)
+  def self.extract_file_sync: (
+    path: String | Pathname,
+    ?mime_type: String?,
+    ?config: config_input?
+  ) -> Result
+
+  def self.extract_bytes_sync: (
+    data: String,
+    mime_type: String,
+    ?config: config_input?
+  ) -> Result
+
+  def self.batch_extract_files_sync: (
+    paths: Array[String | Pathname],
+    ?config: config_input?
+  ) -> Array[Result]
+
+  def self.batch_extract_bytes_sync: (
+    data_array: Array[String],
+    mime_types: Array[String],
+    ?config: config_input?
+  ) -> Array[Result]
+
+  def self.extract_file: (
+    path: String | Pathname,
+    ?mime_type: String?,
+    ?config: config_input?
+  ) -> Result
+
+  def self.extract_bytes: (
+    data: String,
+    mime_type: String,
+    ?config: config_input?
+  ) -> Result
+
+  def self.batch_extract_files: (
+    paths: Array[String | Pathname],
+    ?config: config_input?
+  ) -> Array[Result]
+
+  def self.batch_extract_bytes: (
+    data_array: Array[String],
+    mime_types: Array[String],
+    ?config: config_input?
+  ) -> Array[Result]
+
+  # Cache API
+  def self.clear_cache: () -> void
+  def self.cache_stats: () -> Hash[Symbol | String, Integer]
+
+  # Config loading (native methods)
+  def self._config_from_file_native: (String path) -> Hash[Symbol, untyped]
+  def self._config_discover_native: () -> Hash[Symbol, untyped]?
+
+  # Error introspection (native methods)
+  def self._last_error_code_native: () -> Integer
+  def self._last_panic_context_json_native: () -> String?
+  def self._get_error_details_native: () -> Hash[String, untyped]
+  def self._classify_error_native: (String message) -> Hash[String, untyped]
+  def self._error_code_name_native: (Integer code) -> String
+  def self._error_code_description_native: (Integer code) -> String
+
+  # MIME type detection
+  def self.detect_mime_type: (String data) -> String
+  def self.detect_mime_type_from_path: (String path) -> String
+  def self.get_extensions_for_mime: (String mime_type) -> Array[String]
+  def self.validate_mime_type: (String mime_type) -> String
+
+  # Validation native methods
+  def self._validate_binarization_method_native: (String method) -> Integer
+  def self._validate_ocr_backend_native: (String backend) -> Integer
+  def self._validate_language_code_native: (String code) -> Integer
+  def self._validate_token_reduction_level_native: (String level) -> Integer
+  def self._validate_tesseract_psm_native: (Integer psm) -> Integer
+  def self._validate_tesseract_oem_native: (Integer oem) -> Integer
+  def self._validate_output_format_native: (String format) -> Integer
+  def self._validate_confidence_native: (Float confidence) -> Integer
+  def self._validate_dpi_native: (Integer dpi) -> Integer
+  def self._validate_chunking_params_native: (Integer max_chars, Integer max_overlap) -> Integer
+  def self._get_valid_binarization_methods_native: () -> String
+  def self._get_valid_language_codes_native: () -> String
+  def self._get_valid_ocr_backends_native: () -> String
+  def self._get_valid_token_reduction_levels_native: () -> String
+
+  # Config wrapper functions
+  def self._config_to_json_native: (String config_json) -> String
+  def self._config_get_field_native: (String config_json, String field_name) -> untyped
+  def self._config_merge_native: (String base_json, String override_json) -> String
+
+  # Result wrapper functions
+  def self._result_page_count_native: (untyped result) -> Integer
+  def self._result_chunk_count_native: (untyped result) -> Integer
+  def self._result_detected_language_native: (untyped result) -> String?
+  def self._result_metadata_field_native: (untyped result, String field_name) -> untyped
+
+  # Plugin registration
+  def self.register_post_processor: (String name, _PostProcessor processor, ?stage: Symbol?) -> void
+  def self.unregister_post_processor: (String name) -> void
+  def self.clear_post_processors: () -> void
+  def self.register_validator: (String name, _Validator validator, ?priority: Integer?) -> void
+  def self.unregister_validator: (String name) -> void
+  def self.clear_validators: () -> void
+  def self.register_ocr_backend: (String name, _OcrBackend backend) -> void
+  def self.unregister_ocr_backend: (String name) -> void
+  def self.list_ocr_backends: () -> Array[String]
+  def self.clear_ocr_backends: () -> void
+  def self.unregister_document_extractor: (String name) -> void
+  def self.list_document_extractors: () -> Array[String]
+  def self.clear_document_extractors: () -> void
+  def self.list_post_processors: () -> Array[String]
+  def self.list_validators: () -> Array[String]
+
+  interface _PostProcessor
+    def call: (extraction_result_hash result) -> extraction_result_hash
+  end
+
+  interface _Validator
+    def call: (extraction_result_hash result) -> void
+  end
+
+  interface _OcrBackend
+    def name: () -> String
+    def process_image: (String image_bytes, Hash[Symbol, untyped] config) -> String
+  end
+
+  module ErrorContext
+    def self.last_error_code: () -> Integer
+    def self.last_panic_context: () -> Errors::PanicContext?
+    def self.last_panic_context_json: () -> String?
+    def self.error_details: () -> Hash[String, untyped]
+    def self.classify_error: (String message) -> (Hash[String, untyped] | Integer)
+    def self.error_code_name: (Integer code) -> String
+    def self.error_code_description: (Integer code) -> String
+  end
+
+  module Errors
+    # Panic context information from FFI error introspection
+    class PanicContext
+      attr_reader file: String
+      attr_reader line: Integer
+      attr_reader function: String
+      attr_reader message: String
+      attr_reader timestamp_secs: Integer
+
+      def initialize: (
+        file: String,
+        line: Integer,
+        function: String,
+        message: String,
+        timestamp_secs: Integer
+      ) -> void
+      def to_s: () -> String
+      def to_h: () -> Hash[Symbol, String | Integer]
+      def self.from_json: (String) -> PanicContext?
+
+      private
+
+      def self.with_defaults: (Hash[Symbol, untyped] sliced) -> {file: String, line: Integer, function: String, message: String, timestamp_secs: Integer}
+    end
+
+    class Error < StandardError
+      attr_reader panic_context: PanicContext?
+      attr_reader error_code: Integer?
+
+      def initialize: (String message, ?panic_context: PanicContext?, ?error_code: Integer?) -> void
+    end
+
+    class ValidationError < Error
+    end
+
+    class ParsingError < Error
+      attr_reader context: Hash[untyped, untyped]?
+
+      def initialize: (String message, ?context: Hash[untyped, untyped]?, ?panic_context: PanicContext?, ?error_code: Integer?) -> void
+    end
+
+    class OCRError < Error
+      attr_reader context: Hash[untyped, untyped]?
+
+      def initialize: (String message, ?context: Hash[untyped, untyped]?, ?panic_context: PanicContext?, ?error_code: Integer?) -> void
+    end
+
+    class MissingDependencyError < Error
+      attr_reader dependency: String?
+
+      def initialize: (String message, ?dependency: String?, ?panic_context: PanicContext?, ?error_code: Integer?) -> void
+    end
+
+    class IOError < Error
+    end
+
+    class PluginError < Error
+    end
+
+    class UnsupportedFormatError < Error
+    end
+  end
+
+  # Internal modules (prepended to Kreuzberg singleton)
+  # These are not checked by steep - see Steepfile
+  module CacheAPI : Object
+  end
+
+  module ExtractionAPI : Object
+  end
+
+  module PostProcessorProtocol
+    def call: (extraction_result_hash result) -> extraction_result_hash
+  end
+
+  module ValidatorProtocol
+    def call: (extraction_result_hash result) -> void
+  end
+
+  module OcrBackendProtocol
+    def name: () -> String
+    def process_image: (String image_bytes, Hash[Symbol, untyped] config) -> String
+  end
+end