omniai-mistral 2.0.0 → 2.0.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 29008395f4e0da0a72899f2ddb6817f727a1a2c5be46063aa6693043164bcfcf
-  data.tar.gz: 92a5bd8542912d968aecb7d13d1d64080144d8d20a0b53cd47fd8f757a060fb7
+  metadata.gz: 360410f4a0d40370818852aafbb3d7d8680d8758b30e42e9c00d2b0d2f3508fb
+  data.tar.gz: 115604d511fe504e8cd88ff0f846e811d8f6d3e2da69258fd029654bf737f277
 SHA512:
-  metadata.gz: 14faceab19d119a03fee9b6226b25569c0cbea0a6b3bb680ea47db5d4c483487e45a0c8e91f6415a49583a7a0ec6a0836a3817d4265dcc4998bf2f795bc8a106
-  data.tar.gz: 7fb98a39c2e34db5ee12f62dd794d7f7ffb33d5b1bcb0099fdb5fcc3d3b36096d2972eb9061d033bdf969276d1ec7d573bda89184bc2a8180d8aa1e85d52d874
+  metadata.gz: efe4e899a3b518133788cb9385db9e24cdd7e269223e686cf118ad4e775b61000f1a7a97a9c380653c8f511f8624970c89fecca240f10e3eae0d7c507b5603c3
+  data.tar.gz: 90f89f6856daeff1974ccea7f4858affa6c501f3011e9a3a1a16ff59948f025b97a50e2c6b92e01b18891c91b4ad5868303f3e5c82124f9bf35df06c775bf200

data/lib/omniai/mistral/client.rb

@@ -54,7 +54,7 @@ module OmniAI
       # @yield [prompt] optional
       # @yieldparam prompt [OmniAI::Chat::Prompt]
       #
-      # @return [OmniAI::Chat::Completion]
+      # @return [OmniAI::Chat::Response]
       def chat(messages = nil, model: Chat::DEFAULT_MODEL, temperature: nil, format: nil, stream: nil, tools: nil, &)
         Chat.process!(messages, model:, temperature:, format:, stream:, tools:, client: self, &)
       end
@@ -63,9 +63,25 @@ module OmniAI
       #
       # @param input [String, Array<String>, Array<Integer>] required
       # @param model [String] optional
+      #
+      # @return [OmniAI::Embed::Response]
       def embed(input, model: Embed::DEFAULT_MODEL)
         Embed.process!(input, model:, client: self)
       end
+
+      # @raise [OmniAI::Error]
+      #
+      # @param document [String] e.g. "https://vancouver.ca/files/cov/other-sectors-tourism.PDF"
+      # @param kind [Symbol] optional - `:document` or `:image` - default = `:document``
+      # @param model [String] optional - default = "mistral-ocr-latest"
+      # @param options [Hash] optional - e.g. `{ image_limit: 4 }`
+      #
+      # @raise [OmniAI::Error]
+      #
+      # @return [OmniAI::Mistral::OCR::Response]
+      def ocr(document, kind: OCR::DEFAULT_KIND, model: OCR::DEFAULT_MODEL, options: {})
+        OCR.process!(document, kind:, model:, options:, client: self)
+      end
     end
   end
 end

data/lib/omniai/mistral/ocr/dimensions.rb

@@ -0,0 +1,38 @@
+# frozen_string_literal: true
+
+module OmniAI
+  module Mistral
+    class OCR
+      # Dimensions returned by the OCR API.
+      class Dimensions
+        # !@attribute [rw] width
+        #   @return [Integer]
+        attr_accessor :width
+
+        # !@attribute [rw] height
+        #   @return [Integer]
+        attr_accessor :height
+
+        # !@attribute [rw] dpi
+        #   @return [Integer]
+        attr_accessor :dpi
+
+        # @param width [Integer]
+        # @param height [Integer]
+        # @param dpi [Integer]
+        def initialize(width:, height:, dpi:)
+          @width = width
+          @height = height
+          @dpi = dpi
+        end
+
+        # @param data [Hash]
+        #
+        # @return [Dimensions]
+        def self.parse(data:)
+          new(width: data["width"], height: data["height"], dpi: data["dpi"])
+        end
+      end
+    end
+  end
+end

data/lib/omniai/mistral/ocr/image.rb

@@ -0,0 +1,63 @@
+# frozen_string_literal: true
+
+module OmniAI
+  module Mistral
+    class OCR
+      # An image returned by the OCR API.
+      class Image
+        # !@attribute [rw] id
+        #   @return [Integer]
+        attr_accessor :id
+
+        # !@attribute [rw] top_left_x
+        #   @return [Integer]
+        attr_accessor :top_left_x
+
+        # !@attribute [rw] top_left_y
+        #   @return [Integer]
+        attr_accessor :top_left_y
+
+        # !@attribute [rw] bottom_right_x
+        #   @return [Integer]
+        attr_accessor :bottom_right_x
+
+        # !@attribute [rw] bottom_right_y
+        #   @return [Integer]
+        attr_accessor :bottom_right_y
+
+        # !@attribute [rw] image_base64
+        #   @return [String]
+        attr_accessor :image_base64
+
+        # @param id [Integer]
+        # @param top_left_x [Integer]
+        # @param top_left_y [Integer]
+        # @param bottom_right_x [Integer]
+        # @param bottom_right_y [Integer]
+        # @param image_base64 [String]
+        def initialize(id:, top_left_x:, top_left_y:, bottom_right_x:, bottom_right_y:, image_base64:)
+          @id = id
+          @top_left_x = top_left_x
+          @top_left_y = top_left_y
+          @bottom_right_x = bottom_right_x
+          @bottom_right_y = bottom_right_y
+          @image_base64 = image_base64
+        end
+
+        # @param data [Hash]
+        #
+        # @return [Image]
+        def self.parse(data:)
+          new(
+            id: data["id"],
+            top_left_x: data["top_left_x"],
+            top_left_y: data["top_left_y"],
+            bottom_right_x: data["bottom_right_x"],
+            bottom_right_y: data["bottom_right_y"],
+            image_base64: data["image_base64"]
+          )
+        end
+      end
+    end
+  end
+end

data/lib/omniai/mistral/ocr/page.rb

@@ -0,0 +1,49 @@
+# frozen_string_literal: true
+
+module OmniAI
+  module Mistral
+    class OCR
+      # A page returned by the OCR API.
+      class Page
+        # !@attribute [rw] index
+        #   @return [Integer]
+        attr_accessor :index
+
+        # !@attribute [rw] markdown
+        #   @return [String]
+        attr_accessor :markdown
+
+        # !@attribute [rw] images
+        #   @return [Array<Image>]
+        attr_accessor :images
+
+        # !@attribute [rw] dimensions
+        #   @return [Dimensions]
+        attr_accessor :dimensions
+
+        # @param index [Integer]
+        # @param markdown [String]
+        # @param images [Array<Image>]
+        # @param dimensions [Dimensions]
+        def initialize(index:, markdown:, images:, dimensions:)
+          @index = index
+          @markdown = markdown
+          @images = images
+          @dimensions = dimensions
+        end
+
+        # @param data [Hash]
+        #
+        # @return [Page]
+        def self.parse(data:)
+          new(
+            index: data["index"],
+            markdown: data["markdown"],
+            images: data["images"].map { |image_data| Image.parse(data: image_data) },
+            dimensions: Dimensions.parse(data: data["dimensions"])
+          )
+        end
+      end
+    end
+  end
+end

data/lib/omniai/mistral/ocr/response.rb

@@ -0,0 +1,32 @@
+# frozen_string_literal: true
+
+module OmniAI
+  module Mistral
+    class OCR
+      # The response from the OCR API.
+      class Response
+        # !@attribute [rw] model
+        #   @return [Array<Model>]
+        attr_accessor :model
+
+        # !@attribute [rw] pages
+        #   @return [Array<Page>]
+        attr_accessor :pages
+
+        # @param model [String]
+        # @param pages [Array<Page>]
+        def initialize(model:, pages:)
+          @model = model
+          @pages = pages
+        end
+
+        # @param data [Hash]
+        #
+        # @return [Response]
+        def self.parse(data:)
+          new(model: data["model"], pages: data["pages"].map { |page| Page.parse(data: page) })
+        end
+      end
+    end
+  end
+end

data/lib/omniai/mistral/ocr.rb

@@ -0,0 +1,70 @@
+# frozen_string_literal: true
+
+module OmniAI
+  module Mistral
+    # An implementation of the OCR API.
+    class OCR
+      VERSION = "v1"
+
+      module Kind
+        DOCUMENT = :document
+        IMAGE = :image
+      end
+
+      module Model
+        OCR_LATEST = "mistral-ocr-latest"
+      end
+
+      DEFAULT_KIND = Kind::DOCUMENT
+      DEFAULT_MODEL = Model::OCR_LATEST
+
+      # @raise [Error]
+      #
+      # @param client [OmniAI::Mistral::Client]
+      # @param document [String] a URL
+      # @param kind [Symbol] optional
+      # @param model [String] optional
+      # @param options [Hash] optional (e.g. `{ image_limit: 4 }`)
+      #
+      # @return [Response]
+      def self.process!(document, client:, kind: DEFAULT_KIND, model: DEFAULT_MODEL, options: {})
+        new(document, client:, kind:, model:, options:).process!
+      end
+
+      # @param client [OmniAI::Mistral::Client]
+      # @param document [String] a URL
+      # @param kind [Symbol] optional
+      # @param model [String] optional
+      # @param options [Hash] optional (e.g. `{ image_limit: 4 }`)
+      def initialize(document, client:, kind: DEFAULT_KIND, model: DEFAULT_MODEL, options: {})
+        @document = document
+        @client = client
+        @options = options
+        @kind = kind
+        @model = model
+      end
+
+      # @raise [Error]
+      #
+      # @return [Response]
+      def process!
+        response = @client.connection.accept(:json).post("/#{VERSION}/ocr", json: @options.merge({
+          model: @model,
+          document: document!,
+        }).compact)
+
+        raise HTTPError, response.flush unless response.status.ok?
+
+        Response.parse(data: response.parse)
+      end
+
+      # @return [Hash]
+      def document!
+        case @kind
+        when Kind::DOCUMENT then { document_url: @document, type: "document_url" }
+        when Kind::IMAGE then { image_url: @document, type: "image_url" }
+        end
+      end
+    end
+  end
+end

data/lib/omniai/mistral/version.rb

@@ -2,6 +2,6 @@
 
 module OmniAI
   module Mistral
-    VERSION = "2.0.0"
+    VERSION = "2.0.1"
   end
 end

data/lib/omniai/mistral.rb

@@ -6,6 +6,7 @@ require "zeitwerk"
 
 loader = Zeitwerk::Loader.for_gem
 loader.push_dir(__dir__, namespace: OmniAI)
+loader.inflector.inflect "ocr" => "OCR"
 loader.setup
 
 module OmniAI

metadata

@@ -1,13 +1,13 @@
 --- !ruby/object:Gem::Specification
 name: omniai-mistral
 version: !ruby/object:Gem::Version
-  version: 2.0.0
+  version: 2.0.1
 platform: ruby
 authors:
 - Kevin Sylvestre
 bindir: exe
 cert_chain: []
-date: 2025-03-03 00:00:00.000000000 Z
+date: 2025-03-10 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: event_stream_parser
@@ -65,6 +65,11 @@ files:
 - lib/omniai/mistral/client.rb
 - lib/omniai/mistral/config.rb
 - lib/omniai/mistral/embed.rb
+- lib/omniai/mistral/ocr.rb
+- lib/omniai/mistral/ocr/dimensions.rb
+- lib/omniai/mistral/ocr/image.rb
+- lib/omniai/mistral/ocr/page.rb
+- lib/omniai/mistral/ocr/response.rb
 - lib/omniai/mistral/version.rb
 homepage: https://github.com/ksylvest/omniai-mistral
 licenses: