omniai-google 2.6.5 → 2.7.7

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: a1f3c06628af28183c2e2224b794ed0f105b136208dc06231ad9fc43e77254e5
-  data.tar.gz: f267290388bc2a067f202d0da53581c0142363b69d99d47c6a4e1d5ef722f720
+  metadata.gz: eb0ebf5227f6a88f24418703712cccfad44d0fa41251f7e84e279211e45bf8aa
+  data.tar.gz: 465f216908801d1c440bf6bb22f1d15df8a39558d3cb2ec91bae290ff959f48d
 SHA512:
-  metadata.gz: 4f013c574fc90c3fe31dccdc4ac424020908866fb8efd0b027c3d92b6e901b2a7656bfba527a37af80af17be9b33e42f1b6caa3b9a784f80e678558e1bbeee53
-  data.tar.gz: '08dea5482ac8679c3d382ecf1830f634e85e652a94f470654ebcb5468898ec30473b7be6047535a056475542d30e1ef0f29dbba851fc1f252050ececfeb3af1a'
+  metadata.gz: 7c2dacf677dc673f2b3c2b8ccf439f59be0cd901b23c886d40b2f8f04e11edd1574616ca2a07a59f4c4406d5bb5e06d5aca0285ba45481b22aafb552efdc5741
+  data.tar.gz: 704d882c87af6d51800d86eaef592b1bf84e749686ad0129d7b837a7e8b800417ec1c2757c7d3728f796d0c53b640175b2eb13e369d8a179028f6f9e48fa064b

data/README.md

@@ -58,6 +58,8 @@ OmniAI::Google.configure do |config|
 end
 ```
 
+**Note for Transcription**: When using transcription features, ensure your service account has the necessary permissions for Google Cloud Speech-to-Text API and Google Cloud Storage (for automatic file uploads). See the [GCS Setup](#gcs-setup-for-transcription) section below for detailed configuration.
+
 Credentials may be configured using:
 
 1. A `File` / `String` / `Pathname`.
@@ -143,6 +145,204 @@ end
 
 [Google API Reference `stream`](https://ai.google.dev/gemini-api/docs/api-overview#stream)
 
+### Transcribe
+
+Audio files can be transcribed using Google's Speech-to-Text API. The implementation automatically handles both synchronous and asynchronous recognition based on file size and model type.
+
+#### Basic Usage
+
+```ruby
+# Transcribe a local audio file
+result = client.transcribe("path/to/audio.mp3")
+result.text # "Hello, this is the transcribed text..."
+
+# Transcribe with specific model
+result = client.transcribe("path/to/audio.mp3", model: "latest_long")
+result.text # "Hello, this is the transcribed text..."
+```
+
+#### Multi-Language Detection
+
+The transcription automatically detects multiple languages when no specific language is provided:
+
+```ruby
+# Auto-detect English and Spanish
+result = client.transcribe("bilingual_audio.mp3", model: "latest_long")
+result.text # "Hello, how are you? Hola, ¿cómo estás?"
+
+# Specify expected languages explicitly
+result = client.transcribe("audio.mp3", language: ["en-US", "es-US"], model: "latest_long")
+```
+
+#### Detailed Transcription with Timestamps
+
+Use `VERBOSE_JSON` format to get detailed timing information, confidence scores, and language detection per segment:
+
+```ruby
+result = client.transcribe("audio.mp3", 
+  model: "latest_long", 
+  format: OmniAI::Transcribe::Format::VERBOSE_JSON
+)
+
+# Access the full transcript
+result.text # "Complete transcribed text..."
+
+# Access detailed segment information
+result.segments.each do |segment|
+  puts "Segment #{segment[:segment_id]}: #{segment[:text]}"
+  puts "Language: #{segment[:language_code]}"
+  puts "Confidence: #{segment[:confidence]}"
+  puts "End time: #{segment[:end_time]}"
+  
+  # Word-level timing (if available)
+  segment[:words].each do |word|
+    puts "  #{word[:word]} (#{word[:start_time]} - #{word[:end_time]})"
+  end
+end
+
+# Total audio duration
+puts "Total duration: #{result.total_duration}"
+```
+
+#### Models
+
+The transcription supports various models optimized for different use cases:
+
+```ruby
+# For short audio (< 60 seconds)
+client.transcribe("short_audio.mp3", model: OmniAI::Google::Transcribe::Model::LATEST_SHORT)
+
+# For long-form audio (> 60 seconds) - automatically uses async processing
+client.transcribe("long_audio.mp3", model: OmniAI::Google::Transcribe::Model::LATEST_LONG)
+
+# For phone/telephony audio
+client.transcribe("phone_call.mp3", model: OmniAI::Google::Transcribe::Model::TELEPHONY_LONG)
+
+# For medical conversations
+client.transcribe("medical_interview.mp3", model: OmniAI::Google::Transcribe::Model::MEDICAL_CONVERSATION)
+
+# Other available models
+client.transcribe("audio.mp3", model: OmniAI::Google::Transcribe::Model::CHIRP_2) # Enhanced model
+client.transcribe("audio.mp3", model: OmniAI::Google::Transcribe::Model::CHIRP)   # Universal model
+```
+
+**Available Model Constants:**
+- `OmniAI::Google::Transcribe::Model::LATEST_SHORT` - Optimized for audio < 60 seconds
+- `OmniAI::Google::Transcribe::Model::LATEST_LONG` - Optimized for long-form audio
+- `OmniAI::Google::Transcribe::Model::TELEPHONY_SHORT` - For short phone calls
+- `OmniAI::Google::Transcribe::Model::TELEPHONY_LONG` - For long phone calls  
+- `OmniAI::Google::Transcribe::Model::MEDICAL_CONVERSATION` - For medical conversations
+- `OmniAI::Google::Transcribe::Model::MEDICAL_DICTATION` - For medical dictation
+- `OmniAI::Google::Transcribe::Model::CHIRP_2` - Enhanced universal model
+- `OmniAI::Google::Transcribe::Model::CHIRP` - Universal model
+
+#### Supported Formats
+
+- **Input**: MP3, WAV, FLAC, and other common audio formats
+- **GCS URIs**: Direct transcription from Google Cloud Storage
+- **File uploads**: Automatic upload to GCS for files > 10MB or long-form models
+
+#### Advanced Features
+
+**Automatic Processing Selection:**
+- Files < 60 seconds: Uses synchronous recognition
+- Files > 60 seconds or long-form models: Uses asynchronous batch recognition
+- Large files: Automatically uploaded to Google Cloud Storage
+
+**GCS Integration:**
+- Automatic file upload and cleanup
+- Support for existing GCS URIs
+- Configurable bucket names
+
+**Error Handling:**
+- Automatic retry logic for temporary failures
+- Clear error messages for common issues
+- Graceful handling of network timeouts
+
+[Google Speech-to-Text API Reference](https://cloud.google.com/speech-to-text/docs)
+
+#### GCS Setup for Transcription
+
+For transcription to work properly with automatic file uploads, you need to set up Google Cloud Storage and configure the appropriate permissions.
+
+##### 1. Create a GCS Bucket
+
+You must create a bucket named `{project_id}-speech-audio` manually before using transcription features:
+
+```bash
+# Using gcloud CLI
+gsutil mb gs://your-project-id-speech-audio
+
+# Or create via Google Cloud Console
+# Navigate to Cloud Storage > Browser > Create Bucket
+```
+
+##### 2. Service Account Permissions
+
+Your service account needs the following IAM roles for transcription to work:
+
+**Required Roles:**
+- **Cloud Speech Editor** - Grants access to edit resources in Speech-to-Text
+- **Storage Bucket Viewer** - Grants permission to view buckets and their metadata, excluding IAM policies
+- **Storage Object Admin** - Grants full control over objects, including listing, creating, viewing, and deleting objects
+
+**To assign roles via gcloud CLI:**
+
+```bash
+# Replace YOUR_SERVICE_ACCOUNT_EMAIL and YOUR_PROJECT_ID with actual values
+SERVICE_ACCOUNT="your-service-account@your-project-id.iam.gserviceaccount.com"
+PROJECT_ID="your-project-id"
+
+# Grant Speech-to-Text permissions
+gcloud projects add-iam-policy-binding $PROJECT_ID \
+    --member="serviceAccount:$SERVICE_ACCOUNT" \
+    --role="roles/speech.editor"
+
+# Grant Storage permissions
+gcloud projects add-iam-policy-binding $PROJECT_ID \
+    --member="serviceAccount:$SERVICE_ACCOUNT" \
+    --role="roles/storage.objectAdmin"
+
+gcloud projects add-iam-policy-binding $PROJECT_ID \
+    --member="serviceAccount:$SERVICE_ACCOUNT" \
+    --role="roles/storage.legacyBucketReader"
+```
+
+**Or via Google Cloud Console:**
+1. Go to IAM & Admin > IAM
+2. Find your service account
+3. Click "Edit Principal" 
+4. Add the required roles listed above
+
+##### 3. Enable Required APIs
+
+Ensure the following APIs are enabled in your Google Cloud Project:
+
+```bash
+# Enable Speech-to-Text API
+gcloud services enable speech.googleapis.com
+
+# Enable Cloud Storage API  
+gcloud services enable storage.googleapis.com
+```
+
+##### 4. Bucket Configuration (Optional)
+
+You can customize the bucket name by configuring it in your application:
+
+```ruby
+# Custom bucket name in your transcription calls
+# The bucket must exist and your service account must have access
+client.transcribe("audio.mp3", bucket_name: "my-custom-audio-bucket")
+```
+
+**Important Notes:**
+- The default bucket name follows the pattern: `{project_id}-speech-audio`
+- You must create the bucket manually before using transcription features
+- Choose an appropriate region for your bucket based on your location and compliance requirements
+- Audio files are automatically deleted after successful transcription
+- If transcription fails, temporary files may remain and should be cleaned up manually
+
 ### Embed
 
 Text can be converted into a vector embedding for similarity comparison usage via:

data/lib/omniai/google/bucket.rb

@@ -0,0 +1,115 @@
+# frozen_string_literal: true
+
+require "google/cloud/storage"
+
+module OmniAI
+  module Google
+    # Uploads audio files to Google Cloud Storage for transcription.
+    class Bucket
+      class UploadError < StandardError; end
+
+      # @param client [Client]
+      # @param io [IO, String]
+      # @param bucket_name [String] optional - bucket name (defaults to project_id-speech-audio)
+      def self.process!(client:, io:, bucket_name: nil)
+        new(client:, io:, bucket_name:).process!
+      end
+
+      # @param client [Client]
+      # @param io [File, String]
+      # @param bucket_name [String] optional - bucket name
+      def initialize(client:, io:, bucket_name: nil)
+        @client = client
+        @io = io
+        @bucket_name = bucket_name || default_bucket_name
+      end
+
+      # @raise [UploadError]
+      #
+      # @return [String] GCS URI (gs://bucket/object)
+      def process!
+        # Create storage client with same credentials as main client
+        credentials = @client.instance_variable_get(:@credentials)
+        storage = ::Google::Cloud::Storage.new(
+          project_id:,
+          credentials:
+        )
+
+        # Get bucket (don't auto-create if it doesn't exist)
+        bucket = storage.bucket(@bucket_name)
+        unless bucket
+          raise UploadError, "Bucket '#{@bucket_name}' not found. " \
+            "Please create it manually or ensure the service account has access."
+        end
+
+        # Generate unique filename
+        timestamp = Time.now.strftime("%Y%m%d_%H%M%S")
+        random_suffix = SecureRandom.hex(4)
+        filename = "audio_#{timestamp}_#{random_suffix}.#{file_extension}"
+
+        # Upload file - create StringIO for binary content
+        content = file_content
+        if content.is_a?(String) && content.include?("\0")
+          # Binary content - wrap in StringIO
+          require "stringio"
+          content = StringIO.new(content)
+        end
+
+        bucket.create_file(content, filename)
+
+        # Return GCS URI
+        "gs://#{@bucket_name}/#{filename}"
+      rescue ::Google::Cloud::Error => e
+        raise UploadError, "Failed to upload to GCS: #{e.message}"
+      end
+
+    private
+
+      # @return [String]
+      def project_id
+        @client.instance_variable_get(:@project_id) ||
+          raise(ArgumentError, "project_id is required for GCS upload")
+      end
+
+      # @return [String]
+      def location_id
+        @client.instance_variable_get(:@location_id) || "global"
+      end
+
+      # @return [String]
+      def default_bucket_name
+        "#{project_id}-speech-audio"
+      end
+
+      # @return [String]
+      def file_content
+        case @io
+        when String
+          # Check if it's a file path or binary content
+          if @io.include?("\0") || !File.exist?(@io)
+            # It's binary content, return as-is
+            @io
+          else
+            # It's a file path, read the file
+            File.read(@io)
+          end
+        when File, IO, StringIO
+          @io.rewind if @io.respond_to?(:rewind)
+          @io.read
+        else
+          raise ArgumentError, "Unsupported input type: #{@io.class}"
+        end
+      end
+
+      # @return [String]
+      def file_extension
+        case @io
+        when String
+          File.extname(@io)[1..] || "wav"
+        else
+          "wav" # Default extension
+        end
+      end
+    end
+  end
+end

data/lib/omniai/google/chat.rb

@@ -15,7 +15,7 @@ module OmniAI
       module Model
         GEMINI_1_0_PRO = "gemini-1.0-pro"
         GEMINI_1_5_PRO = "gemini-1.5-pro"
-        GEMINI_2_5_PRO = "gemini-2.5-pro-preview-05-06"
+        GEMINI_2_5_PRO = "gemini-2.5-pro-preview-06-05"
         GEMINI_1_5_FLASH = "gemini-1.5-flash"
         GEMINI_2_0_FLASH = "gemini-2.0-flash"
         GEMINI_2_5_FLASH = "gemini-2.5-flash-preview-04-17"

data/lib/omniai/google/client.rb

@@ -88,6 +88,16 @@ module OmniAI
         Embed.process!(input, model:, client: self)
       end
 
+      # @raise [OmniAI::Error]
+      #
+      # @param input [String, File, IO] required - audio file path, file object, or GCS URI
+      # @param model [String] optional
+      # @param language [String, Array<String>] optional - language codes for transcription
+      # @param format [Symbol] optional - :json or :verbose_json
+      def transcribe(input, model: Transcribe::DEFAULT_MODEL, language: nil, format: nil)
+        Transcribe.process!(input, model:, language:, format:, client: self)
+      end
+
       # @return [String]
       def path
         if @project_id && @location_id

data/lib/omniai/google/config.rb

@@ -55,6 +55,14 @@ module OmniAI
       def credentials=(value)
         @credentials = Credentials.parse(value)
       end
+
+      # @return [Hash]
+      def transcribe_options
+        @transcribe_options ||= {}
+      end
+
+      # @param value [Hash]
+      attr_writer :transcribe_options
     end
   end
 end

data/lib/omniai/google/transcribe.rb

@@ -0,0 +1,143 @@
+# frozen_string_literal: true
+
+module OmniAI
+  module Google
+    # A Google transcribe implementation.
+    #
+    # Usage:
+    #
+    #   transcribe = OmniAI::Google::Transcribe.new(client: client)
+    #   transcribe.process!(audio_file)
+    class Transcribe < OmniAI::Transcribe
+      include TranscribeHelpers
+      module Model
+        CHIRP_2 = "chirp_2"
+        CHIRP = "chirp"
+        LATEST_LONG = "latest_long"
+        LATEST_SHORT = "latest_short"
+        TELEPHONY_LONG = "telephony_long"
+        TELEPHONY_SHORT = "telephony_short"
+        MEDICAL_CONVERSATION = "medical_conversation"
+        MEDICAL_DICTATION = "medical_dictation"
+      end
+
+      DEFAULT_MODEL = Model::LATEST_SHORT
+      DEFAULT_RECOGNIZER = "_"
+
+      # @return [Context]
+      CONTEXT = Context.build do |context|
+        # No custom deserializers needed - let base class handle parsing
+      end
+
+      # @raise [HTTPError]
+      #
+      # @return [OmniAI::Transcribe::Transcription]
+      def process!
+        if needs_async_recognition?
+          process_async!
+        else
+          process_sync!
+        end
+      end
+
+    private
+
+      # @return [Boolean]
+      def needs_async_recognition?
+        # Use async for long-form models or when GCS is needed
+        needs_long_form_recognition? || needs_gcs_upload?
+      end
+
+      # @raise [HTTPError]
+      #
+      # @return [OmniAI::Transcribe::Transcription]
+      def process_sync!
+        response = request!
+        handle_sync_response_errors(response)
+
+        data = response.parse
+        transcript = data.dig("results", 0, "alternatives", 0, "transcript") || ""
+
+        transformed_data = build_sync_response_data(data, transcript)
+        Transcription.parse(model: @model, format: @format, data: transformed_data)
+      end
+
+      # @raise [HTTPError]
+      #
+      # @return [OmniAI::Transcribe::Transcription]
+      def process_async!
+        # Track if we uploaded the file for cleanup
+        uploaded_gcs_uri = nil
+
+        # Start the batch recognition job
+        response = request_batch!
+
+        raise HTTPError, response unless response.status.ok?
+
+        operation_data = response.parse
+        operation_name = operation_data["name"]
+
+        raise HTTPError, "No operation name returned from batch recognition request" unless operation_name
+
+        # Extract GCS URI for cleanup if we uploaded it
+        if operation_data.dig("metadata", "batchRecognizeRequest", "files")
+          file_uri = operation_data.dig("metadata", "batchRecognizeRequest", "files", 0, "uri")
+          # Only mark for cleanup if it's not a user-provided GCS URI
+          uploaded_gcs_uri = file_uri unless @io.is_a?(String) && @io.start_with?("gs://")
+        end
+
+        # Poll for completion
+        result = poll_operation!(operation_name)
+
+        # Extract transcript from completed operation
+        transcript_data = extract_batch_transcript(result)
+
+        # Clean up uploaded file if we created it
+        cleanup_gcs_file(uploaded_gcs_uri) if uploaded_gcs_uri
+
+        Transcription.parse(model: @model, format: @format, data: transcript_data)
+      end
+
+    protected
+
+      # @return [Context]
+      def context
+        CONTEXT
+      end
+
+      # @return [HTTP::Response]
+      def request!
+        # Speech-to-Text API uses different endpoints for regional vs global
+        endpoint = speech_endpoint
+        speech_connection = HTTP.persistent(endpoint)
+          .timeout(connect: @client.timeout, write: @client.timeout, read: @client.timeout)
+          .accept(:json)
+
+        # Add authentication if using credentials
+        speech_connection = speech_connection.auth("Bearer #{@client.send(:auth).split.last}") if @client.credentials?
+
+        speech_connection.post(path, params:, json: payload)
+      end
+
+      # @return [Hash]
+      def payload
+        config = build_config
+        payload_data = { config: }
+        add_audio_data(payload_data)
+        payload_data
+      end
+
+      # @return [String]
+      def path
+        # Always use Speech-to-Text API v2 with recognizers
+        recognizer_path = "projects/#{project_id}/locations/#{location_id}/recognizers/#{recognizer_name}"
+        "/v2/#{recognizer_path}:recognize"
+      end
+
+      # @return [Hash]
+      def params
+        { key: (@client.api_key unless @client.credentials?) }.compact
+      end
+    end
+  end
+end

data/lib/omniai/google/transcribe_helpers.rb

@@ -0,0 +1,461 @@
+# frozen_string_literal: true
+
+module OmniAI
+  module Google
+    # Helper methods for transcription functionality
+    module TranscribeHelpers # rubocop:disable Metrics/ModuleLength
+    private
+
+      # @return [String]
+      def project_id
+        @client.instance_variable_get(:@project_id) ||
+          raise(ArgumentError, "project_id is required for transcription")
+      end
+
+      # @return [String]
+      def location_id
+        case @model
+        when "chirp_2"
+          "us-central1"
+        else
+          @client.instance_variable_get(:@location_id) || "global"
+        end
+      end
+
+      # @return [String]
+      def speech_endpoint
+        location_id == "global" ? "https://speech.googleapis.com" : "https://#{location_id}-speech.googleapis.com"
+      end
+
+      # @return [Array<String>, nil]
+      def language_codes
+        case @language
+        when String
+          [@language] unless @language.strip.empty?
+        when Array
+          cleaned = @language.compact.reject(&:empty?)
+          cleaned if cleaned.any?
+        when nil, ""
+          nil # Auto-detect language when not specified
+        else
+          ["en-US"] # Default to English (multi-language only supported in global/us/eu locations)
+        end
+      end
+
+      # @param input [String, Pathname, File, IO]
+      # @return [String] Base64 encoded audio content
+      def encode_audio(input)
+        case input
+        when String
+          if File.exist?(input)
+            Base64.strict_encode64(File.read(input))
+          else
+            input # Assume it's already base64 encoded
+          end
+        when Pathname, File, IO, StringIO
+          Base64.strict_encode64(input.read)
+        else
+          raise ArgumentError, "Unsupported input type: #{input.class}"
+        end
+      end
+
+      # @return [Boolean]
+      def needs_gcs_upload?
+        return false if @io.is_a?(String) && @io.start_with?("gs://")
+
+        file_size = calculate_file_size
+        # Force GCS upload for files > 10MB or if using long models for longer audio
+        file_size > 10_000_000 || needs_long_form_recognition?
+      end
+
+      # @return [Boolean]
+      def needs_long_form_recognition?
+        # Use long-form models for potentially longer audio files
+        return true if @model&.include?("long")
+
+        # Chirp models process speech in larger chunks and prefer BatchRecognize
+        return true if @model&.include?("chirp")
+
+        # For large files, assume they might be longer than 60 seconds
+        # Approximate: files larger than 1MB might be longer than 60 seconds
+        calculate_file_size > 1_000_000
+      end
+
+      # @return [Integer]
+      def calculate_file_size
+        case @io
+        when String
+          File.exist?(@io) ? File.size(@io) : 0
+        when File, IO, StringIO
+          @io.respond_to?(:size) ? @io.size : 0
+        else
+          0
+        end
+      end
+
+      # @return [Hash]
+      def build_config
+        config = {
+          model: @model,
+          autoDecodingConfig: {},
+        }
+
+        # Only include languageCodes if specified and non-empty (omit for auto-detection)
+        lang_codes = language_codes
+        config[:languageCodes] = if lang_codes&.any?
+                                   lang_codes
+                                 else
+                                   # Handle language detection based on model capabilities
+                                   default_language_codes
+                                 end
+
+        features = build_features
+        config[:features] = features unless features.empty?
+
+        if OmniAI::Google.config.respond_to?(:transcribe_options)
+          config.merge!(OmniAI::Google.config.transcribe_options)
+        end
+
+        config
+      end
+
+      # @return [Array<String>] Default language codes based on model
+      def default_language_codes
+        if @model&.include?("chirp")
+          # Chirp models use "auto" for automatic language detection
+          ["auto"]
+        else
+          # Other models use multiple languages for auto-detection
+          %w[en-US es-US]
+        end
+      end
+
+      # @return [Hash]
+      def build_features
+        case @format
+        when "verbose_json"
+          {
+            enableAutomaticPunctuation: true,
+            enableWordTimeOffsets: true,
+            enableWordConfidence: true,
+          }
+        when "json"
+          { enableAutomaticPunctuation: true }
+        else
+          {}
+        end
+      end
+
+      # @param payload_data [Hash]
+      def add_audio_data(payload_data)
+        if @io.is_a?(String) && @io.start_with?("gs://")
+          payload_data[:uri] = @io
+        elsif needs_gcs_upload?
+          gcs_uri = Bucket.process!(client: @client, io: @io)
+          payload_data[:uri] = gcs_uri
+        else
+          payload_data[:content] = encode_audio(@io)
+        end
+      end
+
+      # @return [Hash] Payload for batch recognition
+      def batch_payload
+        config = build_config
+
+        # Get audio URI for batch processing
+        audio_uri = if @io.is_a?(String) && @io.start_with?("gs://")
+                      @io
+                    else
+                      # Force GCS upload for batch recognition
+                      Bucket.process!(client: @client, io: @io)
+                    end
+
+        {
+          config:,
+          files: [{ uri: audio_uri }],
+          recognitionOutputConfig: {
+            inlineResponseConfig: {},
+          },
+        }
+      end
+
+      # @param operation_name [String]
+      # @raise [HTTPError]
+      #
+      # @return [Hash]
+      def poll_operation!(operation_name)
+        endpoint = speech_endpoint
+        connection = HTTP.persistent(endpoint)
+          .timeout(connect: @client.timeout, write: @client.timeout, read: @client.timeout)
+          .accept(:json)
+
+        # Add authentication if using credentials
+        connection = connection.auth("Bearer #{@client.send(:auth).split.last}") if @client.credentials?
+
+        max_attempts = 60 # Maximum 15 minutes (15 second intervals)
+        attempt = 0
+
+        loop do
+          attempt += 1
+
+          raise HTTPError, "Operation timed out after #{max_attempts * 15} seconds" if attempt > max_attempts
+
+          operation_response = connection.get("/v2/#{operation_name}", params: operation_params)
+
+          raise HTTPError, operation_response unless operation_response.status.ok?
+
+          operation_data = operation_response.parse
+
+          # Check for errors
+          if operation_data["error"]
+            error_message = operation_data.dig("error", "message") || "Unknown error"
+            raise HTTPError, "Operation failed: #{error_message}"
+          end
+
+          # Check if done
+          return operation_data if operation_data["done"]
+
+          # Wait before polling again
+          sleep(15)
+        end
+      end
+
+      # @return [HTTP::Response]
+      def request_batch!
+        endpoint = speech_endpoint
+        connection = HTTP.persistent(endpoint)
+          .timeout(connect: @client.timeout, write: @client.timeout, read: @client.timeout)
+          .accept(:json)
+
+        # Add authentication if using credentials
+        connection = connection.auth("Bearer #{@client.send(:auth).split.last}") if @client.credentials?
+
+        connection.post(batch_path, params: operation_params, json: batch_payload)
+      end
+
+      # @return [String]
+      def batch_path
+        # Use batchRecognize endpoint for async recognition
+        recognizer_path = "projects/#{project_id}/locations/#{location_id}/recognizers/#{recognizer_name}"
+        "/v2/#{recognizer_path}:batchRecognize"
+      end
+
+      # @return [Hash]
+      def operation_params
+        { key: (@client.api_key unless @client.credentials?) }.compact
+      end
+
+      # @return [String]
+      def recognizer_name
+        # Always use the default recognizer - the model is specified in the config
+        "_"
+      end
+
+      # @param result [Hash] Operation result from batch recognition
+      # @return [Hash] Data formatted for OmniAI::Transcribe::Transcription.parse
+      def extract_batch_transcript(result)
+        batch_results = result.dig("response", "results")
+        return empty_transcript_data unless batch_results
+
+        file_result = batch_results.values.first
+        return empty_transcript_data unless file_result
+
+        transcript_segments = file_result.dig("transcript", "results")
+        return empty_transcript_data unless transcript_segments&.any?
+
+        build_transcript_data(transcript_segments, file_result)
+      end
+
+      # @return [Hash]
+      def empty_transcript_data
+        { "text" => "" }
+      end
+
+      # @param transcript_segments [Array]
+      # @param file_result [Hash]
+      # @return [Hash]
+      def build_transcript_data(transcript_segments, file_result)
+        transcript_text = extract_transcript_text(transcript_segments)
+        result_data = { "text" => transcript_text }
+
+        add_duration_if_available(result_data, file_result)
+        add_segments_if_verbose(result_data, transcript_segments)
+
+        result_data
+      end
+
+      # @param transcript_segments [Array]
+      # @return [String]
+      def extract_transcript_text(transcript_segments)
+        text_segments = transcript_segments.map do |segment|
+          segment.dig("alternatives", 0, "transcript")
+        end.compact
+
+        text_segments.join(" ")
+      end
+
+      # @param result_data [Hash]
+      # @param file_result [Hash]
+      def add_duration_if_available(result_data, file_result)
+        duration = file_result.dig("metadata", "totalBilledDuration")
+        result_data["duration"] = parse_duration(duration) if duration
+      end
+
+      # @param result_data [Hash]
+      # @param transcript_segments [Array]
+      def add_segments_if_verbose(result_data, transcript_segments)
+        result_data["segments"] = build_segments(transcript_segments) if @format == "verbose_json"
+      end
+
+      # @param duration_string [String] Duration in Google's format (e.g., "123.456s")
+      # @return [Float] Duration in seconds
+      def parse_duration(duration_string)
+        return nil unless duration_string
+
+        duration_string.to_s.sub(/s$/, "").to_f
+      end
+
+      # @param segments [Array] Transcript segments from Google API
+      # @return [Array<Hash>] Segments formatted for base class
+      def build_segments(segments)
+        segments.map.with_index do |segment, index|
+          alternative = segment.dig("alternatives", 0)
+          next unless alternative
+
+          segment_data = {
+            "id" => index,
+            "text" => alternative["transcript"],
+            "start" => calculate_segment_start(segments, index),
+            "end" => parse_duration(segment["resultEndOffset"]),
+            "confidence" => alternative["confidence"],
+          }
+
+          # Words removed - segments provide sufficient granularity for most use cases
+
+          segment_data
+        end.compact
+      end
+
+      # @param segments [Array] All segments
+      # @param index [Integer] Current segment index
+      # @return [Float] Start time estimated from previous segment end
+      def calculate_segment_start(segments, index)
+        return 0.0 if index.zero?
+
+        prev_segment = segments[index - 1]
+        parse_duration(prev_segment["resultEndOffset"]) || 0.0
+      end
+
+      # @param response [HTTP::Response]
+      # @raise [HTTPError]
+      def handle_sync_response_errors(response)
+        return if response.status.ok?
+
+        error_data = parse_error_data(response)
+        raise_timeout_error(response) if timeout_error?(error_data)
+        raise HTTPError, response
+      end
+
+      # @param response [HTTP::Response]
+      # @return [Hash]
+      def parse_error_data(response)
+        response.parse
+      rescue StandardError
+        {}
+      end
+
+      # @param error_data [Hash]
+      # @return [Boolean]
+      def timeout_error?(error_data)
+        error_data.dig("error", "message")&.include?("60 seconds")
+      end
+
+      # @param response [HTTP::Response]
+      # @raise [HTTPError]
+      def raise_timeout_error(response)
+        raise HTTPError, (response.tap do |r|
+          r.instance_variable_set(:@body, "Audio file exceeds 60-second limit for direct upload. " \
+            "Use a long-form model (e.g., 'latest_long') or upload to GCS first. " \
+            "Original error: #{response.flush}")
+        end)
+      end
+
+      # @param data [Hash]
+      # @param transcript [String]
+      # @return [Hash]
+      def build_sync_response_data(data, transcript)
+        return { "text" => transcript } unless verbose_json_format?(data)
+
+        build_verbose_sync_data(data, transcript)
+      end
+
+      # @param data [Hash]
+      # @return [Boolean]
+      def verbose_json_format?(data)
+        @format == "verbose_json" &&
+          data["results"]&.any? &&
+          data["results"][0]["alternatives"]&.any?
+      end
+
+      # @param data [Hash]
+      # @param transcript [String]
+      # @return [Hash]
+      def build_verbose_sync_data(data, transcript)
+        alternative = data["results"][0]["alternatives"][0]
+        {
+          "text" => transcript,
+          "segments" => [{
+            "id" => 0,
+            "text" => transcript,
+            "start" => 0.0,
+            "end" => nil,
+            "confidence" => alternative["confidence"],
+          }],
+        }
+      end
+
+      # @param gcs_uri [String] GCS URI to delete (e.g., "gs://bucket/file.mp3")
+      def cleanup_gcs_file(gcs_uri)
+        return unless valid_gcs_uri?(gcs_uri)
+
+        bucket_name, object_name = parse_gcs_uri(gcs_uri)
+        return unless bucket_name && object_name
+
+        delete_gcs_object(bucket_name, object_name, gcs_uri)
+      end
+
+      # @param gcs_uri [String]
+      # @return [Boolean]
+      def valid_gcs_uri?(gcs_uri)
+        gcs_uri&.start_with?("gs://")
+      end
+
+      # @param gcs_uri [String]
+      # @return [Array<String>] [bucket_name, object_name]
+      def parse_gcs_uri(gcs_uri)
+        uri_parts = gcs_uri.sub("gs://", "").split("/", 2)
+        [uri_parts[0], uri_parts[1]]
+      end
+
+      # @param bucket_name [String]
+      # @param object_name [String]
+      # @param gcs_uri [String]
+      def delete_gcs_object(bucket_name, object_name, gcs_uri)
+        storage = create_storage_client
+        bucket = storage.bucket(bucket_name)
+        return unless bucket
+
+        file = bucket.file(object_name)
+        file&.delete
+      rescue ::Google::Cloud::Error => e
+        @client.logger&.warn("Failed to cleanup GCS file #{gcs_uri}: #{e.message}")
+      end
+
+      # @return [Google::Cloud::Storage]
+      def create_storage_client
+        credentials = @client.instance_variable_get(:@credentials)
+        ::Google::Cloud::Storage.new(project_id:, credentials:)
+      end
+    end
+  end
+end

data/lib/omniai/google/version.rb

@@ -2,6 +2,6 @@
 
 module OmniAI
   module Google
-    VERSION = "2.6.5"
+    VERSION = "2.7.7"
   end
 end

metadata

@@ -1,13 +1,13 @@
 --- !ruby/object:Gem::Specification
 name: omniai-google
 version: !ruby/object:Gem::Version
-  version: 2.6.5
+  version: 2.7.7
 platform: ruby
 authors:
 - Kevin Sylvestre
 bindir: exe
 cert_chain: []
-date: 2025-05-14 00:00:00.000000000 Z
+date: 2025-06-16 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: event_stream_parser
@@ -37,6 +37,20 @@ dependencies:
     - - ">="
       - !ruby/object:Gem::Version
         version: '0'
+- !ruby/object:Gem::Dependency
+  name: google-cloud-storage
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: '0'
+  type: :runtime
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: '0'
 - !ruby/object:Gem::Dependency
   name: omniai
   requirement: !ruby/object:Gem::Requirement
@@ -75,6 +89,7 @@ files:
 - Gemfile
 - README.md
 - lib/omniai/google.rb
+- lib/omniai/google/bucket.rb
 - lib/omniai/google/chat.rb
 - lib/omniai/google/chat/choice_serializer.rb
 - lib/omniai/google/chat/content_serializer.rb
@@ -92,6 +107,8 @@ files:
 - lib/omniai/google/config.rb
 - lib/omniai/google/credentials.rb
 - lib/omniai/google/embed.rb
+- lib/omniai/google/transcribe.rb
+- lib/omniai/google/transcribe_helpers.rb
 - lib/omniai/google/upload.rb
 - lib/omniai/google/upload/file.rb
 - lib/omniai/google/version.rb