ruby_llm-agents 0.4.0 → 1.0.0.beta.1

data/lib/ruby_llm/agents/audio/speaker.rb

@@ -0,0 +1,553 @@
+# frozen_string_literal: true
+
+require "digest"
+require_relative "../results/speech_result"
+
+module RubyLLM
+  module Agents
+    # Base class for creating text-to-speech speakers using the middleware pipeline
+    #
+    # Speaker provides a DSL for configuring text-to-audio operations with
+    # built-in execution tracking, budget controls, and multi-tenancy support
+    # through the middleware pipeline.
+    #
+    # @example Basic usage
+    #   class ArticleNarrator < RubyLLM::Agents::Speaker
+    #     provider :openai
+    #     model 'tts-1-hd'
+    #     voice 'nova'
+    #   end
+    #
+    #   result = ArticleNarrator.call(text: "Hello world")
+    #   result.audio       # => Binary audio data
+    #   result.save_to("output.mp3")
+    #
+    # @example With voice settings
+    #   class PremiumNarrator < RubyLLM::Agents::Speaker
+    #     provider :elevenlabs
+    #     model 'eleven_multilingual_v2'
+    #     voice 'Rachel'
+    #
+    #     voice_settings do
+    #       stability 0.5
+    #       similarity_boost 0.75
+    #     end
+    #   end
+    #
+    # @api public
+    class Speaker < BaseAgent
+      class << self
+        # Returns the agent type for speakers
+        #
+        # @return [Symbol] :audio
+        def agent_type
+          :audio
+        end
+
+        # @!group Speaker-specific DSL
+
+        # Sets or returns the TTS provider
+        #
+        # @param value [Symbol, nil] The provider (:openai, :elevenlabs, :google, :polly)
+        # @return [Symbol] The current provider setting
+        def provider(value = nil)
+          @provider = value if value
+          return @provider if defined?(@provider) && @provider
+
+          if superclass.respond_to?(:agent_type) && superclass.agent_type == :audio
+            superclass.provider
+          else
+            default_tts_provider
+          end
+        end
+
+        # Sets or returns the TTS model
+        #
+        # @param value [String, nil] The model identifier
+        # @return [String] The current model setting
+        def model(value = nil)
+          @model = value if value
+          return @model if defined?(@model) && @model
+
+          if superclass.respond_to?(:agent_type) && superclass.agent_type == :audio
+            superclass.model
+          else
+            default_tts_model
+          end
+        end
+
+        # Sets or returns the voice name
+        #
+        # @param value [String, nil] The voice name
+        # @return [String] The current voice setting
+        def voice(value = nil)
+          @voice = value if value
+          @voice || inherited_or_default(:voice, default_tts_voice)
+        end
+
+        # Sets or returns the voice ID (for custom/cloned voices)
+        #
+        # @param value [String, nil] The voice ID
+        # @return [String, nil] The current voice ID
+        def voice_id(value = nil)
+          @voice_id = value if value
+          @voice_id || inherited_or_default(:voice_id, nil)
+        end
+
+        # Sets or returns the speech speed
+        #
+        # @param value [Float, nil] Speed multiplier
+        # @return [Float] The current speed
+        def speed(value = nil)
+          @speed = value if value
+          @speed || inherited_or_default(:speed, 1.0)
+        end
+
+        # Sets or returns the output format
+        #
+        # @param value [Symbol, nil] Format (:mp3, :wav, :ogg, etc.)
+        # @return [Symbol] The current format
+        def output_format(value = nil)
+          @output_format = value if value
+          @output_format || inherited_or_default(:output_format, :mp3)
+        end
+
+        # Sets or returns streaming mode
+        #
+        # @param value [Boolean, nil] Enable streaming
+        # @return [Boolean] The current streaming setting
+        def streaming(value = nil)
+          @streaming = value unless value.nil?
+          instance_variable_defined?(:@streaming) ? @streaming : inherited_or_default(:streaming, false)
+        end
+
+        def streaming?
+          streaming
+        end
+
+        # @!endgroup
+
+        # @!group Voice Settings DSL
+
+        # Configures voice settings (ElevenLabs specific)
+        #
+        # @yield Block for configuring voice settings
+        # @return [VoiceSettings] The voice settings configuration
+        def voice_settings(&block)
+          @voice_settings ||= VoiceSettings.new
+          @voice_settings.instance_eval(&block) if block_given?
+          @voice_settings
+        end
+
+        def voice_settings_config
+          @voice_settings || inherited_or_default(:voice_settings_config, nil)
+        end
+
+        # @!endgroup
+
+        # @!group Lexicon DSL
+
+        # Configures pronunciation lexicon
+        #
+        # @yield Block for configuring pronunciations
+        # @return [Lexicon] The lexicon configuration
+        def lexicon(&block)
+          @lexicon ||= Lexicon.new
+          @lexicon.instance_eval(&block) if block_given?
+          @lexicon
+        end
+
+        def lexicon_config
+          @lexicon || inherited_or_default(:lexicon_config, nil)
+        end
+
+        # @!endgroup
+
+        # Factory method to instantiate and execute speaker
+        #
+        # @param text [String] Text to convert to speech
+        # @param options [Hash] Additional options
+        # @yield [audio_chunk] Called for each audio chunk when streaming
+        # @return [SpeechResult] The speech result
+        def call(text:, **options, &block)
+          new(text: text, **options).call(&block)
+        end
+
+        # Streams the speaker output
+        #
+        # @param text [String] Text to convert to speech
+        # @param options [Hash] Additional options
+        # @yield [audio_chunk] Called for each audio chunk
+        # @return [SpeechResult] The speech result
+        def stream(text:, **options, &block)
+          raise ArgumentError, "A block is required for streaming" unless block_given?
+
+          instance = new(text: text, **options.merge(streaming: true))
+          instance.call(&block)
+        end
+
+        private
+
+        def inherited_or_default(method, default)
+          superclass.respond_to?(method) ? superclass.send(method) : default
+        end
+
+        def default_tts_provider
+          RubyLLM::Agents.configuration.default_tts_provider
+        rescue StandardError
+          :openai
+        end
+
+        def default_tts_model
+          RubyLLM::Agents.configuration.default_tts_model
+        rescue StandardError
+          "tts-1"
+        end
+
+        def default_tts_voice
+          RubyLLM::Agents.configuration.default_tts_voice
+        rescue StandardError
+          "nova"
+        end
+      end
+
+      # Voice settings configuration class
+      class VoiceSettings
+        attr_accessor :stability_value, :similarity_boost_value, :style_value, :speaker_boost_value
+
+        def initialize
+          @stability_value = 0.5
+          @similarity_boost_value = 0.75
+          @style_value = 0.0
+          @speaker_boost_value = true
+        end
+
+        def stability(value)
+          @stability_value = value
+        end
+
+        def similarity_boost(value)
+          @similarity_boost_value = value
+        end
+
+        def style(value)
+          @style_value = value
+        end
+
+        def speaker_boost(value)
+          @speaker_boost_value = value
+        end
+
+        def to_h
+          {
+            stability: stability_value,
+            similarity_boost: similarity_boost_value,
+            style: style_value,
+            use_speaker_boost: speaker_boost_value
+          }
+        end
+      end
+
+      # Pronunciation lexicon class
+      class Lexicon
+        attr_reader :pronunciations
+
+        def initialize
+          @pronunciations = {}
+        end
+
+        def pronounce(word, pronunciation)
+          @pronunciations[word] = pronunciation
+        end
+
+        def apply(text)
+          result = text.dup
+          pronunciations.each do |word, pronunciation|
+            result.gsub!(/\b#{Regexp.escape(word)}\b/i, pronunciation)
+          end
+          result
+        end
+
+        def to_h
+          pronunciations.dup
+        end
+      end
+
+      # @!attribute [r] text
+      #   @return [String] Text to convert to speech
+      attr_reader :text
+
+      # Creates a new Speaker instance
+      #
+      # @param text [String] Text to convert to speech
+      # @param options [Hash] Configuration options
+      def initialize(text:, **options)
+        @text = text
+        @streaming_block = nil
+        @runtime_streaming = options.delete(:streaming)
+
+        # Set model to TTS model if not specified
+        options[:model] ||= self.class.model
+
+        super(**options)
+      end
+
+      # Executes the speech through the middleware pipeline
+      #
+      # @yield [audio_chunk] Called for each audio chunk when streaming
+      # @return [SpeechResult] The speech result
+      def call(&block)
+        @streaming_block = block
+        context = build_context
+        result_context = Pipeline::Executor.execute(context)
+        result_context.output
+      end
+
+      # The input for this speech operation
+      #
+      # @return [String] The text being converted
+      def user_prompt
+        text
+      end
+
+      # Core speech execution
+      #
+      # This is called by the Pipeline::Executor after middleware
+      # has been applied. Only contains the speech API logic.
+      #
+      # @param context [Pipeline::Context] The execution context
+      # @return [void] Sets context.output with the SpeechResult
+      def execute(context)
+        execution_started_at = Time.current
+
+        validate_text_input!
+        processed_text = apply_lexicon(text)
+
+        # Execute speech synthesis
+        result = execute_speech(processed_text)
+
+        execution_completed_at = Time.current
+        duration_ms = ((execution_completed_at - execution_started_at) * 1000).to_i
+
+        # Update context
+        context.input_tokens = 0
+        context.output_tokens = 0
+        context.total_cost = calculate_cost(result)
+
+        # Build final result
+        context.output = build_result(
+          result,
+          text,
+          started_at: context.started_at || execution_started_at,
+          completed_at: execution_completed_at,
+          duration_ms: duration_ms,
+          tenant_id: context.tenant_id
+        )
+      end
+
+      # Generates the cache key for this speech
+      #
+      # @return [String] Cache key
+      def agent_cache_key
+        components = [
+          "ruby_llm_agents",
+          "speech",
+          self.class.name,
+          self.class.version,
+          resolved_provider,
+          resolved_model,
+          resolved_voice,
+          resolved_voice_id,
+          resolved_speed,
+          resolved_output_format,
+          Digest::SHA256.hexdigest(text)
+        ].compact
+
+        components.join("/")
+      end
+
+      private
+
+      # Builds context for pipeline execution
+      #
+      # @return [Pipeline::Context] The context object
+      def build_context
+        Pipeline::Context.new(
+          input: user_prompt,
+          agent_class: self.class,
+          agent_instance: self,
+          model: resolved_model,
+          tenant: @options[:tenant],
+          skip_cache: @options[:skip_cache] || streaming_enabled?,
+          stream_block: (@streaming_block if streaming_enabled?)
+        )
+      end
+
+      # Validates text input
+      def validate_text_input!
+        raise ArgumentError, "text is required" if text.nil?
+        raise ArgumentError, "text must be a String, got #{text.class}" unless text.is_a?(String)
+        raise ArgumentError, "text cannot be empty" if text.empty?
+      end
+
+      # Applies lexicon pronunciations
+      def apply_lexicon(text)
+        lexicon = self.class.lexicon_config
+        return text unless lexicon
+
+        lexicon.apply(text)
+      end
+
+      # Executes speech synthesis
+      def execute_speech(processed_text)
+        speak_options = build_speak_options
+
+        if streaming_enabled? && @streaming_block
+          execute_streaming_speech(processed_text, speak_options)
+        else
+          execute_standard_speech(processed_text, speak_options)
+        end
+      end
+
+      # Executes standard (non-streaming) speech synthesis
+      def execute_standard_speech(text, options)
+        response = RubyLLM.speak(text, **options)
+
+        {
+          audio: response.audio,
+          duration: response.respond_to?(:duration) ? response.duration : nil,
+          format: resolved_output_format,
+          provider: resolved_provider,
+          model: resolved_model,
+          voice: resolved_voice,
+          characters: text.length,
+          raw_response: response
+        }
+      end
+
+      # Executes streaming speech synthesis
+      def execute_streaming_speech(text, options)
+        audio_chunks = []
+
+        RubyLLM.speak(text, **options.merge(stream: true)) do |chunk|
+          audio_chunks << chunk.audio if chunk.respond_to?(:audio)
+          @streaming_block.call(chunk) if @streaming_block
+        end
+
+        {
+          audio: audio_chunks.join,
+          duration: nil,
+          format: resolved_output_format,
+          provider: resolved_provider,
+          model: resolved_model,
+          voice: resolved_voice,
+          characters: text.length,
+          streamed: true
+        }
+      end
+
+      # Builds options for RubyLLM.speak
+      def build_speak_options
+        options = {
+          model: resolved_model,
+          voice: resolved_voice_id || resolved_voice
+        }
+
+        speed = resolved_speed
+        options[:speed] = speed if speed && speed != 1.0
+        options[:response_format] = resolved_output_format.to_s
+
+        if resolved_provider == :elevenlabs
+          voice_settings = self.class.voice_settings_config
+          options[:voice_settings] = voice_settings.to_h if voice_settings
+        end
+
+        options
+      end
+
+      # Builds the final result object
+      def build_result(raw_result, original_text, started_at:, completed_at:, duration_ms:, tenant_id:)
+        SpeechResult.new(
+          audio: raw_result[:audio],
+          duration: raw_result[:duration],
+          format: raw_result[:format],
+          file_size: raw_result[:audio]&.bytesize,
+          characters: raw_result[:characters],
+          text_length: original_text.length,
+          provider: raw_result[:provider],
+          model_id: raw_result[:model],
+          voice_id: resolved_voice_id,
+          voice_name: raw_result[:voice],
+          duration_ms: duration_ms,
+          started_at: started_at,
+          completed_at: completed_at,
+          total_cost: calculate_cost(raw_result),
+          status: :success,
+          tenant_id: tenant_id
+        )
+      end
+
+      # Calculates cost for speech synthesis
+      def calculate_cost(raw_result)
+        characters = raw_result[:characters] || 0
+
+        if raw_result[:raw_response].respond_to?(:cost) && raw_result[:raw_response].cost
+          return raw_result[:raw_response].cost
+        end
+
+        provider = raw_result[:provider]
+        model_name = raw_result[:model].to_s
+
+        price_per_1k_chars = case provider
+                            when :openai
+                              model_name.include?("hd") ? 0.030 : 0.015
+                            when :elevenlabs
+                              0.30
+                            when :google
+                              0.016
+                            when :polly
+                              0.016
+                            else
+                              0.015
+                            end
+
+        (characters / 1000.0) * price_per_1k_chars
+      end
+
+      # Resolves the provider to use
+      def resolved_provider
+        @options[:provider] || self.class.provider
+      end
+
+      # Resolves the model to use
+      def resolved_model
+        @model || self.class.model
+      end
+
+      # Resolves the voice to use
+      def resolved_voice
+        @options[:voice] || self.class.voice
+      end
+
+      # Resolves the voice ID to use
+      def resolved_voice_id
+        @options[:voice_id] || self.class.voice_id
+      end
+
+      # Resolves the speed to use
+      def resolved_speed
+        @options[:speed] || self.class.speed
+      end
+
+      # Resolves the output format to use
+      def resolved_output_format
+        @options[:format] || self.class.output_format
+      end
+
+      # Returns whether streaming is enabled
+      def streaming_enabled?
+        @runtime_streaming || self.class.streaming?
+      end
+    end
+  end
+end