smart_prompt 0.4.4 → 0.5.0

data/workers/stt_workers.rb

@@ -0,0 +1,195 @@
+# STT Workers for SmartPrompt
+# These workers demonstrate the new speech-to-text capabilities
+
+# Basic speech-to-text worker
+SmartPrompt.define_worker :stt_transcriber do
+  use "stt_service"
+  model "FunAudioLLM/CosyVoice2-0.5B"
+
+  # Prepare parameters for STT transcription
+  stt_params = {
+    audio_file: params[:audio_file],
+    language: params[:language],
+    prompt: params[:prompt],
+    temperature: params[:temperature] || 0.0,
+    response_format: params[:response_format] || "json"
+  }
+
+  # Call the STT adapter directly
+  adapter = engine.llms["stt_service"]
+
+  # Transcribe audio
+  transcription_data = adapter.transcribe_audio(
+    stt_params[:audio_file],
+    model: params[:model],
+    language: stt_params[:language],
+    prompt: stt_params[:prompt],
+    temperature: stt_params[:temperature],
+    response_format: stt_params[:response_format]
+  )
+
+  { transcription: transcription_data }
+end
+
+# URL-based speech-to-text worker
+SmartPrompt.define_worker :stt_url_transcriber do
+  use "stt_service"
+  model "FunAudioLLM/CosyVoice2-0.5B"
+
+  # Prepare parameters for URL-based STT
+  stt_params = {
+    audio_url: params[:audio_url],
+    language: params[:language],
+    prompt: params[:prompt],
+    temperature: params[:temperature] || 0.0,
+    response_format: params[:response_format] || "json"
+  }
+
+  adapter = engine.llms["stt_service"]
+
+  # Transcribe audio from URL
+  transcription_data = adapter.transcribe_audio_url(
+    stt_params[:audio_url],
+    model: params[:model],
+    language: stt_params[:language],
+    prompt: stt_params[:prompt],
+    temperature: stt_params[:temperature],
+    response_format: stt_params[:response_format]
+  )
+
+  { transcription: transcription_data }
+end
+
+# Batch speech-to-text worker
+SmartPrompt.define_worker :batch_stt do
+  use "stt_service"
+  model "FunAudioLLM/CosyVoice2-0.5B"
+
+  # Process multiple audio files
+  audio_files = params[:audio_files] || [params[:audio_file]]
+
+  adapter = engine.llms["stt_service"]
+
+  # Batch transcribe audio files
+  batch_result = adapter.transcribe_batch(
+    audio_files,
+    model: params[:model],
+    language: params[:language],
+    prompt: params[:prompt],
+    temperature: params[:temperature] || 0.0
+  )
+
+  { batch_result: batch_result }
+end
+
+# Audio file information worker
+SmartPrompt.define_worker :audio_info do
+  use "stt_service"
+
+  adapter = engine.llms["stt_service"]
+
+  # Get audio file information
+  audio_info = adapter.get_audio_info(params[:audio_file])
+
+  { audio_info: audio_info }
+end
+
+# Language detection worker
+SmartPrompt.define_worker :language_detector do
+  use "stt_service"
+
+  adapter = engine.llms["stt_service"]
+
+  if params[:audio_file]
+    # Transcribe and detect language
+    transcription_data = adapter.transcribe_audio(
+      params[:audio_file],
+      model: params[:model],
+      language: params[:language],
+      temperature: params[:temperature] || 0.0
+    )
+
+    # Detect language from transcribed text
+    detected_language = adapter.detect_language(transcription_data[:text])
+
+    {
+      transcription: transcription_data,
+      detected_language: detected_language
+    }
+  elsif params[:text]
+    # Detect language from text directly
+    detected_language = adapter.detect_language(params[:text])
+
+    {
+      text: params[:text],
+      detected_language: detected_language
+    }
+  else
+    { error: "Either audio_file or text parameter is required" }
+  end
+end
+
+# Multi-language STT worker
+SmartPrompt.define_worker :multilingual_stt do
+  use "stt_service"
+  model "FunAudioLLM/CosyVoice2-0.5B"
+
+  adapter = engine.llms["stt_service"]
+
+  # First transcribe without language specification
+  transcription_data = adapter.transcribe_audio(
+    params[:audio_file],
+    model: params[:model],
+    temperature: params[:temperature] || 0.0
+  )
+
+  # Detect language from transcribed text
+  detected_language = adapter.detect_language(transcription_data[:text])
+
+  # Re-transcribe with detected language for better accuracy
+  if detected_language && detected_language != "en"
+    improved_transcription = adapter.transcribe_audio(
+      params[:audio_file],
+      model: params[:model],
+      language: detected_language,
+      temperature: params[:temperature] || 0.0
+    )
+
+    {
+      initial_transcription: transcription_data,
+      improved_transcription: improved_transcription,
+      detected_language: detected_language
+    }
+  else
+    {
+      transcription: transcription_data,
+      detected_language: detected_language
+    }
+  end
+end
+
+# Format conversion worker
+SmartPrompt.define_worker :stt_format_converter do
+  use "stt_service"
+  model "FunAudioLLM/CosyVoice2-0.5B"
+
+  adapter = engine.llms["stt_service"]
+
+  # Generate transcriptions in different formats
+  formats = params[:formats] || %w[json text srt vtt]
+  results = {}
+
+  formats.each do |format|
+    transcription_data = adapter.transcribe_audio(
+      params[:audio_file],
+      model: params[:model],
+      language: params[:language],
+      temperature: params[:temperature] || 0.0,
+      response_format: format
+    )
+
+    results[format] = transcription_data
+  end
+
+  { format_results: results }
+end

data/workers/tts_workers.rb

@@ -0,0 +1,388 @@
+# TTS Workers for SmartPrompt
+# These workers demonstrate the new text-to-speech capabilities
+
+# Basic text-to-speech worker
+SmartPrompt.define_worker :tts_synthesizer do
+  use "tts_service"
+  model "FunAudioLLM/CosyVoice2-0.5B"
+
+  # Prepare parameters for TTS synthesis
+  tts_params = {
+    text: params[:text],
+    voice: params[:voice] || "alloy",
+    speed: params[:speed] || 1.0,
+    response_format: params[:response_format] || "mp3",
+    language: params[:language]
+  }
+
+  # Call the TTS adapter directly
+  adapter = engine.llms["tts_service"]
+
+  if params[:save_to_file]
+    # Synthesize and save to file
+    output_dir = params[:output_dir] || "./generated_audio"
+    filename_prefix = params[:filename_prefix] || "tts_audio"
+    output_path = File.join(output_dir, "#{filename_prefix}_#{Time.now.to_i}.#{tts_params[:response_format]}")
+
+    result = adapter.synthesize_to_file(
+      tts_params[:text],
+      output_path,
+      voice: tts_params[:voice],
+      model: params[:model],
+      speed: tts_params[:speed],
+      response_format: tts_params[:response_format],
+      language: tts_params[:language]
+    )
+
+    { audio_file: result }
+  else
+    # Synthesize and return audio data
+    audio_data = adapter.synthesize_speech(
+      tts_params[:text],
+      voice: tts_params[:voice],
+      model: params[:model],
+      speed: tts_params[:speed],
+      response_format: tts_params[:response_format],
+      language: tts_params[:language]
+    )
+
+    { audio_data: audio_data }
+  end
+end
+
+# Multi-language TTS worker
+SmartPrompt.define_worker :multilingual_tts do
+  use "tts_service"
+  model "FunAudioLLM/CosyVoice2-0.5B"
+
+  # Prepare parameters for multilingual TTS
+  tts_params = {
+    text: params[:text],
+    voice: params[:voice] || "alloy",
+    speed: params[:speed] || 1.0,
+    response_format: params[:response_format] || "mp3",
+    language: params[:language]
+  }
+
+  # Auto-detect language if not specified
+  unless tts_params[:language]
+    # Simple language detection based on text content
+    if params[:text] =~ /[\u4e00-\u9fff]/
+      tts_params[:language] = "zh"
+    elsif params[:text] =~ /[\u3040-\u309f\u30a0-\u30ff]/
+      tts_params[:language] = "ja"
+    elsif params[:text] =~ /[\uac00-\ud7af]/
+      tts_params[:language] = "ko"
+    else
+      tts_params[:language] = "en"
+    end
+  end
+
+  adapter = engine.llms["tts_service"]
+
+  if params[:save_to_file]
+    output_dir = params[:output_dir] || "./multilingual_audio"
+    filename_prefix = params[:filename_prefix] || "multilingual_tts"
+    output_path = File.join(output_dir, "#{filename_prefix}_#{tts_params[:language]}_#{Time.now.to_i}.#{tts_params[:response_format]}")
+
+    result = adapter.synthesize_to_file(
+      tts_params[:text],
+      output_path,
+      voice: tts_params[:voice],
+      model: params[:model],
+      speed: tts_params[:speed],
+      response_format: tts_params[:response_format],
+      language: tts_params[:language]
+    )
+
+    {
+      audio_file: result,
+      detected_language: tts_params[:language]
+    }
+  else
+    audio_data = adapter.synthesize_speech(
+      tts_params[:text],
+      voice: tts_params[:voice],
+      model: params[:model],
+      speed: tts_params[:speed],
+      response_format: tts_params[:response_format],
+      language: tts_params[:language]
+    )
+
+    {
+      audio_data: audio_data,
+      detected_language: tts_params[:language]
+    }
+  end
+end
+
+# Voice selection worker
+SmartPrompt.define_worker :voice_selector do
+  use "tts_service"
+
+  adapter = engine.llms["tts_service"]
+
+  # Get available voices
+  available_voices = adapter.available_voices
+
+  # If voice parameter is provided, use it for synthesis
+  if params[:text]
+    tts_params = {
+      text: params[:text],
+      voice: params[:voice] || "alloy",
+      speed: params[:speed] || 1.0,
+      response_format: params[:response_format] || "mp3"
+    }
+
+    if params[:save_to_file]
+      output_dir = params[:output_dir] || "./voice_samples"
+      filename_prefix = params[:filename_prefix] || "voice_sample"
+      output_path = File.join(output_dir, "#{filename_prefix}_#{tts_params[:voice]}_#{Time.now.to_i}.#{tts_params[:response_format]}")
+
+      result = adapter.synthesize_to_file(
+        tts_params[:text],
+        output_path,
+        voice: tts_params[:voice],
+        model: params[:model],
+        speed: tts_params[:speed],
+        response_format: tts_params[:response_format]
+      )
+
+      {
+        available_voices: available_voices,
+        selected_voice: tts_params[:voice],
+        audio_file: result
+      }
+    else
+      audio_data = adapter.synthesize_speech(
+        tts_params[:text],
+        voice: tts_params[:voice],
+        model: params[:model],
+        speed: tts_params[:speed],
+        response_format: tts_params[:response_format]
+      )
+
+      {
+        available_voices: available_voices,
+        selected_voice: tts_params[:voice],
+        audio_data: audio_data
+      }
+    end
+  else
+    # Just return available voices
+    { available_voices: available_voices }
+  end
+end
+
+# Speed variation worker
+SmartPrompt.define_worker :speed_variation_tts do
+  use "tts_service"
+  model "FunAudioLLM/CosyVoice2-0.5B"
+
+  results = []
+
+  # Generate audio at different speeds
+  speeds = params[:speeds] || [0.5, 0.75, 1.0, 1.25, 1.5, 2.0]
+
+  speeds.each do |speed|
+    tts_params = {
+      text: params[:text],
+      voice: params[:voice] || "alloy",
+      speed: speed,
+      response_format: params[:response_format] || "mp3"
+    }
+
+    adapter = engine.llms["tts_service"]
+
+    if params[:save_to_file]
+      output_dir = params[:output_dir] || "./speed_variations"
+      filename_prefix = params[:filename_prefix] || "speed_#{speed.to_s.gsub('.', '_')}"
+      output_path = File.join(output_dir, "#{filename_prefix}_#{Time.now.to_i}.#{tts_params[:response_format]}")
+
+      result = adapter.synthesize_to_file(
+        tts_params[:text],
+        output_path,
+        voice: tts_params[:voice],
+        model: params[:model],
+        speed: tts_params[:speed],
+        response_format: tts_params[:response_format]
+      )
+
+      results << {
+        speed: speed,
+        audio_file: result
+      }
+    else
+      audio_data = adapter.synthesize_speech(
+        tts_params[:text],
+        voice: tts_params[:voice],
+        model: params[:model],
+        speed: tts_params[:speed],
+        response_format: tts_params[:response_format]
+      )
+
+      results << {
+        speed: speed,
+        audio_data: audio_data
+      }
+    end
+  end
+
+  { speed_variations: results }
+end
+
+# Custom voice management worker
+SmartPrompt.define_worker :custom_voice_manager do
+  use "tts_service"
+
+  adapter = engine.llms["tts_service"]
+
+  case params[:action]
+  when "list"
+    # List custom voices
+    custom_voices = adapter.list_custom_voices
+    { action: "list", custom_voices: custom_voices }
+
+  when "create"
+    # Create custom voice
+    if params[:reference_audio_file]
+      voice_data = adapter.create_custom_voice(
+        params[:name],
+        params[:reference_audio_file],
+        description: params[:description]
+      )
+      { action: "create", voice_data: voice_data }
+    else
+      { error: "reference_audio_file is required for creating custom voice" }
+    end
+
+  when "delete"
+    # Delete custom voice
+    if params[:voice_id]
+      result = adapter.delete_custom_voice(params[:voice_id])
+      { action: "delete", result: result }
+    else
+      { error: "voice_id is required for deleting custom voice" }
+    end
+
+  when "synthesize"
+    # Synthesize using custom voice
+    if params[:voice_id] && params[:text]
+      tts_params = {
+        text: params[:text],
+        voice: params[:voice_id], # Use voice_id as custom voice
+        speed: params[:speed] || 1.0,
+        response_format: params[:response_format] || "mp3"
+      }
+
+      if params[:save_to_file]
+        output_dir = params[:output_dir] || "./custom_voice_audio"
+        filename_prefix = params[:filename_prefix] || "custom_voice"
+        output_path = File.join(output_dir, "#{filename_prefix}_#{params[:voice_id]}_#{Time.now.to_i}.#{tts_params[:response_format]}")
+
+        result = adapter.synthesize_to_file(
+          tts_params[:text],
+          output_path,
+          voice: tts_params[:voice],
+          model: params[:model],
+          speed: tts_params[:speed],
+          response_format: tts_params[:response_format]
+        )
+
+        {
+          action: "synthesize",
+          voice_id: params[:voice_id],
+          audio_file: result
+        }
+      else
+        audio_data = adapter.synthesize_speech(
+          tts_params[:text],
+          voice: tts_params[:voice],
+          model: params[:model],
+          speed: tts_params[:speed],
+          response_format: tts_params[:response_format]
+        )
+
+        {
+          action: "synthesize",
+          voice_id: params[:voice_id],
+          audio_data: audio_data
+        }
+      end
+    else
+      { error: "voice_id and text are required for synthesis" }
+    end
+
+  else
+    # Default action: list voices
+    predefined_voices = adapter.available_voices
+    custom_voices = adapter.list_custom_voices
+    {
+      action: "default",
+      predefined_voices: predefined_voices,
+      custom_voices: custom_voices
+    }
+  end
+end
+
+# Batch TTS worker for multiple texts
+SmartPrompt.define_worker :batch_tts do
+  use "tts_service"
+  model "FunAudioLLM/CosyVoice2-0.5B"
+
+  results = []
+
+  # Process multiple texts
+  texts = params[:texts] || [params[:text]]
+
+  texts.each_with_index do |text, index|
+    tts_params = {
+      text: text,
+      voice: params[:voice] || "alloy",
+      speed: params[:speed] || 1.0,
+      response_format: params[:response_format] || "mp3",
+      language: params[:language]
+    }
+
+    adapter = engine.llms["tts_service"]
+
+    if params[:save_to_file]
+      output_dir = params[:output_dir] || "./batch_audio"
+      filename_prefix = params[:filename_prefix] || "batch_tts_#{index}"
+      output_path = File.join(output_dir, "#{filename_prefix}_#{Time.now.to_i}.#{tts_params[:response_format]}")
+
+      result = adapter.synthesize_to_file(
+        tts_params[:text],
+        output_path,
+        voice: tts_params[:voice],
+        model: params[:model],
+        speed: tts_params[:speed],
+        response_format: tts_params[:response_format],
+        language: tts_params[:language]
+      )
+
+      results << {
+        text: text,
+        index: index,
+        audio_file: result
+      }
+    else
+      audio_data = adapter.synthesize_speech(
+        tts_params[:text],
+        voice: tts_params[:voice],
+        model: params[:model],
+        speed: tts_params[:speed],
+        response_format: tts_params[:response_format],
+        language: tts_params[:language]
+      )
+
+      results << {
+        text: text,
+        index: index,
+        audio_data: audio_data
+      }
+    end
+  end
+
+  { batch_results: results }
+end