refactor: decompose gui.Application god object into focused services

Extract CardService (file I/O, persistence, card directory management) and
GenerationOrchestrator (audio/image/phonetics generation) from the 2852-line
Application struct. Application is now thin UI event-wiring. Also split all
functions over 50 lines into focused helpers throughout app.go.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

1 files changed, 51 insertions, 445 deletions

diff --git a/internal/gui/generator.go b/internal/gui/generator.go
index 4b55608..312baba 100644
--- a/internal/gui/generator.go
+++ b/internal/gui/generator.go
@@ -2,15 +2,9 @@ package gui
 
 import (
 	"context"
-	"fmt"
 	"math/rand"
-	"os"
-	"path/filepath"
-	"strings"
 	"time"
 
-	"fyne.io/fyne/v2"
-
 	"codeberg.org/snonux/totalrecall/internal/audio"
 	"codeberg.org/snonux/totalrecall/internal/image"
 )
@@ -22,489 +16,101 @@ type promptAwareImageClient interface {
 	SetPromptCallback(func(prompt string))
 }
 
+// randomVoice picks a random voice from the provided list.
+// Used by GenerationOrchestrator for both OpenAI and Gemini voice selection.
 func randomVoice(voices []string) string {
 	rng := rand.New(rand.NewSource(time.Now().UnixNano()))
 	return voices[rng.Intn(len(voices))]
 }
 
+// randomOpenAISpeed picks a random speed in [0.90, 1.00) for OpenAI TTS to
+// add slight variation across generations.
 func randomOpenAISpeed() float64 {
 	rng := rand.New(rand.NewSource(time.Now().UnixNano()))
 	return 0.90 + rng.Float64()*0.10
 }
 
-func (a *Application) audioProviderName() string {
-	if a != nil && a.audioConfig != nil {
-		if provider := strings.ToLower(strings.TrimSpace(a.audioConfig.Provider)); provider != "" {
-			return provider
-		}
-	}
-	return audio.DefaultProviderConfig().Provider
-}
-
-// audioVoices returns the voice list for the configured provider.
-func (a *Application) audioVoices() []string {
-	return audio.VoicesFor(a.audioProviderName())
-}
-
-func (a *Application) audioVoiceAndSpeed() (string, float64) {
-	switch a.audioProviderName() {
-	case "gemini":
-		if a.audioConfig != nil {
-			if voice := strings.TrimSpace(a.audioConfig.GeminiVoice); voice != "" {
-				return voice, a.geminiSpeed()
-			}
-		}
-		return randomVoice(a.audioVoices()), a.geminiSpeed()
-	default:
-		return randomVoice(a.audioVoices()), randomOpenAISpeed()
-	}
-}
-
-func (a *Application) geminiSpeed() float64 {
-	if a != nil && a.audioConfig != nil && a.audioConfig.GeminiSpeed > 0 {
-		return a.audioConfig.GeminiSpeed
-	}
-	return audio.DefaultProviderConfig().GeminiSpeed
-}
+// --- Application delegation methods ---
+// Each method delegates to getOrchestrator() so tests that create Application
+// directly (setting newAudioProvider / audioConfig / config) continue to work
+// without modification, while production code uses the pre-built orchestrator.
 
-func (a *Application) geminiVoicePinned() bool {
-	return a != nil && a.audioConfig != nil && strings.TrimSpace(a.audioConfig.GeminiVoice) != ""
+// audioProviderName returns the lowercase TTS provider name.
+func (a *Application) audioProviderName() string {
+	return a.getOrchestrator().audioProviderName()
 }
 
+// audioOutputFormat resolves the effective audio output format.
 func (a *Application) audioOutputFormat() string {
-	if a != nil && a.config != nil && strings.TrimSpace(a.config.AudioFormat) != "" {
-		return a.config.AudioFormat
-	}
-
-	if a != nil && a.audioConfig != nil && strings.TrimSpace(a.audioConfig.OutputFormat) != "" {
-		return a.audioConfig.OutputFormat
-	}
-
-	return audio.DefaultProviderConfig().OutputFormat
+	return a.getOrchestrator().audioOutputFormat()
 }
 
-func (a *Application) audioConfigForGeneration(voice string, speed float64) audio.Config {
-	audioConfig := audio.Config{}
-	if a != nil && a.audioConfig != nil {
-		audioConfig = *a.audioConfig
-	}
-
-	audioConfig.Provider = a.audioProviderName()
-	if a != nil && a.config != nil {
-		audioConfig.OutputDir = a.config.OutputDir
-	}
-	audioConfig.OutputFormat = a.audioOutputFormat()
-
-	switch audioConfig.Provider {
-	case "gemini":
-		audioConfig.GeminiVoice = voice
-		audioConfig.GeminiSpeed = speed
-		if strings.TrimSpace(audioConfig.GeminiTTSModel) == "" {
-			audioConfig.GeminiTTSModel = audio.DefaultProviderConfig().GeminiTTSModel
-		}
-	default:
-		audioConfig.OpenAIVoice = voice
-		audioConfig.OpenAISpeed = speed
-	}
-
-	return audioConfig
-}
-
-func (a *Application) generateAudioFile(ctx context.Context, text, outputFile, voice string, speed float64) error {
-	audioConfig := a.audioConfigForGeneration(voice, speed)
-
-	provider, err := a.newAudioProvider(&audioConfig)
-	if err != nil {
-		return err
-	}
-
-	return provider.GenerateAudio(ctx, text, outputFile)
-}
-
-// translateWord translates a Bulgarian word to English
+// translateWord translates a Bulgarian word to English.
 func (a *Application) translateWord(word string) (string, error) {
-	if a.translator == nil {
-		return "", fmt.Errorf("translation service not configured")
-	}
-
-	return a.translator.TranslateWord(word)
+	return a.getOrchestrator().TranslateWord(word)
 }
 
-// translateEnglishToBulgarian translates an English word to Bulgarian
+// translateEnglishToBulgarian translates an English word to Bulgarian.
 func (a *Application) translateEnglishToBulgarian(word string) (string, error) {
-	if a.translator == nil {
-		return "", fmt.Errorf("translation service not configured")
-	}
-
-	return a.translator.TranslateEnglishToBulgarian(word)
+	return a.getOrchestrator().TranslateEnglishToBulgarian(word)
 }
 
-// generateAudio generates audio for a word
-func (a *Application) generateAudio(ctx context.Context, word string, cardDir string) (string, error) {
-	// Check if this is a regeneration by looking for existing audio file
-	isRegeneration := false
-	if cardDir != "" {
-		audioFile := filepath.Join(cardDir, fmt.Sprintf("audio.%s", a.audioOutputFormat()))
-		if _, err := os.Stat(audioFile); err == nil {
-			isRegeneration = true
-		}
-	}
-
-	voice, speed := a.audioVoiceAndSpeed()
-
-	// Log the audio generation details
-	if isRegeneration {
-		fmt.Printf("Regenerating audio for '%s' with voice: %s, speed: %.2f\n", word, voice, speed)
-	} else {
-		fmt.Printf("Generating audio for '%s' with voice: %s, speed: %.2f\n", word, voice, speed)
-	}
-
-	// Use the provided card directory
-	if cardDir == "" {
-		return "", fmt.Errorf("card directory not provided")
-	}
-
-	// Generate filename in subdirectory
-	outputFile := filepath.Join(cardDir, fmt.Sprintf("audio.%s", a.audioOutputFormat()))
-
-	finalVoice := voice
-	var err error
-	if a.audioProviderName() == "gemini" && !a.geminiVoicePinned() {
-		finalVoice, err = audio.RunWithVoiceFallbacks(voice, func(candidate string) error {
-			if candidate != voice {
-				fmt.Printf("Retrying Gemini audio with voice: %s\n", candidate)
-			}
-			return a.generateAudioFile(ctx, word, outputFile, candidate, speed)
-		}, nil)
-	} else {
-		err = a.generateAudioFile(ctx, word, outputFile, voice, speed)
-	}
-	if err != nil {
-		return "", err
-	}
-
-	audioConfig := a.audioConfigForGeneration(finalVoice, speed)
-
-	// Save audio attribution
-	if err := a.saveAudioAttribution(word, outputFile, finalVoice, speed); err != nil {
-		// Non-fatal error, just log it
-		fmt.Printf("Warning: Failed to save audio attribution: %v\n", err)
-	}
-
-	// Save voice metadata for GUI display
-	if err := a.saveAudioMetadata(cardDir, audioConfig, finalVoice, speed, "en-bg", outputFile, ""); err != nil {
-		fmt.Printf("Warning: Failed to save audio metadata: %v\n", err)
-	}
-
-	return outputFile, nil
+// generateAudio generates audio for an en-bg card's single audio file.
+func (a *Application) generateAudio(ctx context.Context, word, cardDir string) (string, error) {
+	return a.getOrchestrator().GenerateAudio(ctx, word, cardDir)
 }
 
-// generateAudioFront generates front audio for a bg-bg card
-func (a *Application) generateAudioFront(ctx context.Context, word string, cardDir string) (string, error) {
-	if cardDir == "" {
-		return "", fmt.Errorf("card directory not provided")
-	}
-
-	voice, speed := a.audioVoiceAndSpeed()
-	fmt.Printf("Generating front audio for '%s' with voice: %s, speed: %.2f\n", word, voice, speed)
-	frontFile := filepath.Join(cardDir, fmt.Sprintf("audio_front.%s", a.audioOutputFormat()))
-
-	finalVoice := voice
-	var err error
-	if a.audioProviderName() == "gemini" && !a.geminiVoicePinned() {
-		finalVoice, err = audio.RunWithVoiceFallbacks(voice, func(candidate string) error {
-			if candidate != voice {
-				fmt.Printf("Retrying Gemini audio with voice: %s\n", candidate)
-			}
-			return a.generateAudioFile(ctx, word, frontFile, candidate, speed)
-		}, nil)
-	} else {
-		err = a.generateAudioFile(ctx, word, frontFile, voice, speed)
-	}
-	if err != nil {
-		return "", fmt.Errorf("failed to generate front audio: %w", err)
-	}
-
-	audioConfig := a.audioConfigForGeneration(finalVoice, speed)
-
-	if err := a.saveAudioAttribution(word, frontFile, finalVoice, speed); err != nil {
-		fmt.Printf("Warning: Failed to save audio attribution: %v\n", err)
-	}
-
-	// Update metadata
-	if err := a.saveAudioMetadata(cardDir, audioConfig, finalVoice, speed, "bg-bg", frontFile, a.currentAudioFileBack); err != nil {
-		fmt.Printf("Warning: Failed to save audio metadata: %v\n", err)
-	}
-
-	return frontFile, nil
+// generateAudioFront generates the front audio file for a bg-bg card.
+func (a *Application) generateAudioFront(ctx context.Context, word, cardDir string) (string, error) {
+	return a.getOrchestrator().GenerateAudioFront(ctx, word, cardDir)
 }
 
-// generateAudioBack generates back audio for a bg-bg card
-func (a *Application) generateAudioBack(ctx context.Context, text string, cardDir string) (string, error) {
-	if cardDir == "" {
-		return "", fmt.Errorf("card directory not provided")
-	}
-
-	voice, speed := a.audioVoiceAndSpeed()
-	fmt.Printf("Generating back audio for '%s' with voice: %s, speed: %.2f\n", text, voice, speed)
-	backFile := filepath.Join(cardDir, fmt.Sprintf("audio_back.%s", a.audioOutputFormat()))
-
-	finalVoice := voice
-	var err error
-	if a.audioProviderName() == "gemini" && !a.geminiVoicePinned() {
-		finalVoice, err = audio.RunWithVoiceFallbacks(voice, func(candidate string) error {
-			if candidate != voice {
-				fmt.Printf("Retrying Gemini audio with voice: %s\n", candidate)
-			}
-			return a.generateAudioFile(ctx, text, backFile, candidate, speed)
-		}, nil)
-	} else {
-		err = a.generateAudioFile(ctx, text, backFile, voice, speed)
-	}
-	if err != nil {
-		return "", fmt.Errorf("failed to generate back audio: %w", err)
-	}
-
-	audioConfig := a.audioConfigForGeneration(finalVoice, speed)
-
-	if err := a.saveAudioAttribution(text, backFile, finalVoice, speed); err != nil {
-		fmt.Printf("Warning: Failed to save audio attribution: %v\n", err)
-	}
-
-	// Update metadata
-	if err := a.saveAudioMetadata(cardDir, audioConfig, finalVoice, speed, "bg-bg", a.currentAudioFile, backFile); err != nil {
-		fmt.Printf("Warning: Failed to save audio metadata: %v\n", err)
-	}
-
-	return backFile, nil
+// generateAudioBack generates the back audio file for a bg-bg card.
+func (a *Application) generateAudioBack(ctx context.Context, text, cardDir string) (string, error) {
+	return a.getOrchestrator().GenerateAudioBack(ctx, text, cardDir)
 }
 
-// generateAudioBgBg generates audio for both sides of a bg-bg card
+// generateAudioBgBg generates audio for both sides of a bg-bg card.
 func (a *Application) generateAudioBgBg(ctx context.Context, front, back, cardDir string) (string, string, error) {
-	if cardDir == "" {
-		return "", "", fmt.Errorf("card directory not provided")
-	}
-
-	voice, speed := a.audioVoiceAndSpeed()
-
-	// Generate front audio
-	fmt.Printf("Generating front audio for '%s' with voice: %s, speed: %.2f\n", front, voice, speed)
-	frontFile := filepath.Join(cardDir, fmt.Sprintf("audio_front.%s", a.audioOutputFormat()))
-	backFile := filepath.Join(cardDir, fmt.Sprintf("audio_back.%s", a.audioOutputFormat()))
-
-	runPair := func(candidate string) error {
-		if err := a.generateAudioFile(ctx, front, frontFile, candidate, speed); err != nil {
-			return fmt.Errorf("failed to generate front audio: %w", err)
-		}
-
-		fmt.Printf("Generating back audio for '%s' with voice: %s, speed: %.2f\n", back, candidate, speed)
-		if err := a.generateAudioFile(ctx, back, backFile, candidate, speed); err != nil {
-			return fmt.Errorf("failed to generate back audio: %w", err)
-		}
-
-		return nil
-	}
-
-	finalVoice := voice
-	var err error
-	if a.audioProviderName() == "gemini" && !a.geminiVoicePinned() {
-		finalVoice, err = audio.RunWithVoiceFallbacks(voice, func(candidate string) error {
-			if candidate != voice {
-				fmt.Printf("Retrying Gemini audio with voice: %s\n", candidate)
-			}
-			return runPair(candidate)
-		}, nil)
-	} else {
-		err = runPair(voice)
-	}
-	if err != nil {
-		return "", "", err
-	}
-
-	audioConfig := a.audioConfigForGeneration(finalVoice, speed)
-
-	// Save audio attribution
-	if err := a.saveAudioAttribution(front, frontFile, finalVoice, speed); err != nil {
-		fmt.Printf("Warning: Failed to save audio attribution: %v\n", err)
-	}
-	if err := a.saveAudioAttribution(back, backFile, finalVoice, speed); err != nil {
-		fmt.Printf("Warning: Failed to save audio attribution: %v\n", err)
-	}
-
-	// Save metadata for both sides
-	if err := a.saveAudioMetadata(cardDir, audioConfig, finalVoice, speed, "bg-bg", frontFile, backFile); err != nil {
-		fmt.Printf("Warning: Failed to save audio metadata: %v\n", err)
-	}
-
-	return frontFile, backFile, nil
-}
-
-// generateImagesWithPrompt downloads a single image for a word with optional custom prompt and translation
-func (a *Application) generateImagesWithPrompt(ctx context.Context, word string, customPrompt string, translation string, cardDir string) (string, error) {
-	searcher, err := a.newImageSearcher()
-	if err != nil {
-		return "", err
-	}
-
-	// Use the provided card directory
-	if cardDir == "" {
-		return "", fmt.Errorf("card directory not provided")
-	}
-
-	// Create downloader
-	downloadOpts := &image.DownloadOptions{
-		OutputDir:         cardDir,
-		OverwriteExisting: true,
-		CreateDir:         true,
-		FileNamePattern:   "image",
-		MaxSizeBytes:      5 * 1024 * 1024, // 5MB
-	}
-
-	downloader := image.NewDownloader(searcher, downloadOpts)
-
-	// Set up a prompt callback so the GUI and on-disk metadata update as soon as the prompt exists.
-	searcher.SetPromptCallback(a.imagePromptCallback(cardDir, word))
-
-	// Create search options with custom prompt and translation if provided
-	searchOpts := image.DefaultSearchOptions(word)
-	if customPrompt != "" {
-		searchOpts.CustomPrompt = customPrompt
-	}
-	if translation != "" {
-		searchOpts.Translation = translation
-	}
-
-	// Download single image
-	_, path, err := downloader.DownloadBestMatchWithOptions(ctx, searchOpts)
-	if err != nil {
-		return "", err
-	}
-
-	// The prompt has already been saved and UI updated via the callback
-
-	return path, nil
+	return a.getOrchestrator().GenerateAudioBgBg(ctx, front, back, cardDir)
 }
 
-func (a *Application) newImageSearcher() (promptAwareImageClient, error) {
-	switch a.config.ImageProvider {
-	case imageProviderOpenAI:
-		if a.config.OpenAIKey == "" {
-			return nil, fmt.Errorf("OpenAI API key is required for image generation")
-		}
-
-		openaiConfig := &image.OpenAIConfig{
-			APIKey:  a.config.OpenAIKey,
-			Model:   "dall-e-2", // DALL-E 2 supports 512x512
-			Size:    "512x512",  // Half of 1024x1024
-			Quality: "standard",
-			Style:   "natural",
-		}
-
-		return a.newOpenAIImageClient(openaiConfig), nil
-	case imageProviderNanoBanana:
-		config := a.config
-		if config == nil {
-			config = DefaultConfig()
-		}
-		if config.GoogleAPIKey == "" {
-			return nil, fmt.Errorf("google API key is required for image generation")
-		}
-
-		nanoBananaConfig := &image.NanoBananaConfig{
-			APIKey:    config.GoogleAPIKey,
-			Model:     config.NanoBananaModel,
-			TextModel: config.NanoBananaTextModel,
-		}
-
-		return a.newNanoBananaImageClient(nanoBananaConfig), nil
-	default:
-		return nil, fmt.Errorf("unknown image provider: %s", a.config.ImageProvider)
-	}
+// generateAudioFile generates a single audio file using the given voice/speed.
+// Kept as a thin wrapper so the fallback tests in generator_test.go can call it.
+func (a *Application) generateAudioFile(ctx context.Context, text, outputFile, voice string, speed float64) error {
+	return a.getOrchestrator().generateAudioFile(ctx, text, outputFile, voice, speed)
 }
 
-func (a *Application) imagePromptCallback(cardDir, word string) func(prompt string) {
-	return func(prompt string) {
-		// Save the prompt to disk immediately for this word.
-		promptFile := filepath.Join(cardDir, "image_prompt.txt")
-		if err := os.WriteFile(promptFile, []byte(prompt), 0644); err != nil {
-			fmt.Printf("Warning: Failed to save prompt for '%s': %v\n", word, err)
-		}
+// generateImagesWithPrompt downloads a single image for a word with optional
+// custom prompt and translation hint.
+func (a *Application) generateImagesWithPrompt(ctx context.Context, word, customPrompt, translation, cardDir string) (string, error) {
+	o := a.getOrchestrator()
 
-		// Only update UI if this word is still the current word.
+	// Wrap with a UI update callback for the current word's image prompt entry.
+	promptUI := func(prompt string) {
 		a.mu.Lock()
 		isCurrentWord := a.currentWord == word
 		a.mu.Unlock()
 
 		if isCurrentWord && a.imagePromptEntry != nil {
-			fyne.Do(func() {
-				a.imagePromptEntry.SetText(prompt)
-			})
+			a.imagePromptEntry.SetText(prompt)
 		}
 	}
-}
-
-// saveAudioAttribution saves attribution info for generated audio.
-// Uses BuildAttributionFor so no switch on provider name is needed here.
-func (a *Application) saveAudioAttribution(word, audioFile, voice string, speed float64) error {
-	processedText := audio.ProcessedTextForWord(word)
-	providerName := a.audioProviderName()
-
-	// Build an ephemeral Config from the current audioConfig so we can use
-	// AttributionParamsFrom to read provider-specific fields without a switch.
-	cfg := a.audioConfig
-	if cfg == nil {
-		cfg = audio.DefaultProviderConfig()
-	}
-	// Override voice and speed with the values used for this specific generation.
-	cfgCopy := *cfg
-	cfgCopy.Provider = providerName
-	cfgCopy.GeminiVoice = voice
-	cfgCopy.GeminiSpeed = speed
-	cfgCopy.OpenAIVoice = voice
-	cfgCopy.OpenAISpeed = speed
-
-	instruction := audio.InstructionForProvider(providerName, &cfgCopy)
-	params := audio.AttributionParamsFrom(&cfgCopy, word, instruction, processedText, time.Now())
-	attribution := audio.BuildAttributionFor(providerName, params)
-
-	// Save to file
-	attrPath := audio.AttributionPath(audioFile)
-	if err := os.WriteFile(attrPath, []byte(attribution), 0644); err != nil {
-		return fmt.Errorf("failed to write audio attribution file: %w", err)
-	}
 
-	return nil
+	return o.generateImagesWithPromptAndNotify(ctx, word, customPrompt, translation, cardDir, promptUI)
 }
 
-func (a *Application) saveAudioMetadata(cardDir string, audioConfig audio.Config, voice string, speed float64, cardType string, audioFile string, audioFileBack string) error {
-	metadataFile := filepath.Join(cardDir, "audio_metadata.txt")
-	if cardType == "bg-bg" {
-		if audioFile == "" {
-			audioFile, _ = a.resolveBgBgAudioFiles(cardDir)
-		}
-		if audioFileBack == "" {
-			_, audioFileBack = a.resolveBgBgAudioFiles(cardDir)
-		}
-	}
-
-	metadata := audio.BuildSidecarMetadata(audio.SidecarMetadataParams{
-		Provider:          audioConfig.Provider,
-		OutputFormat:      audioConfig.OutputFormat,
-		CardType:          cardType,
-		AudioFile:         audioFile,
-		AudioFileBack:     audioFileBack,
-		OpenAIModel:       audioConfig.OpenAIModel,
-		OpenAIVoice:       voice,
-		OpenAISpeed:       speed,
-		OpenAIInstruction: audioConfig.OpenAIInstruction,
-		GeminiTTSModel:    audioConfig.GeminiTTSModel,
-		GeminiVoice:       voice,
-		GeminiSpeed:       speed,
-	})
+// getPhoneticInfo fetches phonetic information for a Bulgarian word.
+func (a *Application) getPhoneticInfo(word string) (string, error) {
+	return a.getOrchestrator().GetPhoneticInfo(word)
+}
 
-	if err := os.WriteFile(metadataFile, []byte(metadata), 0644); err != nil {
-		return fmt.Errorf("failed to write audio metadata file: %w", err)
-	}
+// saveAudioAttribution saves attribution metadata for a generated audio file.
+func (a *Application) saveAudioAttribution(word, audioFile, voice string, speed float64) error {
+	return a.getOrchestrator().saveAudioAttribution(word, audioFile, voice, speed)
+}
 
-	return nil
+// saveAudioMetadata writes the sidecar metadata file for a generated audio file.
+func (a *Application) saveAudioMetadata(cardDir string, audioCfg audio.Config, voice string, speed float64, cardType, audioFile, audioFileBack string) error {
+	return a.getOrchestrator().saveAudioMetadata(cardDir, audioCfg, voice, speed, cardType, audioFile, audioFileBack)
 }