Refactor TTS architecture

internal/config/config.go

@@ -11,11 +11,41 @@ import (
 	"code.chimeric.al/chimerical/odidere/internal/llm"
 	"code.chimeric.al/chimerical/odidere/internal/stt"
 	"code.chimeric.al/chimerical/odidere/internal/tool"
-	"code.chimeric.al/chimerical/odidere/internal/tts"
 
 	"gopkg.in/yaml.v3"
 )
 
+// TTS holds the configuration for TTS.
+type TTS struct {
+	// URL is the base URL of kokoro-fastapi.
+	// For example, "http://localhost:8880".
+	URL string `yaml:"url"`
+	// Voice is the default voice ID to use (e.g., "af_heart").
+	Voice string `yaml:"voice"`
+	// Timeout is the maximum duration for a synthesis request.
+	// Defaults to 60s if empty.
+	Timeout string `yaml:"timeout"`
+	// VoiceMap maps whisper language names to voice IDs.
+	// Used by SelectVoice to auto-select voices based on detected
+	// language.
+	VoiceMap map[string]string `yaml:"voice_map"`
+}
+
+func (cfg TTS) Validate() error {
+	if cfg.URL == "" {
+		return fmt.Errorf("missing URL")
+	}
+	if cfg.Voice == "" {
+		return fmt.Errorf("missing voice")
+	}
+	if cfg.Timeout != "" {
+		if _, err := time.ParseDuration(cfg.Timeout); err != nil {
+			return fmt.Errorf("invalid timeout: %w", err)
+		}
+	}
+	return nil
+}
+
 // Config holds all application configuration sections.
 type Config struct {
 	// Address is the host:port to listen on.
@@ -36,7 +66,7 @@ type Config struct {
 	// Tools defines external commands available to the LLM.
 	Tools []tool.Tool `yaml:"tools"`
 	// TTS configures the text-to-speech client.
-	TTS tts.Config `yaml:"tts"`
+	TTS TTS `yaml:"tts"`
 }
 
 // ApplyDefaults sets default values for optional configuration fields.

internal/service/service.go

@@ -1,5 +1,5 @@
 // Package service orchestrates the odidere voice assistant server.
-// It coordinates the HTTP server, STT/LLM/TTS clients, and handles
+// It coordinates the HTTP server, STT/LLM clients, and handles
 // graceful shutdown.
 package service
 
@@ -26,7 +26,6 @@ import (
 	"code.chimeric.al/chimerical/odidere/internal/service/templates"
 	"code.chimeric.al/chimerical/odidere/internal/stt"
 	"code.chimeric.al/chimerical/odidere/internal/tool"
-	"code.chimeric.al/chimerical/odidere/internal/tts"
 
 	"github.com/google/uuid"
 	"github.com/robfig/cron/v3"
@@ -42,6 +41,19 @@ const (
 	ipKey  key = "ip"
 )
 
+// defaultVoices maps whisper language names to default Kokoro voices.
+var defaultVoices = map[string]string{
+	"chinese":    "zf_xiaobei",
+	"english":    "af_heart",
+	"french":     "ff_siwis",
+	"hindi":      "hf_alpha",
+	"italian":    "if_sara",
+	"japanese":   "jf_alpha",
+	"korean":     "kf_sarah",
+	"portuguese": "pf_dora",
+	"spanish":    "ef_dora",
+}
+
 //go:embed all:static/*
 var static embed.FS
 
@@ -58,7 +70,6 @@ type Service struct {
 	stt    *stt.Client
 	tmpl   *template.Template
 	tools  *tool.Registry
-	tts    *tts.Client
 }
 
 // New creates a Service from the provided configuration.
@@ -91,13 +102,6 @@ func New(cfg *config.Config, log *slog.Logger) (*Service, error) {
 	}
 	svc.llm = llmClient
 
-	// Create TTS client.
-	ttsClient, err := tts.NewClient(cfg.TTS, log)
-	if err != nil {
-		return nil, fmt.Errorf("create TTS client: %v", err)
-	}
-	svc.tts = ttsClient
-
 	// Convert job configs to jobs.
 	jobs := make([]*job.Job, 0, len(cfg.Jobs))
 	for _, jc := range cfg.Jobs {
@@ -133,7 +137,6 @@ func New(cfg *config.Config, log *slog.Logger) (*Service, error) {
 	)
 	svc.mux.HandleFunc("POST /v1/chat/voice", svc.voice)
 	svc.mux.HandleFunc("POST /v1/chat/voice/stream", svc.voiceStream)
-	svc.mux.HandleFunc("GET /v1/voices", svc.voices)
 	svc.mux.HandleFunc("GET /v1/models", svc.models)
 
 	svc.server = &http.Server{
@@ -236,7 +239,7 @@ func (svc *Service) Run(ctx context.Context) error {
 			),
 		),
 		slog.Group(
-			"tts",
+			"tts_url",
 			slog.String("url", svc.cfg.TTS.URL),
 			slog.String("default_voice", svc.cfg.TTS.Voice),
 		),
@@ -352,6 +355,19 @@ func (svc *Service) JobCount() int {
 	return len(svc.jobs)
 }
 
+// selectVoice returns the voice ID for the given language.
+// It uses the configured voice map, falling back to defaults.
+func (svc *Service) selectVoice(lang string) string {
+	voiceMap := svc.cfg.TTS.VoiceMap
+	if len(voiceMap) == 0 {
+		voiceMap = defaultVoices
+	}
+	if voice, ok := voiceMap[lang]; ok {
+		return voice
+	}
+	return ""
+}
+
 func (svc *Service) home(w http.ResponseWriter, r *http.Request) {
 	var (
 		ctx = r.Context()
@@ -364,7 +380,13 @@ func (svc *Service) home(w http.ResponseWriter, r *http.Request) {
 	}
 	w.Header().Set("Content-Type", "text/html; charset=utf-8")
 	if err := svc.tmpl.ExecuteTemplate(
-		w, "index.gohtml", nil,
+		w, "index.gohtml", struct {
+			TTSURL          string
+			TTSDefaultVoice string
+		}{
+			TTSURL:          svc.cfg.TTS.URL,
+			TTSDefaultVoice: svc.cfg.TTS.Voice,
+		},
 	); err != nil {
 		log.ErrorContext(
 			ctx, "template error", slog.Any("error", err),
@@ -398,8 +420,6 @@ type Request struct {
 
 // Response is the response format for chat and voice endpoints.
 type Response struct {
-	// Audio is the base64-encoded WAV audio response.
-	Audio string `json:"audio,omitempty"`
 	// DetectedLanguage is the language detected in the input speech.
 	DetectedLanguage string `json:"detected_language,omitempty"`
 	// Messages is the full list of messages generated during the query,
@@ -565,7 +585,7 @@ func (svc *Service) voice(w http.ResponseWriter, r *http.Request) {
 	// Determine voice to use.
 	var voice = req.Voice
 	if req.Voice == "" && detectedLang != "" {
-		if autoVoice := svc.tts.SelectVoice(
+		if autoVoice := svc.selectVoice(
 			detectedLang,
 		); autoVoice != "" {
 			voice = autoVoice
@@ -581,17 +601,8 @@ func (svc *Service) voice(w http.ResponseWriter, r *http.Request) {
 		)
 	}
 
-	// Generate audio response with selected voice.
-	audio, err := svc.tts.Synthesize(ctx, final.Content, voice)
-	if err != nil {
-		log.ErrorContext(ctx, "TTS failed", slog.Any("error", err))
-		http.Error(w, "TTS error", http.StatusInternalServerError)
-		return
-	}
-
 	w.Header().Set("Content-Type", "application/json")
 	if err := json.NewEncoder(w).Encode(Response{
-		Audio:            base64.StdEncoding.EncodeToString(audio),
 		DetectedLanguage: detectedLang,
 		Messages:         msgs,
 		Model:            model,
@@ -608,8 +619,6 @@ func (svc *Service) voice(w http.ResponseWriter, r *http.Request) {
 
 // StreamMessage is the SSE event payload for the streaming voice endpoint.
 type StreamMessage struct {
-	// Audio is the base64-encoded WAV audio response.
-	Audio string `json:"audio,omitempty"`
 	// DetectedLanguage is the language detected in the input speech.
 	DetectedLanguage string `json:"detected_language,omitempty"`
 	// Error is an error message, if any.
@@ -784,7 +793,7 @@ func (svc *Service) voiceStream(w http.ResponseWriter, r *http.Request) {
 	// Determine voice to use.
 	var voice = req.Voice
 	if req.Voice == "" && detectedLang != "" {
-		if autoVoice := svc.tts.SelectVoice(
+		if autoVoice := svc.selectVoice(
 			detectedLang,
 		); autoVoice != "" {
 			voice = autoVoice
@@ -834,20 +843,6 @@ func (svc *Service) voiceStream(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
-	// Synthesize TTS for the final assistant message.
-	if last.Message.Content != "" {
-		audio, err := svc.tts.Synthesize(
-			ctx, last.Message.Content, voice,
-		)
-		if err != nil {
-			log.ErrorContext(
-				ctx, "TTS failed", slog.Any("error", err),
-			)
-			last.Error = fmt.Sprintf("TTS error: %v", err)
-		} else {
-			last.Audio = base64.StdEncoding.EncodeToString(audio)
-		}
-	}
 	last.Model = model
 	last.Voice = voice
 	send(last)
@@ -890,37 +885,3 @@ func (svc *Service) models(w http.ResponseWriter, r *http.Request) {
 		)
 	}
 }
-
-// voices returns available TTS voices.
-func (svc *Service) voices(w http.ResponseWriter, r *http.Request) {
-	var (
-		ctx = r.Context()
-		log = ctx.Value(logKey).(*slog.Logger)
-	)
-
-	voices, err := svc.tts.ListVoices(ctx)
-	if err != nil {
-		log.ErrorContext(
-			ctx,
-			"failed to list voices",
-			slog.Any("error", err),
-		)
-		http.Error(
-			w,
-			"failed to list voices",
-			http.StatusInternalServerError,
-		)
-		return
-	}
-
-	w.Header().Set("Content-Type", "application/json")
-	if err := json.NewEncoder(w).Encode(map[string][]string{
-		"voices": voices,
-	}); err != nil {
-		log.ErrorContext(
-			ctx,
-			"failed to encode voices response",
-			slog.Any("error", err),
-		)
-	}
-}

internal/service/static/main.css

@@ -144,6 +144,13 @@ body {
   padding: var(--s-2) var(--s-1);
 }
 
+.collapsible__summary::before {
+  content: "▶";
+  font-size: 0.625em;
+  color: var(--color-text-muted);
+  transition: transform 0.15s ease;
+}
+
 .collapsible[open] .collapsible__summary::before {
   transform: rotate(90deg);
 }
@@ -204,13 +211,6 @@ body {
   display: none;
 }
 
-.collapsible__summary::before {
-  content: "▶";
-  font-size: 0.625em;
-  color: var(--color-text-muted);
-  transition: transform 0.15s ease;
-}
-
 .collapsible__summary .icon {
   width: var(--icon-size);
   height: var(--icon-size);
@@ -530,10 +530,24 @@ body {
   flex-direction: column;
 }
 
+.message__icon {
+  width: var(--action-icon-size);
+  height: var(--action-icon-size);
+  flex-shrink: 0;
+  color: var(--color-text-muted);
+}
+
 .message--assistant .message__icon {
   color: var(--color-primary);
 }
 
+.message__debug {
+  display: none;
+  padding: var(--s-2) var(--s-1);
+  border-top: 1px dotted var(--color-border-light);
+  background: var(--color-surface);
+}
+
 .message--debug-open .message__debug {
   display: block;
 }
@@ -579,6 +593,10 @@ body {
   color: var(--color-primary);
 }
 
+.message__action-btn--play.playing {
+  color: var(--color-red);
+}
+
 @media (hover: hover) {
   .message__action-btn--delete:hover {
     color: var(--color-red);
@@ -637,13 +655,6 @@ body {
   border-radius: var(--radius);
 }
 
-.message__debug {
-  display: none;
-  padding: var(--s-2) var(--s-1);
-  border-top: 1px dotted var(--color-border-light);
-  background: var(--color-surface);
-}
-
 .message__debug-list {
   display: grid;
   grid-template-columns: auto 1fr;
@@ -663,13 +674,6 @@ body {
   white-space: nowrap;
 }
 
-.message__icon {
-  width: var(--action-icon-size);
-  height: var(--action-icon-size);
-  flex-shrink: 0;
-  color: var(--color-text-muted);
-}
-
 /* ==================== */
 /* Settings Modal       */
 /* ==================== */

internal/service/static/main.js

@@ -4,7 +4,6 @@ const MODELS_ENDPOINT = '/v1/models';
 const MODEL_KEY = 'odidere_model';
 const STORAGE_KEY = 'odidere_history';
 const SYSTEM_MESSAGE_KEY = 'odidere_system_message';
-const VOICES_ENDPOINT = '/v1/voices';
 const VOICE_KEY = 'odidere_voice';
 
 /**
@@ -38,6 +37,14 @@ class Odidere {
     this.leafId = null;
     this.rootId = null;
 
+    // TTS state
+    this.ttsUrl = document.querySelector('meta[name="tts-url"]')?.content || '';
+    this.ttsDefaultVoice =
+      document.querySelector('meta[name="tts-default-voice"]')?.content || '';
+    this.playQueue = [];
+    this.isPlaying = false;
+    this.currentMessageId = null;
+
     // Auto-scroll: enabled by default, disabled when user scrolls up,
     // re-enabled when they scroll back to bottom or send a new message.
     this.isAutoScrollEnabled = true;
@@ -92,6 +99,7 @@ class Odidere {
       track.stop();
     }
     this.#stopCurrentAudio();
+    this.#stopTTS();
     this.currentController?.abort();
     this.#releaseWakeLock();
   }
@@ -102,6 +110,7 @@ class Odidere {
    */
   reset() {
     this.#stopCurrentAudio();
+    this.#stopTTS();
 
     if (this.currentController) {
       this.currentController.abort();
@@ -234,7 +243,7 @@ class Odidere {
 
   // ====================
   // CONVERSATION TREE
-  // =====================
+  // ====================
 
   /**
    * #loadHistory loads chat history from localStorage and renders it.
@@ -466,7 +475,7 @@ class Odidere {
     let current = startId;
     while (current) {
       const msg = this.messagesMap.get(current);
-      if (!msg || !msg.children || msg.children.length === 0) break;
+      if (!msg?.children || msg.children.length === 0) break;
       // Follow the first child. With branching, this picks the first branch.
       current = msg.children[0];
     }
@@ -811,6 +820,7 @@ class Odidere {
    */
   #stopConversation() {
     this.#stopCurrentAudio();
+    this.#stopTTS();
 
     if (this.currentController) {
       this.currentController.abort();
@@ -833,6 +843,7 @@ class Odidere {
     if (this.isProcessing) return;
 
     this.#stopCurrentAudio();
+    this.#stopTTS();
 
     // Initialize MediaRecorder on first use
     if (!this.mediaRecorder) {
@@ -911,20 +922,92 @@ class Odidere {
   }
 
   // ====================
-  // AUDIO PLAYBACK
+  // TTS PLAYBACK
   // ====================
+
   /**
-   * #playAudio decodes and plays base64-encoded WAV audio.
+   * #enqueueTTS adds a message to the TTS playback queue and starts
-   * @param {string} base64Audio
+   * playback if not already playing.
+   * @param {string} text
+   * @param {string} voice
+   * @param {string} messageId
    */
-  async #playAudio(base64Audio) {
+  #enqueueTTS(text, voice, messageId) {
+    if (!text || !this.ttsUrl) return;
+    this.playQueue.push({ text, voice, messageId });
+    if (!this.isPlaying) {
+      this.#processQueue();
+    }
+  }
+
+  /**
+   * #processQueue pops the next item from the TTS queue and synthesizes it.
+   */
+  async #processQueue() {
+    if (this.playQueue.length === 0) {
+      this.isPlaying = false;
+      this.currentMessageId = null;
+      return;
+    }
+
+    this.isPlaying = true;
+    const item = this.playQueue.shift();
+    this.currentMessageId = item.messageId;
+
+    // Update the play button on the message to show stop state.
+    this.#setPlayButtonState(item.messageId, true);
+
+    await this.#synthesizeAndPlay(item.text, item.voice, item.messageId);
+
+    // Reset this message's button back to play state before moving on.
+    this.#setPlayButtonState(item.messageId, false);
+
+    // Move to next item in queue.
+    this.#processQueue();
+  }
+
+  /**
+   * #stopTTS stops current audio playback, clears the queue, and resets UI.
+   */
+  #stopTTS() {
+    this.#stopCurrentAudio();
+    this.playQueue = [];
+    this.isPlaying = false;
+
+    // Reset all play buttons.
+    if (this.currentMessageId) {
+      this.#setPlayButtonState(this.currentMessageId, false);
+    }
+    this.currentMessageId = null;
+  }
+
+  /**
+   * #synthesizeAndPlay calls kokoro-fastapi to synthesize text and plays
+   * the resulting audio. On error, logs and returns without blocking the queue.
+   * @param {string} text
+   * @param {string} voice
+   * @param {string} _messageId
+   */
+  async #synthesizeAndPlay(text, voice, _messageId) {
     try {
       this.#stopCurrentAudio();
 
-      const audioData = Uint8Array.from(atob(base64Audio), (c) =>
+      const res = await fetch(`${this.ttsUrl}/v1/audio/speech`, {
-        c.charCodeAt(0),
+        method: 'POST',
-      );
+        headers: { 'Content-Type': 'application/json' },
-      const audioBlob = new Blob([audioData], { type: 'audio/wav' });
+        body: JSON.stringify({
+          model: 'tts-1',
+          input: text,
+          voice: voice || this.$voice.value || this.ttsDefaultVoice,
+          response_format: 'wav',
+        }),
+      });
+
+      if (!res.ok) {
+        throw new Error(`TTS error ${res.status}`);
+      }
+
+      const audioBlob = await res.blob();
       const audioUrl = URL.createObjectURL(audioBlob);
 
       this.currentAudioUrl = audioUrl;
@@ -951,11 +1034,59 @@ class Odidere {
         this.currentAudio.play().catch(reject);
       });
     } catch (e) {
-      console.error('failed to play audio:', e);
+      console.error('TTS synthesis failed:', e);
       this.#cleanupAudio();
     }
   }
 
+  /**
+   * #setPlayButtonState updates the play/stop button icon and state for a
+   * given message.
+   * @param {string} messageId
+   * @param {boolean} playing
+   */
+  #setPlayButtonState(messageId, playing) {
+    const $msg = this.$chat.querySelector(`[data-id="${messageId}"]`);
+    if (!$msg) return;
+    const $btn = $msg.querySelector('[data-action="play"]');
+    if (!$btn) return;
+
+    if (playing) {
+      $btn.classList.add('playing');
+      $btn.innerHTML = `<svg class="icon"><use href="${ICONS_URL}#stop"></use></svg>`;
+      $btn.setAttribute('aria-label', 'Stop');
+    } else {
+      $btn.classList.remove('playing');
+      $btn.innerHTML = `<svg class="icon"><use href="${ICONS_URL}#volume"></use></svg>`;
+      $btn.setAttribute('aria-label', 'Play');
+    }
+  }
+
+  /**
+   * #handlePlayClick handles the play/stop button on an assistant message.
+   * If not playing: stop any current audio, enqueue this message, start playing.
+   * If currently playing this message: stop playback, clear remaining queue, reset.
+   * @param {MouseEvent} event
+   * @param {string} messageId
+   * @param {string} text
+   * @param {string} voice
+   */
+  #handlePlayClick(event, messageId, text, voice) {
+    event.stopPropagation();
+
+    if (this.currentMessageId === messageId) {
+      // Currently playing this message: stop everything.
+      this.#stopTTS();
+    } else {
+      // Not playing this message: stop current playback, enqueue this one.
+      this.#stopTTS();
+      this.#enqueueTTS(text, voice, messageId);
+    }
+  }
+
+  // ====================
+  // AUDIO PLAYBACK
+  // ====================
   /**
    * #stopCurrentAudio stops and cleans up any playing audio.
    */
@@ -1002,13 +1133,17 @@ class Odidere {
    */
   async #fetchVoices() {
     try {
-      const res = await fetch(VOICES_ENDPOINT);
+      const res = await fetch(`${this.ttsUrl}/v1/audio/voices`);
       if (!res.ok) throw new Error(`${res.status}`);
       const data = await res.json();
 
       // Filter out legacy v0 voices.
-      const filtered = data.voices.filter((v) => !v.includes('_v0'));
+      const filtered = data.voices.filter((v) => {
-      this.#populateVoiceSelect(filtered);
+        const id = typeof v === 'string' ? v : v.id;
+        return !id.includes('_v0');
+      });
+      const voiceIds = filtered.map((v) => (typeof v === 'string' ? v : v.id));
+      this.#populateVoiceSelect(voiceIds);
     } catch (e) {
       console.error('failed to fetch voices:', e);
       this.$voice.innerHTML = '';
@@ -1034,6 +1169,7 @@ class Odidere {
     if (this.isProcessing) return;
 
     this.#stopCurrentAudio();
+    this.#stopTTS();
     this.$textInput.value = '';
     this.$textInput.style.height = 'auto';
 
@@ -1050,7 +1186,7 @@ class Odidere {
    *
    * Messages are rendered incrementally as SSE events arrive:
    * - user (transcription), assistant (tool calls), tool (results),
-   *   and a final assistant message with synthesized audio.
+   *   and a final assistant message with content.
    *
    * @param {Object} options
    * @param {Blob} [options.audio] - Recorded audio blob
@@ -1234,10 +1370,13 @@ class Odidere {
           const appended = this.#appendHistory([message], meta);
           this.#renderMessages(appended, meta);
 
-          // For the final assistant message with audio, play it.
+          // Enqueue TTS for the final assistant message with content.
-          if (event.audio) {
+          if (message.content) {
-            this.#setLoadingState(false);
+            this.#enqueueTTS(
-            await this.#playAudio(event.audio);
+              message.content,
+              event.voice || this.$voice.value || '',
+              message.id,
+            );
           }
         }
       }
@@ -1766,6 +1905,17 @@ class Odidere {
       $inspect.setAttribute('aria-expanded', String(isOpen));
     });
 
+    // Play button: only for messages with non-empty content.
+    const $play = $msg.querySelector('[data-action="play"]');
+    if (message.content) {
+      const voice = meta?.voice || this.$voice.value || '';
+      $play.addEventListener('click', (e) =>
+        this.#handlePlayClick(e, message.id, message.content, voice),
+      );
+    } else {
+      $play.remove();
+    }
+
     // Assemble text from all available content.
     const copyParts = [];
     if (message.reasoning_content) copyParts.push(message.reasoning_content);

internal/service/templates/static/head.gohtml

@@ -4,6 +4,8 @@
     <meta charset="utf-8">
     <meta name="viewport"
           content="width=device-width, initial-scale=1.0, shrink-to-fit=no, interactive-widget=resizes-content">
+    <meta name="tts-url" content="{{ .TTSURL }}">
+    <meta name="tts-default-voice" content="{{ .TTSDefaultVoice }}">
     <title>Odidere</title>
     <link rel="stylesheet" href="/static/main.css">
     <script defer src="/static/main.js"></script>

internal/service/templates/static/templates.gohtml

@@ -47,6 +47,14 @@
         <div class="message__actions">
           <svg class="message__icon" aria-hidden="true"><use href="/static/icons.svg#assistant"></use></svg>
           <div class="message__actions-buttons">
+            <button
+              type="button"
+              class="message__action-btn message__action-btn--play"
+              data-action="play"
+              aria-label="Play"
+            >
+              <svg class="icon"><use href="/static/icons.svg#volume"></use></svg>
+            </button>
             <button
               type="button"
               class="message__action-btn message__action-btn--delete"

internal/tts/tts.go

@@ -1,208 +0,0 @@
-// Package tts provides a client for kokoro-fastapi text-to-speech.
-package tts
-
-import (
-	"bytes"
-	"context"
-	"encoding/json"
-	"fmt"
-	"io"
-	"log/slog"
-	"net/http"
-	"time"
-)
-
-// Config holds the configuration for a TTS client.
-type Config struct {
-	// URL is the base URL of kokoro-fastapi.
-	// For example, "http://localhost:8880".
-	URL string `yaml:"url"`
-	// Voice is the default voice ID to use (e.g., "af_heart").
-	Voice string `yaml:"voice"`
-	// Timeout is the maximum duration for a synthesis request.
-	// Defaults to 60s if empty.
-	Timeout string `yaml:"timeout"`
-	// VoiceMap maps whisper language names to voice IDs.
-	// Used by SelectVoice to auto-select voices based on detected
-	// language.
-	// If empty, defaultVoices is used.
-	VoiceMap map[string]string `yaml:"voice_map"`
-}
-
-// Validate checks that required configuration values are present and valid.
-func (cfg Config) Validate() error {
-	if cfg.URL == "" {
-		return fmt.Errorf("missing URL")
-	}
-	if cfg.Voice == "" {
-		return fmt.Errorf("missing voice")
-	}
-	if cfg.Timeout != "" {
-		if _, err := time.ParseDuration(cfg.Timeout); err != nil {
-			return fmt.Errorf("invalid timeout: %w", err)
-		}
-	}
-	return nil
-}
-
-// Client wraps an HTTP client for kokoro-fastapi requests.
-type Client struct {
-	client   *http.Client
-	log      *slog.Logger
-	timeout  time.Duration
-	url      string
-	voice    string
-	voiceMap map[string]string
-}
-
-// NewClient creates a new TTS client with the provided configuration.
-func NewClient(cfg Config, log *slog.Logger) (*Client, error) {
-	if err := cfg.Validate(); err != nil {
-		return nil, fmt.Errorf("invalid config: %v", err)
-	}
-
-	timeout := 60 * time.Second
-	if cfg.Timeout != "" {
-		d, err := time.ParseDuration(cfg.Timeout)
-		if err != nil {
-			return nil, fmt.Errorf("parse timeout: %v", err)
-		}
-		timeout = d
-	}
-
-	voiceMap := cfg.VoiceMap
-	if len(voiceMap) == 0 {
-		voiceMap = defaultVoices
-	}
-
-	return &Client{
-		client:   &http.Client{Timeout: timeout},
-		log:      log,
-		timeout:  timeout,
-		url:      cfg.URL,
-		voice:    cfg.Voice,
-		voiceMap: voiceMap,
-	}, nil
-}
-
-// defaultVoices maps whisper language names to default Kokoro voices.
-// Used when Config.VoiceMap is empty.
-var defaultVoices = map[string]string{
-	"chinese":    "zf_xiaobei",
-	"english":    "af_heart",
-	"french":     "ff_siwis",
-	"hindi":      "hf_alpha",
-	"italian":    "if_sara",
-	"japanese":   "jf_alpha",
-	"korean":     "kf_sarah",
-	"portuguese": "pf_dora",
-	"spanish":    "ef_dora",
-}
-
-// SelectVoice returns the voice ID for the given whisper language name.
-// Returns an empty string if no mapping exists for the language.
-func (c *Client) SelectVoice(lang string) string {
-	if voice, ok := c.voiceMap[lang]; ok {
-		return voice
-	}
-	return ""
-}
-
-// Synthesize converts text to speech and returns WAV audio.
-// If voice is empty, the default configured voice is used.
-func (c *Client) Synthesize(ctx context.Context, text, voice string) (
-	[]byte, error,
-) {
-	ctx, cancel := context.WithTimeout(ctx, c.timeout)
-	defer cancel()
-	if voice == "" {
-		voice = c.voice
-	}
-
-	body, err := json.Marshal(struct {
-		Input string `json:"input"`
-		Voice string `json:"voice"`
-	}{
-		Input: text,
-		Voice: voice,
-	})
-	if err != nil {
-		return nil, fmt.Errorf("marshal request: %w", err)
-	}
-
-	req, err := http.NewRequestWithContext(
-		ctx,
-		http.MethodPost,
-		c.url+"/v1/audio/speech",
-		bytes.NewReader(body),
-	)
-	if err != nil {
-		return nil, fmt.Errorf("create request: %w", err)
-	}
-	req.Header.Set("Content-Type", "application/json")
-
-	res, err := c.client.Do(req)
-	if err != nil {
-		return nil, fmt.Errorf("send request: %w", err)
-	}
-	defer res.Body.Close()
-
-	if res.StatusCode != http.StatusOK {
-		body, err := io.ReadAll(res.Body)
-		if err != nil {
-			c.log.ErrorContext(
-				ctx,
-				"failed to read response body",
-				slog.Any("error", err),
-			)
-		}
-		return nil, fmt.Errorf(
-			"tts error %d: %s", res.StatusCode, body,
-		)
-	}
-
-	audio, err := io.ReadAll(res.Body)
-	if err != nil {
-		return nil, fmt.Errorf("read response: %w", err)
-	}
-
-	return audio, nil
-}
-
-// ListVoices returns the available voices from kokoro-fastapi.
-func (c *Client) ListVoices(ctx context.Context) ([]string, error) {
-	ctx, cancel := context.WithTimeout(ctx, c.timeout)
-	defer cancel()
-
-	req, err := http.NewRequestWithContext(
-		ctx,
-		http.MethodGet,
-		c.url+"/v1/audio/voices",
-		nil,
-	)
-	if err != nil {
-		return nil, fmt.Errorf("create request: %w", err)
-	}
-
-	res, err := c.client.Do(req)
-	if err != nil {
-		return nil, fmt.Errorf("send request: %w", err)
-	}
-	defer res.Body.Close()
-
-	if res.StatusCode != http.StatusOK {
-		body, _ := io.ReadAll(res.Body)
-		return nil, fmt.Errorf(
-			"voices error %d: %s", res.StatusCode, body,
-		)
-	}
-
-	var voices = struct {
-		Voices []string `json:"voices"`
-	}{}
-	if err := json.NewDecoder(res.Body).Decode(&voices); err != nil {
-		return nil, fmt.Errorf("decode response: %w", err)
-	}
-
-	return voices.Voices, nil
-}

internal/tts/tts_test.go

@@ -1,236 +0,0 @@
-package tts
-
-import (
-	"testing"
-)
-
-func TestConfigValidate(t *testing.T) {
-	tests := []struct {
-		name    string
-		cfg     Config
-		wantErr bool
-	}{
-		{
-			name:    "empty config",
-			cfg:     Config{},
-			wantErr: true,
-		},
-		{
-			name: "missing URL",
-			cfg: Config{
-				Voice: "af_heart",
-			},
-			wantErr: true,
-		},
-		{
-			name: "missing voice",
-			cfg: Config{
-				URL: "http://localhost:8880",
-			},
-			wantErr: true,
-		},
-		{
-			name: "invalid timeout",
-			cfg: Config{
-				URL:     "http://localhost:8880",
-				Voice:   "af_heart",
-				Timeout: "not-a-duration",
-			},
-			wantErr: true,
-		},
-		{
-			name: "valid minimal config",
-			cfg: Config{
-				URL:   "http://localhost:8880",
-				Voice: "af_heart",
-			},
-			wantErr: false,
-		},
-		{
-			name: "valid config with timeout",
-			cfg: Config{
-				URL:     "http://localhost:8880",
-				Voice:   "af_heart",
-				Timeout: "120s",
-			},
-			wantErr: false,
-		},
-		{
-			name: "valid config with voice map",
-			cfg: Config{
-				URL:   "http://localhost:8880",
-				Voice: "af_heart",
-				VoiceMap: map[string]string{
-					"english": "af_heart",
-					"chinese": "zf_xiaobei",
-				},
-			},
-			wantErr: false,
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			err := tt.cfg.Validate()
-			if (err != nil) != tt.wantErr {
-				t.Errorf(
-					"Validate() error = %v, wantErr %v",
-					err, tt.wantErr,
-				)
-			}
-		})
-	}
-}
-
-func TestNewClient(t *testing.T) {
-	tests := []struct {
-		name    string
-		cfg     Config
-		wantErr bool
-	}{
-		{
-			name:    "invalid config",
-			cfg:     Config{},
-			wantErr: true,
-		},
-		{
-			name: "valid config without timeout",
-			cfg: Config{
-				URL:   "http://localhost:8880",
-				Voice: "af_heart",
-			},
-			wantErr: false,
-		},
-		{
-			name: "valid config with timeout",
-			cfg: Config{
-				URL:     "http://localhost:8880",
-				Voice:   "af_heart",
-				Timeout: "90s",
-			},
-			wantErr: false,
-		},
-		{
-			name: "valid config with custom voice map",
-			cfg: Config{
-				URL:   "http://localhost:8880",
-				Voice: "af_heart",
-				VoiceMap: map[string]string{
-					"english": "custom_en",
-					"french":  "custom_fr",
-				},
-			},
-			wantErr: false,
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			client, err := NewClient(tt.cfg, nil)
-			if (err != nil) != tt.wantErr {
-				t.Errorf(
-					"NewClient() error = %v, wantErr %v",
-					err, tt.wantErr,
-				)
-				return
-			}
-			if !tt.wantErr && client == nil {
-				t.Error("NewClient() returned nil client")
-			}
-		})
-	}
-}
-
-func TestSelectVoice(t *testing.T) {
-	tests := []struct {
-		name     string
-		voiceMap map[string]string
-		lang     string
-		want     string
-	}{
-		{
-			name:     "default map english",
-			voiceMap: nil,
-			lang:     "english",
-			want:     "af_heart",
-		},
-		{
-			name:     "default map chinese",
-			voiceMap: nil,
-			lang:     "chinese",
-			want:     "zf_xiaobei",
-		},
-		{
-			name:     "default map japanese",
-			voiceMap: nil,
-			lang:     "japanese",
-			want:     "jf_alpha",
-		},
-		{
-			name:     "default map unknown language",
-			voiceMap: nil,
-			lang:     "klingon",
-			want:     "",
-		},
-		{
-			name: "custom map",
-			voiceMap: map[string]string{
-				"english": "custom_english",
-				"german":  "custom_german",
-			},
-			lang: "english",
-			want: "custom_english",
-		},
-		{
-			name: "custom map missing language",
-			voiceMap: map[string]string{
-				"english": "custom_english",
-			},
-			lang: "french",
-			want: "",
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			cfg := Config{
-				URL:      "http://localhost:8880",
-				Voice:    "af_heart",
-				VoiceMap: tt.voiceMap,
-			}
-			client, err := NewClient(cfg, nil)
-			if err != nil {
-				t.Fatalf("NewClient() error = %v", err)
-			}
-
-			got := client.SelectVoice(tt.lang)
-			if got != tt.want {
-				t.Errorf(
-					"SelectVoice(%q) = %q, want %q",
-					tt.lang, got, tt.want,
-				)
-			}
-		})
-	}
-}
-
-func TestDefaultVoiceMap(t *testing.T) {
-	// Verify that DefaultVoiceMap contains expected entries.
-	expected := map[string]string{
-		"english":  "af_heart",
-		"chinese":  "zf_xiaobei",
-		"japanese": "jf_alpha",
-		"spanish":  "ef_dora",
-		"french":   "ff_siwis",
-		"korean":   "kf_sarah",
-	}
-
-	for lang, voice := range expected {
-		if got := defaultVoices[lang]; got != voice {
-			t.Errorf(
-				"DefaultVoiceMap[%q] = %q, want %q",
-				lang, got, voice,
-			)
-		}
-	}
-}