From ab23ad99832bbfb58f2e22d5be7a82e51b6da535 Mon Sep 17 00:00:00 2001
From: Michael Hansen <hansen.mike@gmail.com>
Date: Mon, 9 Sep 2024 14:19:42 -0500
Subject: [PATCH] Add announce (#64)

* Add announce

* Add finished message

* Add triggers (defer did not work)

* Remove url from VoiceAssistantAnnounceFinished

* Add success flag to VoiceAssistantAnnounceFinished
---
 .../voice_assistant/voice_assistant.cpp       | 22 ++++++++++++++++++-
 .../voice_assistant/voice_assistant.h         |  8 +++++++
 2 files changed, 29 insertions(+), 1 deletion(-)

diff --git a/esphome/components/voice_assistant/voice_assistant.cpp b/esphome/components/voice_assistant/voice_assistant.cpp
index f3f95b2..8038133 100644
--- a/esphome/components/voice_assistant/voice_assistant.cpp
+++ b/esphome/components/voice_assistant/voice_assistant.cpp
@@ -435,6 +435,10 @@ void VoiceAssistant::loop() {
         this->set_timeout("playing", 50, [this]() {
           this->cancel_timeout("speaker-timeout");
           this->set_state_(State::IDLE, State::IDLE);
+
+          api::VoiceAssistantAnnounceFinished msg;
+          msg.success = true;
+          this->api_client_->send_voice_assistant_announce_finished(msg);
         });
       }
       break;
@@ -767,7 +771,11 @@ void VoiceAssistant::on_event(const api::VoiceAssistantEventResponse &msg) {
     }
     case api::enums::VOICE_ASSISTANT_RUN_END: {
       ESP_LOGD(TAG, "Assist Pipeline ended");
-      if (this->state_ == State::STREAMING_MICROPHONE) {
+      if (this->state_ == State::STARTING_PIPELINE) {
+        // Pipeline ended before starting microphone
+        this->set_state_(State::IDLE, State::IDLE);
+      }
+      else if (this->state_ == State::STREAMING_MICROPHONE) {
         this->ring_buffer_->reset();
 #ifdef USE_ESP_ADF
         if (this->use_wake_word_) {
@@ -910,6 +918,18 @@ void VoiceAssistant::timer_tick_() {
   this->timer_tick_trigger_->trigger(res);
 }
 
+void VoiceAssistant::on_announce(const api::VoiceAssistantAnnounceRequest &msg) {
+#ifdef USE_MEDIA_PLAYER
+  if (this->media_player_ != nullptr) {
+    this->tts_start_trigger_->trigger(msg.text);
+    this->media_player_->make_call().set_media_url(msg.media_id).set_announcement(true).perform();
+    this->set_state_(State::STREAMING_RESPONSE, State::STREAMING_RESPONSE);
+    this->tts_end_trigger_->trigger(msg.media_id);
+    this->end_trigger_->trigger();
+  }
+#endif
+}
+
 VoiceAssistant *global_voice_assistant = nullptr;  // NOLINT(cppcoreguidelines-avoid-non-const-global-variables)
 
 }  // namespace voice_assistant
diff --git a/esphome/components/voice_assistant/voice_assistant.h b/esphome/components/voice_assistant/voice_assistant.h
index 16d9eaa..33831b8 100644
--- a/esphome/components/voice_assistant/voice_assistant.h
+++ b/esphome/components/voice_assistant/voice_assistant.h
@@ -43,6 +43,7 @@ enum VoiceAssistantFeature : uint32_t {
   FEATURE_SPEAKER = 1 << 1,
   FEATURE_API_AUDIO = 1 << 2,
   FEATURE_TIMERS = 1 << 3,
+  FEATURE_ANNOUNCE = 1 << 4,
 };
 
 enum class State {
@@ -134,6 +135,12 @@ class VoiceAssistant : public Component {
       flags |= VoiceAssistantFeature::FEATURE_TIMERS;
     }
 
+#ifdef USE_MEDIA_PLAYER
+    if (this->media_player_ != nullptr) {
+      flags |= VoiceAssistantFeature::FEATURE_ANNOUNCE;
+    }
+#endif
+
     return flags;
   }
 
@@ -147,6 +154,7 @@ class VoiceAssistant : public Component {
   void on_event(const api::VoiceAssistantEventResponse &msg);
   void on_audio(const api::VoiceAssistantAudio &msg);
   void on_timer_event(const api::VoiceAssistantTimerEventResponse &msg);
+  void on_announce(const api::VoiceAssistantAnnounceRequest &msg);
 
   bool is_running() const { return this->state_ != State::IDLE; }
   void set_continuous(bool continuous) { this->continuous_ = continuous; }