From e882cea47e76c50db674baf8dae0fa8744aabc4c Mon Sep 17 00:00:00 2001
From: Jeff Cooper <jeff@thejeffcooper.com>
Date: Tue, 3 Sep 2024 23:48:13 -0400
Subject: [PATCH] Voice assist improvement - configurable conversation_id
 timeout (#7385)

---
 esphome/components/voice_assistant/__init__.py         | 6 ++++++
 esphome/components/voice_assistant/voice_assistant.cpp | 8 +++++++-
 esphome/components/voice_assistant/voice_assistant.h   | 3 +++
 tests/components/voice_assistant/test.esp32-ard.yaml   | 1 +
 4 files changed, 17 insertions(+), 1 deletion(-)

diff --git a/esphome/components/voice_assistant/__init__.py b/esphome/components/voice_assistant/__init__.py
index 031edbf27a..a4fb572208 100644
--- a/esphome/components/voice_assistant/__init__.py
+++ b/esphome/components/voice_assistant/__init__.py
@@ -43,6 +43,8 @@ CONF_VOLUME_MULTIPLIER = "volume_multiplier"
 
 CONF_WAKE_WORD = "wake_word"
 
+CONF_CONVERSATION_TIMEOUT = "conversation_timeout"
+
 CONF_ON_TIMER_STARTED = "on_timer_started"
 CONF_ON_TIMER_UPDATED = "on_timer_updated"
 CONF_ON_TIMER_CANCELLED = "on_timer_cancelled"
@@ -100,6 +102,9 @@ CONFIG_SCHEMA = cv.All(
                 cv.float_with_unit("decibel full scale", "(dBFS|dbfs|DBFS)"),
                 cv.int_range(0, 31),
             ),
+            cv.Optional(
+                CONF_CONVERSATION_TIMEOUT, default="300s"
+            ): cv.positive_time_period_milliseconds,
             cv.Optional(CONF_VOLUME_MULTIPLIER, default=1.0): cv.float_range(
                 min=0.0, min_included=False
             ),
@@ -182,6 +187,7 @@ async def to_code(config):
     cg.add(var.set_noise_suppression_level(config[CONF_NOISE_SUPPRESSION_LEVEL]))
     cg.add(var.set_auto_gain(config[CONF_AUTO_GAIN]))
     cg.add(var.set_volume_multiplier(config[CONF_VOLUME_MULTIPLIER]))
+    cg.add(var.set_conversation_timeout(config[CONF_CONVERSATION_TIMEOUT]))
 
     if CONF_ON_LISTENING in config:
         await automation.build_automation(
diff --git a/esphome/components/voice_assistant/voice_assistant.cpp b/esphome/components/voice_assistant/voice_assistant.cpp
index e4f388db68..43c7428858 100644
--- a/esphome/components/voice_assistant/voice_assistant.cpp
+++ b/esphome/components/voice_assistant/voice_assistant.cpp
@@ -171,6 +171,11 @@ void VoiceAssistant::deallocate_buffers_() {
 #endif
 }
 
+void VoiceAssistant::reset_conversation_id() {
+  this->conversation_id_ = "";
+  ESP_LOGD(TAG, "reset conversation ID");
+}
+
 int VoiceAssistant::read_microphone_() {
   size_t bytes_read = 0;
   if (this->mic_->is_running()) {  // Read audio into input buffer
@@ -299,7 +304,8 @@ void VoiceAssistant::loop() {
         break;
       }
       this->set_state_(State::STARTING_PIPELINE);
-      this->set_timeout("reset-conversation_id", 5 * 60 * 1000, [this]() { this->conversation_id_ = ""; });
+      this->set_timeout("reset-conversation_id", this->conversation_timeout_,
+                        [this]() { this->reset_conversation_id(); });
       break;
     }
     case State::STARTING_PIPELINE: {
diff --git a/esphome/components/voice_assistant/voice_assistant.h b/esphome/components/voice_assistant/voice_assistant.h
index a160972e22..88cb0dd413 100644
--- a/esphome/components/voice_assistant/voice_assistant.h
+++ b/esphome/components/voice_assistant/voice_assistant.h
@@ -147,6 +147,8 @@ class VoiceAssistant : public Component {
   }
   void set_auto_gain(uint8_t auto_gain) { this->auto_gain_ = auto_gain; }
   void set_volume_multiplier(float volume_multiplier) { this->volume_multiplier_ = volume_multiplier; }
+  void set_conversation_timeout(uint32_t conversation_timeout) { this->conversation_timeout_ = conversation_timeout; }
+  void reset_conversation_id();
 
   Trigger<> *get_intent_end_trigger() const { return this->intent_end_trigger_; }
   Trigger<> *get_intent_start_trigger() const { return this->intent_start_trigger_; }
@@ -262,6 +264,7 @@ class VoiceAssistant : public Component {
   uint8_t noise_suppression_level_;
   uint8_t auto_gain_;
   float volume_multiplier_;
+  uint32_t conversation_timeout_;
 
   uint8_t *send_buffer_;
   int16_t *input_buffer_;
diff --git a/tests/components/voice_assistant/test.esp32-ard.yaml b/tests/components/voice_assistant/test.esp32-ard.yaml
index 2e0209311d..7f6fd85303 100644
--- a/tests/components/voice_assistant/test.esp32-ard.yaml
+++ b/tests/components/voice_assistant/test.esp32-ard.yaml
@@ -33,6 +33,7 @@ speaker:
 voice_assistant:
   microphone: mic_id_external
   speaker: speaker_id
+  conversation_timeout: 60s
   on_listening:
     - logger.log: "Voice assistant microphone listening"
   on_start: