3 #ifdef USE_VOICE_ASSISTANT 11 namespace voice_assistant {
13 static const char *
const TAG =
"voice_assistant";
19 static const size_t SAMPLE_RATE_HZ = 16000;
20 static const size_t INPUT_BUFFER_SIZE = 32 * SAMPLE_RATE_HZ / 1000;
21 static const size_t BUFFER_SIZE = 512 * SAMPLE_RATE_HZ / 1000;
22 static const size_t SEND_BUFFER_SIZE = INPUT_BUFFER_SIZE *
sizeof(int16_t);
23 static const size_t RECEIVE_SIZE = 1024;
24 static const size_t SPEAKER_BUFFER_SIZE = 16 * RECEIVE_SIZE;
33 ESP_LOGE(TAG,
"Could not create socket");
38 int err = this->
socket_->setsockopt(SOL_SOCKET, SO_REUSEADDR, &enable,
sizeof(
int));
40 ESP_LOGW(TAG,
"Socket unable to set reuseaddr: errno %d", err);
43 err = this->
socket_->setblocking(
false);
45 ESP_LOGE(TAG,
"Socket unable to set nonblocking mode: errno %d", err);
56 ESP_LOGE(TAG,
"Socket unable to set sockaddr: errno %d", errno);
63 ESP_LOGE(TAG,
"Socket unable to bind: errno %d", errno);
83 ESP_LOGW(TAG,
"Could not allocate speaker buffer");
92 ESP_LOGW(TAG,
"Could not allocate input buffer");
102 ESP_LOGW(TAG,
"Could not allocate ring buffer");
109 ESP_LOGW(TAG,
"Could not allocate send buffer");
122 memset(this->
input_buffer_, 0, INPUT_BUFFER_SIZE *
sizeof(int16_t));
172 ESP_LOGD(TAG,
"reset conversation ID");
176 size_t bytes_read = 0;
179 if (bytes_read == 0) {
180 memset(this->
input_buffer_, 0, INPUT_BUFFER_SIZE *
sizeof(int16_t));
186 ESP_LOGD(TAG,
"microphone not running");
222 ESP_LOGD(TAG,
"Starting Microphone");
246 ESP_LOGD(TAG,
"Waiting for speech...");
252 if (bytes_read > 0) {
253 vad_state_t vad_state =
255 if (vad_state == VAD_SPEECH) {
259 ESP_LOGD(TAG,
"VAD detected speech");
276 ESP_LOGD(TAG,
"Requesting start...");
296 ESP_LOGW(TAG,
"Could not request start");
314 while (available >= SEND_BUFFER_SIZE) {
318 msg.
data.assign((
char *) this->send_buffer_, read_bytes);
328 sizeof(this->dest_addr_));
354 bool playing =
false;
357 ssize_t received_len = 0;
361 if (received_len > 0) {
367 ESP_LOGD(TAG,
"Receive buffer full");
377 ESP_LOGD(TAG,
"End of audio stream received");
386 #ifdef USE_MEDIA_PLAYER 413 ESP_LOGD(TAG,
"Speaker has finished outputting all audio");
442 this->speaker_buffer_size_ -= written;
446 ESP_LOGV(TAG,
"Speaker buffer full, trying again next loop");
456 ESP_LOGE(TAG,
"Client attempting to unsubscribe that is not the current API Client");
465 ESP_LOGE(TAG,
"Multiple API Clients attempting to connect to Voice Assistant");
475 static const LogString *voice_assistant_state_to_string(
State state) {
478 return LOG_STR(
"IDLE");
480 return LOG_STR(
"START_MICROPHONE");
482 return LOG_STR(
"STARTING_MICROPHONE");
484 return LOG_STR(
"WAIT_FOR_VAD");
486 return LOG_STR(
"WAITING_FOR_VAD");
488 return LOG_STR(
"START_PIPELINE");
490 return LOG_STR(
"STARTING_PIPELINE");
492 return LOG_STR(
"STREAMING_MICROPHONE");
494 return LOG_STR(
"STOP_MICROPHONE");
496 return LOG_STR(
"STOPPING_MICROPHONE");
498 return LOG_STR(
"AWAITING_RESPONSE");
500 return LOG_STR(
"STREAMING_RESPONSE");
502 return LOG_STR(
"RESPONSE_FINISHED");
504 return LOG_STR(
"UNKNOWN");
511 ESP_LOGD(TAG,
"State changed from %s to %s", LOG_STR_ARG(voice_assistant_state_to_string(old_state)),
512 LOG_STR_ARG(voice_assistant_state_to_string(state)));
518 ESP_LOGD(TAG,
"Desired state set to %s", LOG_STR_ARG(voice_assistant_state_to_string(desired_state)));
522 ESP_LOGE(TAG,
"Failed to start server. See Home Assistant logs for more details.");
523 this->
error_trigger_->
trigger(
"failed-to-start",
"Failed to start server. See Home Assistant logs for more details.");
533 ESP_LOGD(TAG,
"Client started, streaming microphone");
549 ESP_LOGD(TAG,
"Client started, streaming microphone");
575 ESP_LOGE(TAG,
"No API client connected");
628 ESP_LOGD(TAG,
"Signaling stop...");
635 ESP_LOGD(TAG,
"Event Type: %" PRId32, msg.
event_type);
638 ESP_LOGD(TAG,
"Assist Pipeline running");
644 ESP_LOGD(TAG,
"Wake word detected");
649 ESP_LOGD(TAG,
"STT started");
654 for (
auto arg : msg.
data) {
655 if (arg.name ==
"text") {
656 text = std::move(arg.value);
660 ESP_LOGW(TAG,
"No text in STT_END event");
663 ESP_LOGD(TAG,
"Speech recognised as: \"%s\"", text.c_str());
668 ESP_LOGD(TAG,
"Intent started");
672 for (
auto arg : msg.
data) {
673 if (arg.name ==
"conversation_id") {
682 for (
auto arg : msg.
data) {
683 if (arg.name ==
"text") {
684 text = std::move(arg.value);
688 ESP_LOGW(TAG,
"No text in TTS_START event");
691 ESP_LOGD(TAG,
"Response: \"%s\"", text.c_str());
692 this->
defer([
this, text]() {
704 for (
auto arg : msg.
data) {
705 if (arg.name ==
"url") {
706 url = std::move(arg.value);
710 ESP_LOGW(TAG,
"No url in TTS_END event");
713 ESP_LOGD(TAG,
"Response URL: \"%s\"", url.c_str());
714 this->
defer([
this, url]() {
715 #ifdef USE_MEDIA_PLAYER 727 ESP_LOGD(TAG,
"Assist Pipeline ended");
747 std::string code =
"";
748 std::string message =
"";
749 for (
auto arg : msg.
data) {
750 if (arg.name ==
"code") {
751 code = std::move(arg.value);
752 }
else if (arg.name ==
"message") {
753 message = std::move(arg.value);
756 if (code ==
"wake-word-timeout" || code ==
"wake_word_detection_aborted" || code ==
"no_wake_word") {
759 }
else if (code ==
"wake-provider-missing" || code ==
"wake-engine-missing") {
761 this->
defer([
this, code, message]() {
767 ESP_LOGE(TAG,
"Error: %s - %s", code.c_str(), message.c_str());
779 ESP_LOGD(TAG,
"TTS stream start");
789 ESP_LOGD(TAG,
"TTS stream end");
795 ESP_LOGD(TAG,
"Starting STT by VAD");
799 ESP_LOGD(TAG,
"STT by VAD end");
804 ESP_LOGD(TAG,
"Unhandled event type: %" PRId32, msg.
event_type);
810 #ifdef USE_SPEAKER // We should never get to this function if there is no speaker anyway 817 ESP_LOGV(TAG,
"Received audio: %u bytes from API", msg.
data.length());
819 ESP_LOGE(TAG,
"Cannot receive audio, buffer is full");
834 ESP_LOGD(TAG,
"Timer Event");
835 ESP_LOGD(TAG,
" Type: %" PRId32, msg.
event_type);
836 ESP_LOGD(TAG,
" %s", timer.
to_string().c_str());
865 std::vector<Timer> res;
866 res.reserve(this->
timers_.size());
867 for (
auto &pair : this->
timers_) {
868 auto &timer = pair.second;
869 if (timer.is_active && timer.seconds_left > 0) {
870 timer.seconds_left--;
872 res.push_back(timer);
878 #ifdef USE_MEDIA_PLAYER 894 #endif // USE_VOICE_ASSISTANT Trigger * tts_stream_start_trigger_
virtual size_t play(const uint8_t *data, size_t length)=0
std::string wake_word_phrase
Trigger< Timer > * timer_cancelled_trigger_
bool wait_for_stream_end_
void set_interval(const std::string &name, uint32_t interval, std::function< void()> &&f)
Set an interval function with a unique name.
const float AFTER_CONNECTION
For components that should be initialized after a data connection (API/MQTT) is connected.
std::unordered_map< std::string, Timer > timers_
bool cancel_timeout(const std::string &name)
Cancel a timeout function.
enums::VoiceAssistantTimerEvent event_type
void deallocate_buffers_()
Trigger * intent_end_trigger_
HighFrequencyLoopRequester high_freq_
VoiceAssistant * global_voice_assistant
std::string conversation_id_
socklen_t set_sockaddr_any(struct sockaddr *addr, socklen_t addrlen, uint16_t port)
Set a sockaddr to the any address and specified port for the IP version used by socket_ip().
std::string conversation_id
api::APIConnection * api_client_
Trigger< Timer > * timer_finished_trigger_
float get_setup_priority() const override
vad_handle_t vad_instance_
std::unique_ptr< socket::Socket > socket_
uint32_t noise_suppression_level
Trigger< std::string > * tts_start_trigger_
void set_timeout(const std::string &name, uint32_t timeout, std::function< void()> &&f)
Set a timeout function with a unique name.
bool cancel_interval(const std::string &name)
Cancel an interval function.
void defer(const std::string &name, std::function< void()> &&f)
Defer a callback to the next loop() call.
void deallocate(T *p, size_t n)
Trigger< std::string > * tts_end_trigger_
VoiceAssistantAudioSettings audio_settings
microphone::Microphone * mic_
Trigger< std::vector< Timer > > * timer_tick_trigger_
enums::VoiceAssistantEvent event_type
void client_subscription(api::APIConnection *client, bool subscribe)
virtual bool has_buffered_data() const =0
std::vector< VoiceAssistantEventData > data
std::string get_client_combined_info() const
uint32_t conversation_timeout_
size_t speaker_bytes_received_
uint8_t * speaker_buffer_
uint8_t noise_suppression_level_
void trigger(Ts... x)
Inform the parent automation that the event has triggered.
bool status_has_error() const
Trigger< Timer > * timer_started_trigger_
Trigger * listening_trigger_
void status_set_error(const char *message="unspecified")
media_player::MediaPlayer * media_player_
void start()
Start running the loop continuously.
Trigger * client_connected_trigger_
bool send_voice_assistant_request(const VoiceAssistantRequest &msg)
void reset_conversation_id()
speaker::Speaker * speaker_
void stop()
Stop running the loop continuously.
void set_state_(State state)
Trigger * stt_vad_start_trigger_
bool send_voice_assistant_announce_finished(const VoiceAssistantAnnounceFinished &msg)
std::unique_ptr< RingBuffer > ring_buffer_
bool send_voice_assistant_audio(const VoiceAssistantAudio &msg)
void on_audio(const api::VoiceAssistantAudio &msg)
Trigger * client_disconnected_trigger_
void status_clear_error()
void on_timer_event(const api::VoiceAssistantTimerEventResponse &msg)
Trigger * stt_vad_end_trigger_
struct sockaddr_storage dest_addr_
virtual size_t read(int16_t *buf, size_t len)=0
Trigger * wake_word_detected_trigger_
Trigger * tts_stream_end_trigger_
virtual void mark_failed()
Mark this component as failed.
Implementation of SPI Controller mode.
size_t speaker_buffer_size_
std::string to_string() const
Trigger * intent_start_trigger_
void on_announce(const api::VoiceAssistantAnnounceRequest &msg)
static std::unique_ptr< RingBuffer > create(size_t len)
Trigger< Timer > * timer_updated_trigger_
size_t speaker_buffer_index_
void on_event(const api::VoiceAssistantEventResponse &msg)
Trigger< std::string > * stt_end_trigger_
void request_start(bool continuous, bool silence_detection)
Trigger< std::string, std::string > * error_trigger_
std::unique_ptr< Socket > socket(int domain, int type, int protocol)
Create a socket of the given domain, type and protocol.