From 746488cabf2d4ac8db4b0d7883dc72dd8c5aa73c Mon Sep 17 00:00:00 2001
From: Jesse Hills <3060199+jesserockz@users.noreply.github.com>
Date: Wed, 19 Jul 2023 11:38:47 +1200
Subject: [PATCH] Fix silence detection flag on voice assistant (#5120)
---
esphome/components/api/api.proto | 1 +
esphome/components/api/api_connection.cpp | 3 ++-
esphome/components/api/api_connection.h | 2 +-
esphome/components/api/api_pb2.cpp | 9 +++++++++
esphome/components/api/api_pb2.h | 1 +
esphome/components/api/api_server.cpp | 6 +++---
esphome/components/api/api_server.h | 2 +-
esphome/components/voice_assistant/voice_assistant.cpp | 2 +-
esphome/components/voice_assistant/voice_assistant.h | 6 +-----
9 files changed, 20 insertions(+), 12 deletions(-)
diff --git a/esphome/components/api/api.proto b/esphome/components/api/api.proto
index 0d68d9fe55..86685aa5e6 100644
--- a/esphome/components/api/api.proto
+++ b/esphome/components/api/api.proto
@@ -1420,6 +1420,7 @@ message VoiceAssistantRequest {
bool start = 1;
string conversation_id = 2;
+ bool use_vad = 3;
}
message VoiceAssistantResponse {
diff --git a/esphome/components/api/api_connection.cpp b/esphome/components/api/api_connection.cpp
index 858ff0e525..a46efd80e5 100644
--- a/esphome/components/api/api_connection.cpp
+++ b/esphome/components/api/api_connection.cpp
@@ -907,12 +907,13 @@ BluetoothConnectionsFreeResponse APIConnection::subscribe_bluetooth_connections_
#endif
#ifdef USE_VOICE_ASSISTANT
-bool APIConnection::request_voice_assistant(bool start, const std::string &conversation_id) {
+bool APIConnection::request_voice_assistant(bool start, const std::string &conversation_id, bool use_vad) {
if (!this->voice_assistant_subscription_)
return false;
VoiceAssistantRequest msg;
msg.start = start;
msg.conversation_id = conversation_id;
+ msg.use_vad = use_vad;
return this->send_voice_assistant_request(msg);
}
void APIConnection::on_voice_assistant_response(const VoiceAssistantResponse &msg) {
diff --git a/esphome/components/api/api_connection.h b/esphome/components/api/api_connection.h
index c146adff02..acc4578661 100644
--- a/esphome/components/api/api_connection.h
+++ b/esphome/components/api/api_connection.h
@@ -124,7 +124,7 @@ class APIConnection : public APIServerConnection {
void subscribe_voice_assistant(const SubscribeVoiceAssistantRequest &msg) override {
this->voice_assistant_subscription_ = msg.subscribe;
}
- bool request_voice_assistant(bool start, const std::string &conversation_id);
+ bool request_voice_assistant(bool start, const std::string &conversation_id, bool use_vad);
void on_voice_assistant_response(const VoiceAssistantResponse &msg) override;
void on_voice_assistant_event_response(const VoiceAssistantEventResponse &msg) override;
#endif
diff --git a/esphome/components/api/api_pb2.cpp b/esphome/components/api/api_pb2.cpp
index 8c7f6d0c4a..3a2d980e57 100644
--- a/esphome/components/api/api_pb2.cpp
+++ b/esphome/components/api/api_pb2.cpp
@@ -6348,6 +6348,10 @@ bool VoiceAssistantRequest::decode_varint(uint32_t field_id, ProtoVarInt value)
this->start = value.as_bool();
return true;
}
+ case 3: {
+ this->use_vad = value.as_bool();
+ return true;
+ }
default:
return false;
}
@@ -6365,6 +6369,7 @@ bool VoiceAssistantRequest::decode_length(uint32_t field_id, ProtoLengthDelimite
void VoiceAssistantRequest::encode(ProtoWriteBuffer buffer) const {
buffer.encode_bool(1, this->start);
buffer.encode_string(2, this->conversation_id);
+ buffer.encode_bool(3, this->use_vad);
}
#ifdef HAS_PROTO_MESSAGE_DUMP
void VoiceAssistantRequest::dump_to(std::string &out) const {
@@ -6377,6 +6382,10 @@ void VoiceAssistantRequest::dump_to(std::string &out) const {
out.append(" conversation_id: ");
out.append("'").append(this->conversation_id).append("'");
out.append("\n");
+
+ out.append(" use_vad: ");
+ out.append(YESNO(this->use_vad));
+ out.append("\n");
out.append("}");
}
#endif
diff --git a/esphome/components/api/api_pb2.h b/esphome/components/api/api_pb2.h
index 769f7aaff5..627165953d 100644
--- a/esphome/components/api/api_pb2.h
+++ b/esphome/components/api/api_pb2.h
@@ -1655,6 +1655,7 @@ class VoiceAssistantRequest : public ProtoMessage {
public:
bool start{false};
std::string conversation_id{};
+ bool use_vad{false};
void encode(ProtoWriteBuffer buffer) const override;
#ifdef HAS_PROTO_MESSAGE_DUMP
void dump_to(std::string &out) const override;
diff --git a/esphome/components/api/api_server.cpp b/esphome/components/api/api_server.cpp
index 87b5f9e63f..f70d45ecd0 100644
--- a/esphome/components/api/api_server.cpp
+++ b/esphome/components/api/api_server.cpp
@@ -323,16 +323,16 @@ void APIServer::on_shutdown() {
}
#ifdef USE_VOICE_ASSISTANT
-bool APIServer::start_voice_assistant(const std::string &conversation_id) {
+bool APIServer::start_voice_assistant(const std::string &conversation_id, bool use_vad) {
for (auto &c : this->clients_) {
- if (c->request_voice_assistant(true, conversation_id))
+ if (c->request_voice_assistant(true, conversation_id, use_vad))
return true;
}
return false;
}
void APIServer::stop_voice_assistant() {
for (auto &c : this->clients_) {
- if (c->request_voice_assistant(false, ""))
+ if (c->request_voice_assistant(false, "", false))
return;
}
}
diff --git a/esphome/components/api/api_server.h b/esphome/components/api/api_server.h
index be124f42ff..9b40a5ef02 100644
--- a/esphome/components/api/api_server.h
+++ b/esphome/components/api/api_server.h
@@ -81,7 +81,7 @@ class APIServer : public Component, public Controller {
#endif
#ifdef USE_VOICE_ASSISTANT
- bool start_voice_assistant(const std::string &conversation_id);
+ bool start_voice_assistant(const std::string &conversation_id, bool use_vad);
void stop_voice_assistant();
#endif
diff --git a/esphome/components/voice_assistant/voice_assistant.cpp b/esphome/components/voice_assistant/voice_assistant.cpp
index 44d640ff39..217ddb6354 100644
--- a/esphome/components/voice_assistant/voice_assistant.cpp
+++ b/esphome/components/voice_assistant/voice_assistant.cpp
@@ -130,7 +130,7 @@ void VoiceAssistant::start(struct sockaddr_storage *addr, uint16_t port) {
void VoiceAssistant::request_start(bool continuous) {
ESP_LOGD(TAG, "Requesting start...");
- if (!api::global_api_server->start_voice_assistant(this->conversation_id_)) {
+ if (!api::global_api_server->start_voice_assistant(this->conversation_id_, this->silence_detection_)) {
ESP_LOGW(TAG, "Could not request start.");
this->error_trigger_->trigger("not-connected", "Could not request start.");
this->continuous_ = false;
diff --git a/esphome/components/voice_assistant/voice_assistant.h b/esphome/components/voice_assistant/voice_assistant.h
index b103584509..e67baaee65 100644
--- a/esphome/components/voice_assistant/voice_assistant.h
+++ b/esphome/components/voice_assistant/voice_assistant.h
@@ -25,10 +25,9 @@ namespace voice_assistant {
// Version 1: Initial version
// Version 2: Adds raw speaker support
-// Version 3: Adds continuous support
+// Version 3: Unused/skip
static const uint32_t INITIAL_VERSION = 1;
static const uint32_t SPEAKER_SUPPORT = 2;
-static const uint32_t SILENCE_DETECTION_SUPPORT = 3;
class VoiceAssistant : public Component {
public:
@@ -48,9 +47,6 @@ class VoiceAssistant : public Component {
uint32_t get_version() const {
#ifdef USE_SPEAKER
if (this->speaker_ != nullptr) {
- if (this->silence_detection_) {
- return SILENCE_DETECTION_SUPPORT;
- }
return SPEAKER_SUPPORT;
}
#endif