github/esphome
2
0
Fork 0
mirror of https://github.com/esphome/esphome.git synced 2026-09-15 09:08:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

[api] Enable zero-copy bytes for VoiceAssistantAudio and other SOURCE_BOTH messages

Browse Source
This commit is contained in:
J. Nick Koston
2026-01-02 10:56:17 -10:00
parent 5d8346cfb6
commit 9906724828
6 changed files with 31 additions and 37 deletions
Download Patch File Download Diff File Expand all files Collapse all files
esphome/components/api/api.proto
+2 -2
View File
@@ -2425,7 +2425,7 @@ message ZWaveProxyFrame {
option (ifdef) = "USE_ZWAVE_PROXY";
option (no_delay) = true;
bytes data = 1 [(pointer_to_buffer) = true];
bytes data = 1;
}
enum ZWaveProxyRequestType {
@@ -2439,5 +2439,5 @@ message ZWaveProxyRequest {
option (ifdef) = "USE_ZWAVE_PROXY";
ZWaveProxyRequestType type = 1;
bytes data = 2 [(pointer_to_buffer) = true];
bytes data = 2;
}
esphome/components/api/api_pb2.cpp
+6 -4
View File
@@ -2527,20 +2527,22 @@ bool VoiceAssistantAudio::decode_varint(uint32_t field_id, ProtoVarInt value) {
}
bool VoiceAssistantAudio::decode_length(uint32_t field_id, ProtoLengthDelimited value) {
switch (field_id) {
case 1:
this->data = value.as_string();
case 1: {
this->data = value.data();
this->data_len = value.size();
break;
}
default:
return false;
}
return true;
}
void VoiceAssistantAudio::encode(ProtoWriteBuffer buffer) const {
buffer.encode_bytes(1, this->data_ptr_, this->data_len_);
buffer.encode_bytes(1, this->data, this->data_len);
buffer.encode_bool(2, this->end);
}
void VoiceAssistantAudio::calculate_size(ProtoSize &size) const {
size.add_length(1, this->data_len_);
size.add_length(1, this->data_len);
size.add_bool(1, this->end);
}
bool VoiceAssistantTimerEventResponse::decode_varint(uint32_t field_id, ProtoVarInt value) {
esphome/components/api/api_pb2.h
+11 -16
View File
@@ -1046,7 +1046,7 @@ class SubscribeLogsRequest final : public ProtoDecodableMessage {
class SubscribeLogsResponse final : public ProtoMessage {
public:
static constexpr uint8_t MESSAGE_TYPE = 29;
static constexpr uint8_t ESTIMATED_SIZE = 11;
static constexpr uint8_t ESTIMATED_SIZE = 21;
#ifdef HAS_PROTO_MESSAGE_DUMP
const char *message_name() const override { return "subscribe_logs_response"; }
#endif
@@ -1069,7 +1069,7 @@ class SubscribeLogsResponse final : public ProtoMessage {
class NoiseEncryptionSetKeyRequest final : public ProtoDecodableMessage {
public:
static constexpr uint8_t MESSAGE_TYPE = 124;
static constexpr uint8_t ESTIMATED_SIZE = 9;
static constexpr uint8_t ESTIMATED_SIZE = 19;
#ifdef HAS_PROTO_MESSAGE_DUMP
const char *message_name() const override { return "noise_encryption_set_key_request"; }
#endif
@@ -1161,7 +1161,7 @@ class HomeassistantActionRequest final : public ProtoMessage {
class HomeassistantActionResponse final : public ProtoDecodableMessage {
public:
static constexpr uint8_t MESSAGE_TYPE = 130;
static constexpr uint8_t ESTIMATED_SIZE = 24;
static constexpr uint8_t ESTIMATED_SIZE = 34;
#ifdef HAS_PROTO_MESSAGE_DUMP
const char *message_name() const override { return "homeassistant_action_response"; }
#endif
@@ -1388,7 +1388,7 @@ class ListEntitiesCameraResponse final : public InfoResponseProtoMessage {
class CameraImageResponse final : public StateResponseProtoMessage {
public:
static constexpr uint8_t MESSAGE_TYPE = 44;
static constexpr uint8_t ESTIMATED_SIZE = 20;
static constexpr uint8_t ESTIMATED_SIZE = 30;
#ifdef HAS_PROTO_MESSAGE_DUMP
const char *message_name() const override { return "camera_image_response"; }
#endif
@@ -2123,7 +2123,7 @@ class BluetoothGATTReadRequest final : public ProtoDecodableMessage {
class BluetoothGATTReadResponse final : public ProtoMessage {
public:
static constexpr uint8_t MESSAGE_TYPE = 74;
static constexpr uint8_t ESTIMATED_SIZE = 17;
static constexpr uint8_t ESTIMATED_SIZE = 27;
#ifdef HAS_PROTO_MESSAGE_DUMP
const char *message_name() const override { return "bluetooth_gatt_read_response"; }
#endif
@@ -2146,7 +2146,7 @@ class BluetoothGATTReadResponse final : public ProtoMessage {
class BluetoothGATTWriteRequest final : public ProtoDecodableMessage {
public:
static constexpr uint8_t MESSAGE_TYPE = 75;
static constexpr uint8_t ESTIMATED_SIZE = 19;
static constexpr uint8_t ESTIMATED_SIZE = 29;
#ifdef HAS_PROTO_MESSAGE_DUMP
const char *message_name() const override { return "bluetooth_gatt_write_request"; }
#endif
@@ -2182,7 +2182,7 @@ class BluetoothGATTReadDescriptorRequest final : public ProtoDecodableMessage {
class BluetoothGATTWriteDescriptorRequest final : public ProtoDecodableMessage {
public:
static constexpr uint8_t MESSAGE_TYPE = 77;
static constexpr uint8_t ESTIMATED_SIZE = 17;
static constexpr uint8_t ESTIMATED_SIZE = 27;
#ifdef HAS_PROTO_MESSAGE_DUMP
const char *message_name() const override { return "bluetooth_gatt_write_descriptor_request"; }
#endif
@@ -2218,7 +2218,7 @@ class BluetoothGATTNotifyRequest final : public ProtoDecodableMessage {
class BluetoothGATTNotifyDataResponse final : public ProtoMessage {
public:
static constexpr uint8_t MESSAGE_TYPE = 79;
static constexpr uint8_t ESTIMATED_SIZE = 17;
static constexpr uint8_t ESTIMATED_SIZE = 27;
#ifdef HAS_PROTO_MESSAGE_DUMP
const char *message_name() const override { return "bluetooth_gatt_notify_data_response"; }
#endif
@@ -2521,17 +2521,12 @@ class VoiceAssistantEventResponse final : public ProtoDecodableMessage {
class VoiceAssistantAudio final : public ProtoDecodableMessage {
public:
static constexpr uint8_t MESSAGE_TYPE = 106;
static constexpr uint8_t ESTIMATED_SIZE = 11;
static constexpr uint8_t ESTIMATED_SIZE = 21;
#ifdef HAS_PROTO_MESSAGE_DUMP
const char *message_name() const override { return "voice_assistant_audio"; }
#endif
std::string data{};
const uint8_t *data_ptr_{nullptr};
size_t data_len_{0};
void set_data(const uint8_t *data, size_t len) {
this->data_ptr_ = data;
this->data_len_ = len;
}
const uint8_t *data{nullptr};
uint16_t data_len{0};
bool end{false};
void encode(ProtoWriteBuffer buffer) const override;
void calculate_size(ProtoSize &size) const override;
esphome/components/api/api_pb2_dump.cpp
+1 -5
View File
@@ -1978,11 +1978,7 @@ void VoiceAssistantEventResponse::dump_to(std::string &out) const {
void VoiceAssistantAudio::dump_to(std::string &out) const {
MessageDumpHelper helper(out, "VoiceAssistantAudio");
out.append(" data: ");
if (this->data_ptr_ != nullptr) {
out.append(format_hex_pretty(this->data_ptr_, this->data_len_));
} else {
out.append(format_hex_pretty(reinterpret_cast<const uint8_t *>(this->data.data()), this->data.size()));
}
out.append(format_hex_pretty(this->data, this->data_len));
out.append("\n");
dump_field(out, "end", this->end);
}
esphome/components/voice_assistant/voice_assistant.cpp
+8 -7
View File
@@ -272,7 +272,8 @@ void VoiceAssistant::loop() {
size_t read_bytes = this->ring_buffer_->read((void *) this->send_buffer_, SEND_BUFFER_SIZE, 0);
if (this->audio_mode_ == AUDIO_MODE_API) {
api::VoiceAssistantAudio msg;
msg.set_data(this->send_buffer_, read_bytes);
msg.data = this->send_buffer_;
msg.data_len = read_bytes;
this->api_client_->send_message(msg, api::VoiceAssistantAudio::MESSAGE_TYPE);
} else {
if (!this->udp_socket_running_) {
@@ -841,12 +842,12 @@ void VoiceAssistant::on_event(const api::VoiceAssistantEventResponse &msg) {
void VoiceAssistant::on_audio(const api::VoiceAssistantAudio &msg) {
#ifdef USE_SPEAKER // We should never get to this function if there is no speaker anyway
if ((this->speaker_ != nullptr) && (this->speaker_buffer_ != nullptr)) {
if (this->speaker_buffer_index_ + msg.data.length() < SPEAKER_BUFFER_SIZE) {
memcpy(this->speaker_buffer_ + this->speaker_buffer_index_, msg.data.data(), msg.data.length());
this->speaker_buffer_index_ += msg.data.length();
this->speaker_buffer_size_ += msg.data.length();
this->speaker_bytes_received_ += msg.data.length();
ESP_LOGV(TAG, "Received audio: %u bytes from API", msg.data.length());
if (this->speaker_buffer_index_ + msg.data_len < SPEAKER_BUFFER_SIZE) {
memcpy(this->speaker_buffer_ + this->speaker_buffer_index_, msg.data, msg.data_len);
this->speaker_buffer_index_ += msg.data_len;
this->speaker_buffer_size_ += msg.data_len;
this->speaker_bytes_received_ += msg.data_len;
ESP_LOGV(TAG, "Received audio: %u bytes from API", msg.data_len);
} else {
ESP_LOGE(TAG, "Cannot receive audio, buffer is full");
}
script/api_protobuf/api_protobuf.py
+3 -3
View File
@@ -362,12 +362,12 @@ def create_field_type_info(
# Traditional fixed array approach with copy (takes priority)
return FixedArrayBytesType(field, fixed_size)
# For SOURCE_CLIENT only messages (decode but no encode), use pointer
# For messages that decode (SOURCE_CLIENT or SOURCE_BOTH), use pointer
# for zero-copy access to the receive buffer
if needs_decode and not needs_encode:
if needs_decode:
return PointerToBytesBufferType(field, None)
# For SOURCE_BOTH/SOURCE_SERVER, explicit annotation is still needed
# For SOURCE_SERVER (encode only), explicit annotation is still needed
if get_field_opt(field, pb.pointer_to_buffer, False):
return PointerToBytesBufferType(field, None)