devttl/xiaozhi-esp32
1
0
Fork 0
forked from xiaozhi/xiaozhi-esp32
Code Pull Requests Activity

Audio stream packet with timestamp

Browse Source
This commit is contained in:
Xiaoxia
2025-04-28 23:10:24 +08:00
parent 7fdf78408e
commit b804343d30
8 changed files with 70 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
main/protocols/mqtt_protocol.cc
View File

@@ -121,24 +121,24 @@ bool MqttProtocol::SendText(const std::string& text) {
return true;
}
void MqttProtocol::SendAudio(const std::vector<uint8_t>& data) {
void MqttProtocol::SendAudio(const AudioStreamPacket& packet) {
std::lock_guard<std::mutex> lock(channel_mutex_);
if (udp_ == nullptr) {
return;
}
std::string nonce(aes_nonce_);
*(uint16_t*)&nonce[2] = htons(data.size());
*(uint16_t*)&nonce[2] = htons(packet.payload.size());
*(uint32_t*)&nonce[12] = htonl(++local_sequence_);
std::string encrypted;
encrypted.resize(aes_nonce_.size() + data.size());
encrypted.resize(aes_nonce_.size() + packet.payload.size());
memcpy(encrypted.data(), nonce.data(), nonce.size());
size_t nc_off = 0;
uint8_t stream_block[16] = {0};
if (mbedtls_aes_crypt_ctr(&aes_ctx_, data.size(), &nc_off, (uint8_t*)nonce.c_str(), stream_block,
(uint8_t*)data.data(), (uint8_t*)&encrypted[nonce.size()]) != 0) {
if (mbedtls_aes_crypt_ctr(&aes_ctx_, packet.payload.size(), &nc_off, (uint8_t*)nonce.c_str(), stream_block,
(uint8_t*)packet.payload.data(), (uint8_t*)&encrypted[nonce.size()]) != 0) {
ESP_LOGE(TAG, "Failed to encrypt audio data");
return;
}
@@ -229,20 +229,20 @@ bool MqttProtocol::OpenAudioChannel() {
ESP_LOGW(TAG, "Received audio packet with wrong sequence: %lu, expected: %lu", sequence, remote_sequence_ + 1);
}
std::vector<uint8_t> decrypted;
size_t decrypted_size = data.size() - aes_nonce_.size();
size_t nc_off = 0;
uint8_t stream_block[16] = {0};
decrypted.resize(decrypted_size);
auto nonce = (uint8_t*)data.data();
auto encrypted = (uint8_t*)data.data() + aes_nonce_.size();
int ret = mbedtls_aes_crypt_ctr(&aes_ctx_, decrypted_size, &nc_off, nonce, stream_block, encrypted, (uint8_t*)decrypted.data());
AudioStreamPacket packet;
packet.payload.resize(decrypted_size);
int ret = mbedtls_aes_crypt_ctr(&aes_ctx_, decrypted_size, &nc_off, nonce, stream_block, encrypted, (uint8_t*)packet.payload.data());
if (ret != 0) {
ESP_LOGE(TAG, "Failed to decrypt audio data, ret: %d", ret);
return;
}
if (on_incoming_audio_ != nullptr) {
on_incoming_audio_(std::move(decrypted));
on_incoming_audio_(std::move(packet));
}
remote_sequence_ = sequence;
last_incoming_time_ = std::chrono::steady_clock::now();

2
main/protocols/mqtt_protocol.h
View File

@@ -26,7 +26,7 @@ public:
~MqttProtocol();
bool Start() override;
void SendAudio(const std::vector<uint8_t>& data) override;
void SendAudio(const AudioStreamPacket& packet) override;
bool OpenAudioChannel() override;
void CloseAudioChannel() override;
bool IsAudioChannelOpened() const override;

2
main/protocols/protocol.cc
View File

@@ -8,7 +8,7 @@ void Protocol::OnIncomingJson(std::function<void(const cJSON* root)> callback) {
on_incoming_json_ = callback;
}
void Protocol::OnIncomingAudio(std::function<void(std::vector<uint8_t>&& data)> callback) {
void Protocol::OnIncomingAudio(std::function<void(AudioStreamPacket&& packet)> callback) {
on_incoming_audio_ = callback;
}

8
main/protocols/protocol.h
View File

@@ -8,7 +8,7 @@
#include <vector>
struct AudioStreamPacket {
uint32_t timestamp;
uint32_t timestamp = 0;
std::vector<uint8_t> payload;
};
@@ -53,7 +53,7 @@ public:
return session_id_;
}
void OnIncomingAudio(std::function<void(std::vector<uint8_t>&& data)> callback);
void OnIncomingAudio(std::function<void(AudioStreamPacket&& packet)> callback);
void OnIncomingJson(std::function<void(const cJSON* root)> callback);
void OnAudioChannelOpened(std::function<void()> callback);
void OnAudioChannelClosed(std::function<void()> callback);
@@ -64,7 +64,7 @@ public:
virtual void CloseAudioChannel() = 0;
virtual bool IsAudioChannelOpened() const = 0;
virtual bool IsAudioChannelBusy() const;
virtual void SendAudio(const std::vector<uint8_t>& data) = 0;
virtual void SendAudio(const AudioStreamPacket& packet) = 0;
virtual void SendWakeWordDetected(const std::string& wake_word);
virtual void SendStartListening(ListeningMode mode);
virtual void SendStopListening();
@@ -74,7 +74,7 @@ public:
protected:
std::function<void(const cJSON* root)> on_incoming_json_;
std::function<void(std::vector<uint8_t>&& data)> on_incoming_audio_;
std::function<void(AudioStreamPacket&& packet)> on_incoming_audio_;
std::function<void()> on_audio_channel_opened_;
std::function<void()> on_audio_channel_closed_;
std::function<void(const std::string& message)> on_network_error_;

45
main/protocols/websocket_protocol.cc
View File

@@ -28,40 +28,40 @@ bool WebsocketProtocol::Start() {
return true;
}
void WebsocketProtocol::SendAudio(const std::vector<uint8_t>& data) {
void WebsocketProtocol::SendAudio(const AudioStreamPacket& packet) {
if (websocket_ == nullptr) {
return;
}
if (version_ == 2) {
std::string packet;
packet.resize(sizeof(BinaryProtocol2) + data.size());
auto bp2 = (BinaryProtocol2*)packet.data();
std::string serialized;
serialized.resize(sizeof(BinaryProtocol2) + packet.payload.size());
auto bp2 = (BinaryProtocol2*)serialized.data();
bp2->version = htons(version_);
bp2->type = 0;
bp2->reserved = 0;
bp2->timestamp = htonl(0);
bp2->payload_size = htonl(data.size());
memcpy(bp2->payload, data.data(), data.size());
bp2->timestamp = htonl(packet.timestamp);
bp2->payload_size = htonl(packet.payload.size());
memcpy(bp2->payload, packet.payload.data(), packet.payload.size());
busy_sending_audio_ = true;
websocket_->Send(packet.data(), packet.size(), true);
websocket_->Send(serialized.data(), serialized.size(), true);
busy_sending_audio_ = false;
} else if (version_ == 3) {
std::string packet;
packet.resize(sizeof(BinaryProtocol3) + data.size());
auto bp3 = (BinaryProtocol3*)packet.data();
std::string serialized;
serialized.resize(sizeof(BinaryProtocol3) + packet.payload.size());
auto bp3 = (BinaryProtocol3*)serialized.data();
bp3->type = 0;
bp3->reserved = 0;
bp3->payload_size = htons(data.size());
memcpy(bp3->payload, data.data(), data.size());
bp3->payload_size = htons(packet.payload.size());
memcpy(bp3->payload, packet.payload.data(), packet.payload.size());
busy_sending_audio_ = true;
websocket_->Send(packet.data(), packet.size(), true);
websocket_->Send(serialized.data(), serialized.size(), true);
busy_sending_audio_ = false;
} else {
busy_sending_audio_ = true;
websocket_->Send(data.data(), data.size(), true);
websocket_->Send(packet.payload.data(), packet.payload.size(), true);
busy_sending_audio_ = false;
}
}
@@ -130,15 +130,24 @@ bool WebsocketProtocol::OpenAudioChannel() {
bp2->timestamp = ntohl(bp2->timestamp);
bp2->payload_size = ntohl(bp2->payload_size);
auto payload = (uint8_t*)bp2->payload;
on_incoming_audio_(std::vector<uint8_t>(payload, payload + bp2->payload_size));
on_incoming_audio_(AudioStreamPacket{
.timestamp = bp2->timestamp,
.payload = std::vector<uint8_t>(payload, payload + bp2->payload_size)
});
} else if (version_ == 3) {
BinaryProtocol3* bp3 = (BinaryProtocol3*)data;
bp3->type = bp3->type;
bp3->payload_size = ntohs(bp3->payload_size);
auto payload = (uint8_t*)bp3->payload;
on_incoming_audio_(std::vector<uint8_t>(payload, payload + bp3->payload_size));
on_incoming_audio_(AudioStreamPacket{
.timestamp = 0,
.payload = std::vector<uint8_t>(payload, payload + bp3->payload_size)
});
} else {
on_incoming_audio_(std::vector<uint8_t>((uint8_t*)data, (uint8_t*)data + len));
on_incoming_audio_(AudioStreamPacket{
.timestamp = 0,
.payload = std::vector<uint8_t>((uint8_t*)data, (uint8_t*)data + len)
});
}
}
} else {

2
main/protocols/websocket_protocol.h
View File

@@ -16,7 +16,7 @@ public:
~WebsocketProtocol();
bool Start() override;
void SendAudio(const std::vector<uint8_t>& data) override;
void SendAudio(const AudioStreamPacket& packet) override;
bool OpenAudioChannel() override;
void CloseAudioChannel() override;
bool IsAudioChannelOpened() const override;