RSDK-12150 Add AudioIn Component

src/viam/api/CMakeLists.txt

@@ -132,6 +132,14 @@ if (VIAMCPPSDK_USE_DYNAMIC_PROTOS)
     ${PROTO_GEN_DIR}/component/arm/v1/arm.grpc.pb.h
     ${PROTO_GEN_DIR}/component/arm/v1/arm.pb.cc
     ${PROTO_GEN_DIR}/component/arm/v1/arm.pb.h
+    ${PROTO_GEN_DIR}/component/audioin/v1/audioin.grpc.pb.cc
+    ${PROTO_GEN_DIR}/component/audioin/v1/audioin.grpc.pb.h
+    ${PROTO_GEN_DIR}/component/audioin/v1/audioin.pb.cc
+    ${PROTO_GEN_DIR}/component/audioin/v1/audioin.pb.h
+    ${PROTO_GEN_DIR}/component/audioout/v1/audioout.grpc.pb.cc
+    ${PROTO_GEN_DIR}/component/audioout/v1/audioout.grpc.pb.h
+    ${PROTO_GEN_DIR}/component/audioout/v1/audioout.pb.cc
+    ${PROTO_GEN_DIR}/component/audioout/v1/audioout.pb.h
     ${PROTO_GEN_DIR}/component/base/v1/base.grpc.pb.cc
     ${PROTO_GEN_DIR}/component/base/v1/base.grpc.pb.h
     ${PROTO_GEN_DIR}/component/base/v1/base.pb.cc
@@ -294,6 +302,10 @@ target_sources(viamapi
     ${PROTO_GEN_DIR}/common/v1/common.pb.cc
     ${PROTO_GEN_DIR}/component/arm/v1/arm.grpc.pb.cc
     ${PROTO_GEN_DIR}/component/arm/v1/arm.pb.cc
+    ${PROTO_GEN_DIR}/component/audioin/v1/audioin.grpc.pb.cc
+    ${PROTO_GEN_DIR}/component/audioin/v1/audioin.pb.cc
+    ${PROTO_GEN_DIR}/component/audioout/v1/audioout.grpc.pb.cc
+    ${PROTO_GEN_DIR}/component/audioout/v1/audioout.pb.cc
     ${PROTO_GEN_DIR}/component/base/v1/base.grpc.pb.cc
     ${PROTO_GEN_DIR}/component/base/v1/base.pb.cc
     ${PROTO_GEN_DIR}/component/board/v1/board.grpc.pb.cc
@@ -358,6 +370,10 @@ target_sources(viamapi
       ${PROTO_GEN_DIR}/../../viam/api/common/v1/common.pb.h
       ${PROTO_GEN_DIR}/../../viam/api/component/arm/v1/arm.grpc.pb.h
       ${PROTO_GEN_DIR}/../../viam/api/component/arm/v1/arm.pb.h
+      ${PROTO_GEN_DIR}/../../viam/api/component/audioin/v1/audioin.grpc.pb.h
+      ${PROTO_GEN_DIR}/../../viam/api/component/audioin/v1/audioin.pb.h
+      ${PROTO_GEN_DIR}/../../viam/api/component/audioout/v1/audioout.grpc.pb.h
+      ${PROTO_GEN_DIR}/../../viam/api/component/audioout/v1/audioout.pb.h
       ${PROTO_GEN_DIR}/../../viam/api/component/base/v1/base.grpc.pb.h
       ${PROTO_GEN_DIR}/../../viam/api/component/base/v1/base.pb.h
       ${PROTO_GEN_DIR}/../../viam/api/component/board/v1/board.grpc.pb.h

src/viam/examples/modules/audioin/client.cpp

@@ -1,9 +1,9 @@
-#include <fstream>
 #include <iostream>
 #include <memory>
 #include <string>
 #include <vector>
 
+#include <viam/sdk/common/audio.hpp>
 #include <viam/sdk/common/instance.hpp>
 #include <viam/sdk/common/proto_value.hpp>
 #include <viam/sdk/components/audio_in.hpp>
@@ -18,22 +18,34 @@ int main() {
     // any other C++ SDK objects and stays alive until all Viam C++ SDK objects are destroyed.
     Instance inst;
 
-    // Update these with your robot's connection details
-    const char* uri = "";  // replace with your robot's URI
-    DialOptions dial_options;
-    dial_options.set_allow_insecure_downgrade(true);  // set to false if connecting securely
+    std::string host("xarm-main.aqb785vhl4.viam.cloud");
+    DialOptions dial_opts;
+    dial_opts.set_entity(std::string("88dcef8e-db7f-47dc-9b0f-eb08fdc5a97d"));
 
-    // Uncomment and fill out your credentials details if connecting securely
-    // std::string type = "api-key";
-    // std::string payload = "your-api-key-here";
-    // Credentials credentials(type, payload);
-    // dial_options.set_credentials(credentials);
+    Credentials credentials("api-key", "eou798fi90d1fytv66c9lntv1ndwxj4g");
 
-    boost::optional<DialOptions> opts(dial_options);
-    std::string address(uri);
-    Options options(1, opts);
+    dial_opts.set_credentials(credentials);
+    boost::optional<DialOptions> opts(dial_opts);
+    Options options(0, opts);
 
-    std::shared_ptr<RobotClient> robot = RobotClient::at_address(address, options);
+    auto robot = RobotClient::at_address(host, options);
+
+    // // Update these with your robot's connection details
+    // const char* uri = "";  // replace with your robot's URI
+    // DialOptions dial_options;
+    // dial_options.set_allow_insecure_downgrade(true);  // set to false if connecting securely
+
+    // // Uncomment and fill out your credentials details if connecting securely
+    // // std::string type = "api-key";
+    // // std::string payload = "your-api-key-here";
+    // // Credentials credentials(type, payload);
+    // // dial_options.set_credentials(credentials);
+
+    // boost::optional<DialOptions> opts(dial_options);
+    // std::string address(uri);
+    // Options options(1, opts);
+
+    // std::shared_ptr<RobotClient> robot = RobotClient::at_address(address, options);
 
     // Print resources
     VIAM_SDK_LOG(info) << "Resources:";
@@ -49,9 +61,8 @@ int main() {
         return EXIT_FAILURE;
     }
 
-    // Get audio properties
     VIAM_SDK_LOG(info) << "Getting audio properties...";
-    AudioIn::properties props = audio_in->get_properties();
+    properties props = audio_in->get_properties();
     VIAM_SDK_LOG(info) << "Audio properties:";
     VIAM_SDK_LOG(info) << "  sample_rate_hz: " << props.sample_rate_hz;
     VIAM_SDK_LOG(info) << "  num_channels: " << props.num_channels;
@@ -66,8 +77,7 @@ int main() {
     auto chunk_handler = [&](AudioIn::audio_chunk&& chunk) -> bool {
         chunk_count++;
         VIAM_SDK_LOG(info) << "Received chunk " << chunk_count
-                           << " - length: " << chunk.audio_data.size()
-                           << " bytes, timestamp: " << chunk.start_timestamp_ns;
+                           << " - length: " << chunk.audio_data.size() << " bytes";
 
         for (const auto& byte : chunk.audio_data) {
             all_audio_data.push_back(static_cast<uint8_t>(byte));
@@ -77,52 +87,24 @@ int main() {
     };
 
     // Get 2 seconds of audio (with previous_timestamp = 0 to start from now)
-    audio_in->get_audio("pcm16", chunk_handler, 2.0, 0);
+    audio_in->get_audio(audio_codecs::PCM_16, chunk_handler, 2.0, 0);
 
     VIAM_SDK_LOG(info) << "Total audio data received: " << all_audio_data.size() << " bytes";
     VIAM_SDK_LOG(info) << "Total chunks: " << chunk_count;
 
-    // Save to a WAV file
     std::string filename = "sine_wave_audio.wav";
-    std::ofstream outfile(filename, std::ios::binary);
-    if (!outfile.is_open()) {
-        VIAM_SDK_LOG(error) << "Failed to open file for writing";
+    try {
+        write_wav_file(filename,
+                       all_audio_data,
+                       audio_codecs::PCM_16,
+                       props.sample_rate_hz,
+                       props.num_channels);
+        VIAM_SDK_LOG(info) << "Audio saved to " << filename;
+        VIAM_SDK_LOG(info) << "To play: open " << filename << " (or use any audio player)";
+    } catch (const std::exception& e) {
+        VIAM_SDK_LOG(error) << "Failed to write WAV file: " << e.what();
         return EXIT_FAILURE;
     }
 
-    // WAV file parameters
-    uint32_t sample_rate = props.sample_rate_hz;
-    uint16_t num_channels = props.num_channels;
-    uint16_t bits_per_sample = 16;  // 16-bit PCM
-    uint32_t data_size = all_audio_data.size();
-    uint32_t byte_rate = sample_rate * num_channels * (bits_per_sample / 8);
-    uint16_t block_align = num_channels * (bits_per_sample / 8);
-
-    // Write WAV header
-    // RIFF chunk descriptor
-    outfile.write("RIFF", 4);
-    uint32_t chunk_size = 36 + data_size;
-    outfile.write(reinterpret_cast<const char*>(&chunk_size), 4);
-    outfile.write("WAVE", 4);
-
-    outfile.write("fmt ", 4);
-    uint32_t subchunk1_size = 16;  // PCM
-    outfile.write(reinterpret_cast<const char*>(&subchunk1_size), 4);
-    uint16_t audio_format = 1;  // PCM
-    outfile.write(reinterpret_cast<const char*>(&audio_format), 2);
-    outfile.write(reinterpret_cast<const char*>(&num_channels), 2);
-    outfile.write(reinterpret_cast<const char*>(&sample_rate), 4);
-    outfile.write(reinterpret_cast<const char*>(&byte_rate), 4);
-    outfile.write(reinterpret_cast<const char*>(&block_align), 2);
-    outfile.write(reinterpret_cast<const char*>(&bits_per_sample), 2);
-
-    outfile.write("data", 4);
-    outfile.write(reinterpret_cast<const char*>(&data_size), 4);
-    outfile.write(reinterpret_cast<const char*>(all_audio_data.data()), data_size);
-
-    outfile.close();
-    VIAM_SDK_LOG(info) << "Audio saved to " << filename;
-    VIAM_SDK_LOG(info) << "To play: open " << filename << " (or use any audio player)";
-
     return EXIT_SUCCESS;
 }

src/viam/examples/modules/audioin/main.cpp

@@ -6,6 +6,7 @@
 #include <sstream>
 #include <vector>
 
+#include <viam/sdk/common/audio.hpp>
 #include <viam/sdk/common/exception.hpp>
 #include <viam/sdk/common/instance.hpp>
 #include <viam/sdk/common/proto_value.hpp>
@@ -46,6 +47,15 @@ class SineWaveAudioIn : public AudioIn, public Reconfigurable {
 
    private:
     double frequency_{440.0};
+
+    static double generate_sine_sample(double frequency, double amplitude, double time_seconds);
+    static int16_t float_to_pcm16(double sample_value);
+    static std::vector<uint8_t> pcm16_samples_to_bytes(const std::vector<int16_t>& samples);
+    static audio_chunk create_audio_chunk(const std::vector<int16_t>& samples,
+                                          const std::string& codec,
+                                          int sample_rate_hz,
+                                          int num_channels,
+                                          int sequence_number);
 };
 
 std::vector<std::string> SineWaveAudioIn::validate(const ResourceConfig& cfg) {
@@ -79,9 +89,53 @@ ProtoStruct SineWaveAudioIn::do_command(const ProtoStruct& command) {
     return command;
 }
 
-AudioIn::properties SineWaveAudioIn::get_properties(const ProtoStruct&) {
-    AudioIn::properties props;
-    props.supported_codecs = {"pcm16"};
+// Generates a single audio sample representing a sine wave at the given frequency, amplitude, and
+// time.
+double SineWaveAudioIn::generate_sine_sample(double frequency,
+                                             double amplitude,
+                                             double time_seconds) {
+    return amplitude * std::sin(2.0 * M_PI * frequency * time_seconds);
+}
+
+// Converts a normalized floating-point sample (-1.0 to 1.0) to 16-bit PCM format.
+int16_t SineWaveAudioIn::float_to_pcm16(double sample_value) {
+    return static_cast<int16_t>(sample_value * 32767.0);
+}
+
+std::vector<uint8_t> SineWaveAudioIn::pcm16_samples_to_bytes(const std::vector<int16_t>& samples) {
+    std::vector<uint8_t> bytes(samples.size() * sizeof(int16_t));
+    std::copy(reinterpret_cast<const uint8_t*>(samples.data()),
+              reinterpret_cast<const uint8_t*>(samples.data()) + bytes.size(),
+              bytes.begin());
+    return bytes;
+}
+
+AudioIn::audio_chunk SineWaveAudioIn::create_audio_chunk(const std::vector<int16_t>& samples,
+                                                         const std::string& codec,
+                                                         int sample_rate_hz,
+                                                         int num_channels,
+                                                         int sequence_number) {
+    audio_chunk chunk;
+    chunk.audio_data = pcm16_samples_to_bytes(samples);
+    chunk.info.codec = codec;
+    chunk.info.sample_rate_hz = sample_rate_hz;
+    chunk.info.num_channels = num_channels;
+
+    auto now = std::chrono::system_clock::now();
+    auto nanos = std::chrono::duration_cast<std::chrono::nanoseconds>(now.time_since_epoch());
+    // Set both start and end timestamps to the current system time in nanoseconds.
+    // In a real application, start_timestamp_ns would mark the time the audio chunk begins,
+    // and end_timestamp_ns would mark when it ends. Here they are equal for simplicity.
+    chunk.start_timestamp_ns = nanos;
+    chunk.end_timestamp_ns = nanos;
+    chunk.sequence_number = sequence_number;
+
+    return chunk;
+}
+
+properties SineWaveAudioIn::get_properties(const ProtoStruct&) {
+    properties props;
+    props.supported_codecs = {audio_codecs::PCM_16};
     props.sample_rate_hz = 44100;
     props.num_channels = 1;
     return props;
@@ -92,9 +146,9 @@ void SineWaveAudioIn::get_audio(std::string const& codec,
                                 double const& duration_seconds,
                                 int64_t const& previous_timestamp,
                                 const ProtoStruct& extra) {
-    const int sample_rate = 44100;  // 44.1 kHz
-    const double amplitude = 0.5;   // Half volume
-    const int chunk_size = 1024;    // Samples per chunk
+    const int sample_rate = 44100;
+    const double amplitude = 0.5;
+    const int chunk_size = 1024;
 
     int total_samples = static_cast<int>(duration_seconds * sample_rate);
     int num_chunks = (total_samples + chunk_size - 1) / chunk_size;
@@ -107,37 +161,16 @@ void SineWaveAudioIn::get_audio(std::string const& codec,
         std::vector<int16_t> samples;
         samples.reserve(samples_in_chunk);
 
-        // Generate sine wave samples
+        // Create each sample and put in chunk.
         for (int i = 0; i < samples_in_chunk; ++i) {
             int sample_idx = chunk_idx * chunk_size + i;
-            double t = static_cast<double>(sample_idx) / sample_rate;
-            double sample_value = amplitude * std::sin(2.0 * M_PI * frequency_ * t);
-
-            // Convert to 16-bit PCM
-            int16_t pcm_sample = static_cast<int16_t>(sample_value * 32767.0);
-            samples.push_back(pcm_sample);
+            double time_seconds = static_cast<double>(sample_idx) / sample_rate;
+            double sample_value = generate_sine_sample(frequency_, amplitude, time_seconds);
+            samples.push_back(float_to_pcm16(sample_value));
         }
 
-        audio_chunk chunk;
-
-        // Convert int16_t samples to uint8_t bytes
-        chunk.audio_data.resize(samples.size() * sizeof(int16_t));
-        std::memcpy(chunk.audio_data.data(), samples.data(), chunk.audio_data.size());
-
-        // Set audio info
-        chunk.info.codec = codec;
-        chunk.info.sample_rate_hz = sample_rate;
-        chunk.info.num_channels = 1;
-
-        // Get current timestamp in nanoseconds
-        auto now = std::chrono::system_clock::now();
-        auto nanos =
-            std::chrono::duration_cast<std::chrono::nanoseconds>(now.time_since_epoch()).count();
-        chunk.start_timestamp_ns = nanos;
-        chunk.end_timestamp_ns = nanos;
-        chunk.sequence = chunk_idx;
+        audio_chunk chunk = create_audio_chunk(samples, codec, sample_rate, 1, chunk_idx);
 
-        // Call the chunk handler callback
         if (!chunk_handler(std::move(chunk))) {
             VIAM_RESOURCE_LOG(info) << "Chunk handler returned false, stopping";
             break;

src/viam/sdk/CMakeLists.txt

@@ -60,6 +60,7 @@ endif()
 
 target_sources(viamsdk
   PRIVATE
+    common/audio.cpp
     common/client_helper.cpp
     common/exception.cpp
     common/instance.cpp
@@ -166,6 +167,7 @@ target_sources(viamsdk
       ../..
       ${CMAKE_CURRENT_BINARY_DIR}/../..
     FILES
+      ../../viam/sdk/common/audio.hpp
       ../../viam/sdk/common/client_helper.hpp
       ../../viam/sdk/common/exception.hpp
       ../../viam/sdk/common/instance.hpp
@@ -177,6 +179,7 @@ target_sources(viamsdk
       ../../viam/sdk/common/version_metadata.hpp
       ../../viam/sdk/common/world_state.hpp
       ../../viam/sdk/components/arm.hpp
+      ../../viam/sdk/components/audio_in.hpp
       ../../viam/sdk/components/base.hpp
       ../../viam/sdk/components/board.hpp
       ../../viam/sdk/components/button.hpp