feat: support configuring audio channels and sample rates of recordings; audio recordings now work!

src/main.rs

@@ -2,8 +2,13 @@ use clap::Parser;
 use fomo_reducer::{CommandRouter, State, Storage, all_commands, initialize_vcs, update_vcs};
 use secrecy::{ExposeSecret, SecretString};
 use snafu::Snafu;
-use songbird::{Songbird, shards::TwilightMap};
+use songbird::{
+    Config, Songbird,
+    driver::{Channels, DecodeConfig, SampleRate},
+    shards::TwilightMap,
+};
 use std::{fmt::Debug, sync::Arc};
+use strum::EnumString;
 use tokio::{select, signal::ctrl_c, task::JoinSet};
 use tokio_util::{sync::CancellationToken, time::FutureExt as _};
 use tracing_subscriber::{EnvFilter, fmt::format::FmtSpan};
@@ -17,6 +22,47 @@ use twilight_model::{
     id::{Id, marker::UserMarker},
 };
 
+#[derive(Clone, Copy, Debug, strum::Display, EnumString)]
+enum AudioChannels {
+    Mono,
+    Stereo,
+}
+
+impl From<AudioChannels> for Channels {
+    fn from(value: AudioChannels) -> Self {
+        match value {
+            AudioChannels::Mono => Channels::Mono,
+            AudioChannels::Stereo => Channels::Stereo,
+        }
+    }
+}
+
+#[derive(Clone, Copy, Debug, strum::Display, EnumString)]
+enum AudioSampleRate {
+    #[strum(serialize = "8000Hz")]
+    Hz8000,
+    #[strum(serialize = "12000Hz")]
+    Hz12000,
+    #[strum(serialize = "16000Hz")]
+    Hz16000,
+    #[strum(serialize = "24000Hz")]
+    Hz24000,
+    #[strum(serialize = "48000Hz")]
+    Hz48000,
+}
+
+impl From<AudioSampleRate> for SampleRate {
+    fn from(value: AudioSampleRate) -> Self {
+        match value {
+            AudioSampleRate::Hz8000 => SampleRate::Hz8000,
+            AudioSampleRate::Hz12000 => SampleRate::Hz12000,
+            AudioSampleRate::Hz16000 => SampleRate::Hz16000,
+            AudioSampleRate::Hz24000 => SampleRate::Hz24000,
+            AudioSampleRate::Hz48000 => SampleRate::Hz48000,
+        }
+    }
+}
+
 #[derive(Debug, Parser)]
 struct AppArgs {
     #[arg(long, env)]
@@ -31,6 +77,12 @@ struct AppArgs {
     #[arg(long, env)]
     discord_status: Option<Arc<str>>,
 
+    #[arg(long, env, default_value_t = AudioChannels::Mono)]
+    audio_channels: AudioChannels,
+
+    #[arg(long, env, default_value_t = AudioSampleRate::Hz12000)]
+    audio_sample_rate: AudioSampleRate,
+
     #[arg(long, env)]
     bot_data: Storage,
 
@@ -89,6 +141,8 @@ async fn main() -> Result<(), MainError> {
         discord_bot_owner_user_id,
         discord_nickname,
         discord_status,
+        audio_channels,
+        audio_sample_rate,
         bot_data,
         user_data,
         recording_data,
@@ -159,9 +213,17 @@ async fn main() -> Result<(), MainError> {
             .collect(),
     );
 
-    let senders = Arc::new(senders);
+    let audio_channels = audio_channels.into();
+    let audio_sample_rate = audio_sample_rate.into();
 
+    let senders = Arc::new(senders);
     let songbird = Songbird::twilight(senders, discord_user_id);
+    songbird.set_config(
+        Config::default().decode_mode(songbird::driver::DecodeMode::Decode(DecodeConfig::new(
+            audio_channels,
+            audio_sample_rate,
+        ))),
+    );
 
     let interaction_client = discord_client.interaction(discord_application_id);
 
@@ -196,6 +258,8 @@ async fn main() -> Result<(), MainError> {
     let user_data = user_data.into_inner();
 
     let state = State {
+        audio_channels,
+        audio_sample_rate,
         bot_data,
         cancellation_token: cancellation_token.clone(),
         discord_application_id,