Client VCData (SSRC -> UserID)

Can use this to store per guild and per channel information

May also store a list of visible users/those that have talked too
while Alan was in the chat

flake.nix

@@ -40,6 +40,8 @@
           buildInputs = with pkgs; [
             # Add additional build inputs here
             libopus
+            stt
+            tts
           ] ++ lib.optionals pkgs.stdenv.isDarwin ([
             # Additional darwin specific inputs can be set here
             pkgs.libiconv

src/commands.rs

@@ -1,7 +1,11 @@
 use serenity::framework::standard::macros::group;
+
 pub mod ping;
 use ping::*;
 
+pub mod vc;
+use vc::*;
+
 #[group]
-#[commands(ping)]
+#[commands(ping, join)]
 pub struct General;

src/commands/vc.rs

@@ -0,0 +1,30 @@
+use serenity::framework::standard::macros::command;
+use serenity::framework::standard::CommandResult;
+use serenity::model::channel::Message;
+use serenity::prelude::Context;
+
+use crate::vc;
+
+#[command]
+#[only_in(guilds)]
+pub async fn join(ctx: &Context, msg: &Message) -> CommandResult {
+    let guild = msg.guild(&ctx.cache).unwrap();
+
+    let channel_id = guild
+        .voice_states
+        .get(&msg.author.id)
+        .and_then(|voice_state| voice_state.channel_id);
+
+    let connect_to = match channel_id {
+        Some(channel) => channel,
+        None => {
+            msg.reply(ctx, "Not in a voice channel").await;
+
+            return Ok(());
+        },
+    };
+
+    vc::join(ctx.clone(), guild.clone(), connect_to).await;
+    
+    Ok(())
+}

src/main.rs

@@ -2,13 +2,13 @@
 
 use std::env;
 
-use serenity::framework::standard::StandardFramework;
+use serenity::framework::standard::{StandardFramework, WithWhiteSpace};
 use serenity::prelude::*;
 
 // This trait adds the `register_songbird` and `register_songbird_with` methods
 // to the client builder below, making it easy to install this voice client.
 // The voice client can be retrieved in any command using `songbird::get(ctx).await`.
-use songbird::SerenityInit;
+use songbird::{SerenityInit, Config, driver::DecodeMode};
 
 mod commands;
 use commands::*;
@@ -17,6 +17,7 @@ mod events;
 use events::*;
 
 mod utils;
+mod vc;
 
 #[tokio::main]
 async fn main() {
@@ -25,7 +26,8 @@ async fn main() {
     // Configure the client with your Discord bot token in the environment.
     let token = env::var("DISCORD_TOKEN").expect("Expected a token in the environment");
     // Set gateway intents, which decides what events the bot will be notified about
-    let intents = GatewayIntents::GUILDS
+    let intents = GatewayIntents::non_privileged()
+	| GatewayIntents::privileged()
         | GatewayIntents::DIRECT_MESSAGES
         | GatewayIntents::DIRECT_MESSAGE_REACTIONS
         | GatewayIntents::DIRECT_MESSAGE_TYPING
@@ -33,19 +35,29 @@ async fn main() {
 	| GatewayIntents::GUILD_VOICE_STATES;
 
     let framework = StandardFramework::new()
-        .configure(|c| c.prefix("ALAN! "))
+        .configure(|c| {
+	    c.prefix("ALAN! ");
+	    c.with_whitespace(WithWhiteSpace {
+		commands: true,
+		groups: true,
+		prefixes: true
+	    })
+	}
+	)
         .group(&GENERAL_GROUP);
 
+    let songbird_config = Config::default()
+        .decode_mode(DecodeMode::Decode);
+    
     let mut client = Client::builder(&token, intents)
         .event_handler(Handler)
         .framework(framework)
-        .register_songbird()
+        .register_songbird_from_config(songbird_config)
         .await
         .expect("Err creating client");
 
-    // TODO: do checks when getting data in add guild etc.
-    println!("Initialize guild voice popin state");
     utils::guild_popin::init(&client).await;
+    vc::init(&client).await;
 
     if let Err(why) = client.start().await {
         println!("Client error: {:?}", why);

src/utils/guild_popin.rs

@@ -1,7 +1,7 @@
 use rand::prelude::*;
 use serenity::model::id::GuildId;
 use serenity::Client;
-use songbird::{create_player, ffmpeg, Call};
+use songbird::Call;
 use std::collections::HashMap;
 use std::ops::Range;
 use std::sync::Arc;
@@ -9,7 +9,9 @@ use std::thread::sleep;
 use std::time::Duration;
 
 use serenity::model::prelude::{ChannelType, Guild};
-use serenity::prelude::{Context, RwLock, TypeMapKey};
+use serenity::prelude::{Context, RwLock, TypeMapKey, Mutex};
+
+use crate::vc;
 
 #[derive(Debug)]
 struct GuildPopIn {
@@ -23,8 +25,8 @@ impl TypeMapKey for GuildPopIn {
 
 fn init_pop_state() -> GuildPopIn {
     GuildPopIn {
-        min: 1000,
-        max: 5000,
+        min: 30*60*1000,
+        max: 180*60*1000,
     }
 }
 
@@ -50,30 +52,16 @@ async fn popin(ctx: Context, guild: Guild) {
             None => {}
         }
     }
+    let mut call = None;
     match most {
         Some(chan) => {
-            let manager = songbird::get(&ctx).await.expect("Songbird: intialization");
-            let (call, status) = manager.join(guild.id, chan).await;
-            match status {
-                Ok(_) => {
-                    let mut call = call.lock().await;
-                    let ding_src =
-                        std::env::var("DING_SOUND").expect("DING not found in DING_SOUND");
-                    let ding = ffmpeg(ding_src).await.expect("no ding.");
-                    let (audio, handle) = create_player(ding);
-                    call.play(audio);
-                    tokio::spawn(popout_soon(ctx.clone(), guild, Some(call.clone())));
-                }
-                Err(_err) => {
-                    println!("Error joining channel");
-                }
-            }
+	    call = vc::join(ctx.clone(), guild.clone(), chan).await;
         }
         None => {
             println!("No good channel to join");
         }
     }
-    // tokio::spawn(popout_soon(ctx.clone(), guild, None));
+    tokio::spawn(popout_soon(ctx.clone(), guild, call));
 }
 
 async fn popin_soon(ctx: Context, guild: Guild) {
@@ -95,17 +83,24 @@ async fn popin_soon(ctx: Context, guild: Guild) {
     tokio::spawn(popin(ctx.clone(), guild));
 }
 
-async fn popout(ctx: Context, guild: Guild, call: Option<Call>) {
+async fn popout(ctx: Context, guild: Guild, call: Option<Arc<Mutex<Call>>>) {
     match call {
-	Some(mut call) => {
-	    call.leave().await;
+	Some(call) => {
+	    let mut call = call.lock().await;
+	    let status = call.leave().await;
+	    match status {
+		Ok(_) => {},
+		Err(_err) => {
+		    println!("Failed to leave call");
+		}
+	    }
 	},
 	None => {}
     }
     start_dingdong(ctx.clone(), guild);
 }
 
-async fn popout_soon(ctx: Context, guild: Guild, call: Option<Call>) {
+async fn popout_soon(ctx: Context, guild: Guild, call: Option<Arc<Mutex<Call>>>) {
     let data = ctx.data.read().await;
     let pops = data
         .get::<GuildPopIn>()

src/vc/mod.rs

@@ -0,0 +1,249 @@
+use std::{collections::HashMap, sync::Arc};
+
+use serenity::{
+    async_trait,
+    model::prelude::{ChannelId, Guild, GuildId},
+    prelude::{Context, Mutex, RwLock, TypeMapKey}, Client,
+};
+
+use songbird::{
+    create_player,
+    events::context_data::{ConnectData, DisconnectData},
+    ffmpeg,
+    model::{
+        id::UserId,
+        payload::{ClientDisconnect, Speaking},
+    },
+    Call, CoreEvent, Event, EventContext, EventHandler,
+};
+
+#[derive(Eq, Hash)]
+struct CallLocation {
+    guild: GuildId,
+    channel: ChannelId
+}
+
+impl PartialEq for CallLocation {
+    fn eq(&self, other: &Self) -> bool {
+	self.guild == other.guild && self.channel == other.channel
+    }
+}
+
+struct VoiceData {
+    call: Arc<Mutex<Call>>,
+    users: Arc<RwLock<HashMap<u32, Option<UserId>>>>,
+}
+
+impl VoiceData {
+    pub fn new(call: Arc<Mutex<Call>>) -> Self {
+	Self {
+	    call,
+	    users: Arc::new(RwLock::new(HashMap::default()))
+	}
+    }
+}
+
+struct VCData {
+    loc: Arc<CallLocation>,
+    data: Arc<RwLock<VoiceData>>
+}
+
+impl TypeMapKey for VCData {
+    type Value = Arc<RwLock<HashMap<CallLocation, VCData>>>;
+}
+
+impl VCData {
+    pub fn new(loc: CallLocation, data: VoiceData) -> Self {
+        // You can manage state here, such as a buffer of audio packet bytes so
+        // you can later store them in intervals.
+        VCData {
+	    loc: Arc::new(loc),
+	    data: Arc::new(RwLock::new(data))
+	}
+    }
+    pub fn clone(&self) -> Self {
+	VCData {
+	    loc: self.loc.clone(),
+	    data: self.data.clone()
+	}
+    }
+}
+
+pub async fn init(client: &Client) {
+    let mut data = client.data.write().await;
+    data.insert::<VCData>(Arc::new(RwLock::new(HashMap::default())))
+}
+
+#[async_trait]
+impl EventHandler for VCData {
+    #[allow(unused_variables)]
+    async fn act(&self, ctx: &EventContext<'_>) -> Option<Event> {
+        use EventContext as Ctx;
+        match ctx {
+            Ctx::SpeakingStateUpdate(Speaking {
+                speaking,
+                ssrc,
+                user_id,
+                ..
+            }) => {
+                // Discord voice calls use RTP, where every sender uses a randomly allocated
+                // *Synchronisation Source* (SSRC) to allow receivers to tell which audio
+                // stream a received packet belongs to. As this number is not derived from
+                // the sender's user_id, only Discord Voice Gateway messages like this one
+                // inform us about which random SSRC a user has been allocated. Future voice
+                // packets will contain *only* the SSRC.
+                //
+                // You can implement logic here so that you can differentiate users'
+                // SSRCs and map the SSRC to the User ID and maintain this state.
+                // Using this map, you can map the `ssrc` in `voice_packet`
+                // to the user ID and handle their audio packets separately.
+                {
+		    let data = self.data.write().await;
+                    let mut users = data.users.write().await;
+                    users.insert(ssrc.clone(), user_id.clone());
+                }
+                println!(
+                    "\n\n\nSpeaking state update: user {:?} has SSRC {:?}, using {:?}\n\n\n",
+                    user_id, ssrc, speaking,
+                );
+            }
+            Ctx::SpeakingUpdate(data) => {
+                // You can implement logic here which reacts to a user starting
+                // or stopping speaking, and to map their SSRC to User ID.
+		let vcdata = self.data.read().await;
+                let users = vcdata.users.read().await;
+                println!(
+                    "Source {}/{:?} has {} speaking.",
+                    data.ssrc,
+                    users.get(&data.ssrc),
+                    if data.speaking { "started" } else { "stopped" },
+                );
+            }
+            Ctx::VoicePacket(data) => {
+                // An event which fires for every received audio packet,
+                // containing the decoded data.
+                if let Some(audio) = data.audio {
+                    // println!("Audio packet's first 5 samples: {:?}", audio.get(..5.min(audio.len())));
+                    // println!(
+                    // 	"Audio packet sequence {:05} has {:04} bytes (decompressed from {}), SSRC {}",
+                    // 	data.packet.sequence.0,
+                    // 	audio.len() * std::mem::size_of::<i16>(),
+                    // 	data.packet.payload.len(),
+                    // 	data.packet.ssrc,
+                    // );
+                } else {
+                    // println!("RTP packet, but no audio. Driver may not be configured to decode.");
+                }
+            }
+            Ctx::RtcpPacket(data) => {
+                // An event which fires for every received rtcp packet,
+                // containing the call statistics and reporting information.
+                // println!("RTCP packet received: {:?}", data.packet);
+            }
+            Ctx::ClientDisconnect(ClientDisconnect { user_id, .. }) => {
+                // You can implement your own logic here to handle a user who has left the
+                // voice channel e.g., finalise processing of statistics etc.
+                // You will typically need to map the User ID to their SSRC; observed when
+                // first speaking.
+
+                println!("Client disconnected: user {:?}", user_id);
+            }
+            Ctx::DriverConnect(ConnectData { channel_id, .. }) => {
+                println!("VoiceDriver is connected.");
+            }
+	    Ctx::DriverDisconnect(DisconnectData {
+		channel_id,
+		guild_id,
+		..
+	    }) => {
+		// TODO: Remove data from GuildVoiceData
+	    }
+            _ => {
+                // We won't be registering this struct for any more event classes.
+                unimplemented!()
+            }
+        }
+
+        None
+    }
+}
+
+pub async fn play_file(call: Arc<Mutex<Call>>, file: String) {
+    let mut call = call.lock().await;
+    let ff_src = ffmpeg(file).await.expect("Unable to find file.");
+    let (audio, handle) = create_player(ff_src);
+    call.play(audio);
+}
+
+pub async fn join(ctx: Context, guild: Guild, cid: ChannelId) -> Option<Arc<Mutex<Call>>> {
+    let manager = songbird::get(&ctx).await.expect("Songbird: intialization");
+    let (call, status) = manager.join(guild.id, cid).await;
+    match status {
+        Ok(_) => {
+	    let vc_data: VCData = VCData::new(
+		CallLocation {
+		    guild: guild.id,
+		    channel: cid
+		}, VoiceData::new(
+		    call.clone()
+		)
+	    );
+            {
+                let data = ctx.data.read().await;
+		match data.get::<VCData>() {
+                    Some(vc_guild) => {
+			let mut vc_guild = vc_guild.write().await;
+			vc_guild.insert(CallLocation {
+			    guild: guild.id,
+			    channel: cid
+			}, vc_data.clone());
+		    }
+		    None => {
+			println!("VoiceData for client hasn't been initialized");
+		    }
+		}
+            }
+            let call_handle = call.clone();
+            {
+                let mut call = call.lock().await;
+
+                call.add_global_event(
+                    CoreEvent::SpeakingStateUpdate.into(),
+                    vc_data.clone()
+                );
+
+                call.add_global_event(
+                    CoreEvent::SpeakingUpdate.into(),
+                    vc_data.clone()
+                );
+
+                call.add_global_event(
+                    CoreEvent::VoicePacket.into(),
+                    vc_data.clone()
+                );
+
+                call.add_global_event(
+                    CoreEvent::RtcpPacket.into(),
+                    vc_data.clone()
+                );
+
+                call.add_global_event(
+                    CoreEvent::ClientDisconnect.into(),
+                    vc_data.clone()
+                );
+
+                call.add_global_event(
+                    CoreEvent::DriverConnect.into(),
+                    vc_data.clone()
+                );
+            }
+            let ding_src = std::env::var("DING_SOUND").expect("DING not found in DING_SOUND");
+            play_file(call_handle, ding_src).await;
+            return Some(call);
+        }
+        Err(_err) => {
+            println!("Error joining channel");
+        }
+    }
+    None
+}