discord で読み上げ bot をつくる

tkt

get started はここから

tkt

こっちにする

tkt

https://github.com/serenity-rs/serenity
これ使う

tkt

example 作れたけど、これはどうやったら動かせるんだ？

tkt

local serve して、/ping 投げても帰ってこない

tkt

動かしてるのはこれ

tkt

さっき動かなかったのは、コマンドが違った説

tkt

.configure(|c| c.prefix("~")) // set the bot's prefix to "~"

これで設定してた。

ので、コマンドは ~ping

tkt

voice 再生系の機能もあるっぽい

tkt

start で読み取り開始して voice に入室。
それ以降はメッセージの追加を検知して読み取る必要がある。
できるかな、

tkt

webhook とか設定してないけど、どうやって動いてるのか…

tkt

このライブラリだと、ボイスチャット系に対応してないかも？

tkt

そのようなライブラリの1つがSerenityです。これは、オプションを利用できるようにしながら、心配する必要のないものを抽象化するライブラリです。これはDiscordAPIをラップアラウンドし、音声サポートとすべての作業を備えた、ほとんど機能が完全です。

気づいてないだけかも

tkt

https://github.com/serenity-rs/songbird
こっちかな

tkt

m1 でビルドできない

error: failed to run custom build command for `audiopus_sys v0.1.8`

Caused by:
  process didn't exit successfully: `/app/target/debug/build/audiopus_sys-688f2c42484975a9/build-script-build` (exit code: 101)

tkt

ここで同じことで悩んでる人がいた

tkt

error: 
        is_x86_feature_detected can only be used on x86 and x86_64 targets.
        You can prevent it from being used in other architectures by
        guarding it behind a cfg(target_arch) as follows:

            #[cfg(any(target_arch = "x86", target_arch = "x86_64"))] {
                if is_x86_feature_detected(...) { ... }
            }
        
  --> /usr/local/cargo/registry/src/github.com-1ecc6299db9ec823/opusic-sys-0.3.4/build.rs:74:13
   |
74 |         if !std::is_x86_feature_detected!("sse") {
   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   |
   = note: this error originates in a macro (in Nightly builds, run with -Z macro-backtrace for more info)

error: aborting due to 8 previous errors

For more information about this error, try `rustc --explain E0658`.
error: could not compile `opusic-sys`

To learn more, run the command again with --verbose.
warning: build failed, waiting for other jobs to finish...
error: build failed

それでもだめだった

tkt

m1 にはまだ早いのか、、、

tkt

https://github.com/DoumanAsh/suzumi
これを clone してきて build しても同じエラーだったので、厳しそう

tkt

intel mac に切り替えて作業することにする

tkt

intel

error: failed to run custom build command for `audiopus_sys v0.1.8`

Caused by:
  process didn't exit successfully: `/app/target/debug/build/audiopus_sys-e2d1122e6f3b14ab/build-script-build` (exit code: 101)
  --- stdout
  cargo:info=No feature or environment variable found, linking by default.
  cargo:rerun-if-env-changed=OPUS_NO_PKG_CONFIG
  cargo:rerun-if-env-changed=PKG_CONFIG
  cargo:rerun-if-env-changed=PKG_CONFIG_PATH_x86_64-unknown-linux-musl
  cargo:rerun-if-env-changed=PKG_CONFIG_PATH_x86_64_unknown_linux_musl
  cargo:rerun-if-env-changed=HOST_PKG_CONFIG_PATH
  cargo:rerun-if-env-changed=PKG_CONFIG_PATH
  cargo:rerun-if-env-changed=PKG_CONFIG_LIBDIR_x86_64-unknown-linux-musl
  cargo:rerun-if-env-changed=PKG_CONFIG_LIBDIR_x86_64_unknown_linux_musl
  cargo:rerun-if-env-changed=HOST_PKG_CONFIG_LIBDIR
  cargo:rerun-if-env-changed=PKG_CONFIG_LIBDIR
  cargo:rerun-if-env-changed=PKG_CONFIG_SYSROOT_DIR_x86_64-unknown-linux-musl
  cargo:rerun-if-env-changed=PKG_CONFIG_SYSROOT_DIR_x86_64_unknown_linux_musl
  cargo:rerun-if-env-changed=HOST_PKG_CONFIG_SYSROOT_DIR
  cargo:rerun-if-env-changed=PKG_CONFIG_SYSROOT_DIR
  cargo:info=`pkg_config` could not find `Opus`.
  cargo:info=Opus source path: "/usr/local/cargo/registry/src/github.com-1ecc6299db9ec823/audiopus_sys-0.1.8/opus".
  cargo:info=Opus will be built as static-library.
  Updating build configuration files, please wait....

  --- stderr
  autogen.sh: line 14: autoreconf: not found
  thread 'main' panicked at 'Failed to autogen Opus.', /usr/local/cargo/registry/src/github.com-1ecc6299db9ec823/audiopus_sys-0.1.8/build.rs:102:9
  note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
warning: build failed, waiting for other jobs to finish...
error: build failed

依存してる何かが入ってない？

tkt

これを順番に入れていく。
alpine でできるかは不明

tkt

/app # apk add libopus-dev
fetch http://dl-cdn.alpinelinux.org/alpine/v3.12/main/x86_64/APKINDEX.tar.gz
fetch http://dl-cdn.alpinelinux.org/alpine/v3.12/community/x86_64/APKINDEX.tar.gz
ERROR: unsatisfiable constraints:
  libopus-dev (missing):
    required by: world[libopus-dev]

どういうこっちゃ

tkt

apk add opus

したらビルドが通った

tkt

/app # cargo run
   Compiling app v0.1.0 (/app)
    Finished dev [unoptimized + debuginfo] target(s) in 15.70s
     Running `target/debug/app`
speach is connected!
Segmentation fault

~join したら落ちた。まだいろいろいるのかも

tkt

RUN apk update && \
    apk upgrade && \
    apk add --update --no-cache alpine-sdk git opus-dev build-base autoconf automake libtool m4 ffmpeg

これだけ入れてもだめっぽい

tkt

alpine だとダメかもなので ubuntu のイメージ引っ張ってくる

tkt

FROM rust:alpine3.12

WORKDIR /app
RUN apk update && \
    apk upgrade && \
    apk add --update --no-cache alpine-sdk git opus-dev build-base autoconf automake libtool m4 ffmpeg

ENV FFMPEG_CORE="bash \
    tzdata \
    libass \
    libstdc++ \
    libpng \
    libjpeg \
    xvidcore \
    x264-libs \
    x265 \
    libvpx \
    libvorbis \
    opus \
    lame \
    fdk-aac \
    freetype \
    libavc1394 \
    yasm"

ENV RUNTIME_DEPS="coreutils \
    make \
    cmake \
    gcc \
    g++ \
    clang \
    linux-headers \
    git \
    curl"

ENV BUILD_DEPS="fdk-aac-dev \
    freetype-dev \
    x264-dev \
    x265-dev \
    yasm-dev \
    libogg-dev \
    libvorbis-dev \
    opus-dev \
    libvpx-dev \
    lame-dev \
    xvidcore-dev \
    libass-dev \
    openssl-dev \
    musl-dev \
    build-base \
    libjpeg-turbo-dev \
    libpng-dev \
    libavc1394-dev \
    libavc1394-dev \
    clang-dev"
ENV FFMPEG_VERSION=4.0

# FFMPEG
RUN apk update && \
    apk add --no-cache apk-tools && \
    echo http://nl.alpinelinux.org/alpine/edge/testing >> /etc/apk/repositories && \
    apk update && \
    apk add --no-cache --virtual .runtime-deps ${RUNTIME_DEPS} && \
    apk add --no-cache ${FFMPEG_CORE} && \
    apk add --no-cache --virtual .build-deps ${BUILD_DEPS}

COPY Cargo.lock Cargo.lock
COPY Cargo.toml Cargo.toml

COPY  . .

CMD [ "/bin/sh",  "-c", "cargo run" ]

これでもだめだった

tkt

あきらめて ubuntu

tkt

うごいた

tkt

Voiceのとこが動かんかったので、次は再生のexampleそのままとってきてちゃれんじ

tkt

参考が増えた

tkt

できたらここにのせたい

tkt

CoefontStudioつかいたい

tkt

youtube の音流すのはできたので

message 読み取り
音声変換
lambda に乗せてみる

tkt

lambda には果たして乗るのだろうか

tkt

厳しそうな予感はする

tkt

これ使えばできそうだけど、読み上げでいちいち / 入れてられないので厳しい

tkt

rust が動く安いコンテナ探すか…

tkt

ローカルで動かして、youtube の url いれたら音が流れるようにはなった

tkt

ローカルで動かして、youtube の url いれたら音が流れるようにはなった

tkt

今回は音声ファイルを取得してきてそれを流したいけど、やり方がよくわからない

tkt

let ffmpeg = Command::new("ffmpeg")
    .args(pre_args)
    .arg("-i")
    .arg("-")
    .args(&ffmpeg_args)
    .stdin(youtube_dl.stdout.ok_or(Error::Stdout)?)
    .stderr(Stdio::null())
    .stdout(Stdio::piped())
    .spawn()?;

let metadata = Metadata::from_ytdl_output(value?);

trace!("ytdl metadata {:?}", metadata);

Ok(Input::new(
    true,
    child_to_reader::<f32>(ffmpeg),
    Codec::FloatPcm,
    Container::Raw,
    Some(metadata),
))

input を作れば流せるみたいだけど、それを作ってるのが独自手法過ぎて(?)よくわからん

tkt

ローカルファイルからinputを作る方法で簡単なのがないか探す

tkt

input はこれ

tkt

// Input
pub fn new(
    stereo: bool,
    reader: Reader,
    kind: Codec,
    container: Container,
    metadata: Option<Metadata>
) -> Self

pub enum Reader {
    Pipe(BufReader<ChildContainer>),
    Memory(Catcher<Box<Reader>>),
    Compressed(TxCatcher<Box<Input>, OpusCompressor>),
    Restartable(Restartable),
    File(BufReader<File>),
    Vec(Cursor<Vec<u8>>),
    Extension(Box<dyn Read + Send>),
    ExtensionSeek(Box<dyn ReadSeek + Send>),
}

pub enum Codec {
    Opus(OpusDecoderState),
    Pcm,
    FloatPcm,
}

pub enum Codec {
    Opus(OpusDecoderState),
    Pcm,
    FloatPcm,
}

pub struct Metadata {
    pub title: Option<String>,
    pub artist: Option<String>,
    pub date: Option<String>,
    pub channels: Option<u8>,
    pub start_time: Option<Duration>,
    pub duration: Option<Duration>,
    pub sample_rate: Option<u32>,
    pub source_url: Option<String>,
    pub thumbnail: Option<String>,
}

enum 多めだしなんとかなりそう…？

tkt

厳しかったら ffmpeg 経由でやる

tkt

metadata 見てると曲であること前提の構造に見えて、喋らせたいとかの場合これではないので貼って思えてきた

tkt

use songbird::{driver::Driver, ffmpeg, tracks::create_player};

// A Call is also valid here!
let mut handler: Driver = Default::default();
let source = ffmpeg("../audio/my-favourite-song.mp3")
    .await
    .expect("This might fail: handle this error!");
let (mut audio, audio_handle) = create_player(source);

audio.set_volume(0.5);

handler.play_only(audio);

// Future access occurs via audio_handle.

それっぽいのがあった

tkt

とりあえずこれでローカルファイルが流せるかやってみる

tkt

ffmpeg とか youtube-dl とか使ってるのがメインだから、rust じゃなくていいのでは？
とか思っちゃったけど、今は考えてはいけない

tkt

これの play_only を呼び出したい
Methods from Deref<Target = Driver> これはどうやって呼び出すのか
Rust 力が足りない

tkt

ソースで定義見ると、ちゃんと impl されてる

tkt

/// The Call handler is responsible for a single voice connection, acting
/// as a clean API above the inner state and gateway message management.
///
/// If the `"driver"` feature is enabled, then a Call exposes all control methods of
/// [`Driver`] via `Deref(Mut)`.
///
/// [`Driver`]: struct@Driver

driver を enable しないとだめ？

tkt

ソース追ってたらいつの間にか使えるようになってた
ロードできてなかっただけ？

tkt

handler.play(track);
だと再生できなかった
次は
handler.enque(track);
で試す

tkt

input どうやって作るんだって言ってたけど

/// Opens an audio file through `ffmpeg` and creates an audio source.
///
/// This source is not seek-compatible.
/// If you need looping or track seeking, then consider using
/// [`Restartable::ffmpeg`].
///
/// [`Restartable::ffmpeg`]: crate::input::restartable::Restartable::ffmpeg
pub async fn ffmpeg<P: AsRef<OsStr>>(path: P) -> Result<Input> {
    _ffmpeg(path.as_ref()).await
}

これでいけそう

tkt

handler.play(track);
だと再生できなかった
次は
handler.enque(track);
で試す

ファイルパスの指定が間違っていた可能性が大

tkt

async fn get_input_from_local() -> Input {
    let root = option_env!("CARGO_MANIFEST_DIR").unwrap();
    let path = Path::new(root);
    let file_path = path.join("binaries").join("2_23_AM_2.mp3");
    return ffmpeg(file_path)
        .await
        .expect("This might fail: handle this error!");
}

project root の取得と Input への変換
これを enqueue_source で再生に成功

tkt

これで GCP の環境を作った

tkt

これでできそうな気もするけど、なんとも言えない

tkt

use google_texttospeech1::Texttospeech;
use google_texttospeech1::{api::ListVoicesResponse, client, Error, Result};

use std::default::Default;
use yup_oauth2;

pub async fn get_speech() -> client::Result<(hyper::Response<hyper::body::Body>, ListVoicesResponse)>
{
    let secret: yup_oauth2::ApplicationSecret = Default::default();
    let auth = yup_oauth2::InstalledFlowAuthenticator::builder(
        secret,
        yup_oauth2::InstalledFlowReturnMethod::HTTPRedirect,
    )
    .build()
    .await
    .unwrap();
    let mut hub = Texttospeech::new(
        hyper::Client::builder().build(hyper_rustls::HttpsConnector::with_native_roots()),
        auth,
    );
    hub.voices().list().language_code("magna").doit().await
}

#[cfg(test)]
mod tests {
    use super::*;

    #[tokio::test]
    async fn get() {
        let result = get_speech().await;
        match result {
            Err(e) => match e {
                // The Error enum provides details about what exactly happened.
                // You can also just use its `Debug`, `Display` or `Error` traits
                Error::HttpError(_)
                | Error::Io(_)
                | Error::MissingAPIKey
                | Error::MissingToken(_)
                | Error::Cancelled
                | Error::UploadSizeLimitExceeded(_, _)
                | Error::Failure(_)
                | Error::BadRequest(_)
                | Error::FieldClash(_)
                | Error::JsonDecodeError(_, _) => println!("{}", e),
            },
            Ok(res) => println!("Success: {:?}", res),
        }
    }
}

$ cargo test
   Compiling app v0.1.0 (/workspace)
.
.
.
test tts::tests::get ... test tts::tests::get has been running for over 60 seconds

サンプル通そうとしてるけど、応答さえない

tkt

lib 使わずにやった方が早そう

tkt

Http Client は使えたのでこれを使ってやっていく

tkt

aws polly の選択肢もありそう

これはライブラリがあった

tkt

aws polly の選択肢もありそう

これはライブラリがあった

tkt

こっちもちょっと試してみよう

tkt

こっちもちょっと試してみよう

tkt

これもある

tkt

stars は少ないけど、 init が 2months ago でこれから伸びてきそう
こっちつかってみる

tkt

example

tkt

使うAPIとしてはこれで良いはず

tkt

running 1 test
I can speak as: Lotte in "Dutch"
I can speak as: Maxim in "Russian"
I can speak as: Salli in "US English"
I can speak as: Geraint in "Welsh English"
I can speak as: Miguel in "US Spanish"
I can speak as: Marlene in "German"
I can speak as: Giorgio in "Italian"
I can speak as: Inês in "Portuguese"
I can speak as: Zeina in "Arabic"
I can speak as: Zhiyu in "Chinese Mandarin"
I can speak as: Gwyneth in "Welsh"
I can speak as: Karl in "Icelandic"
I can speak as: Joanna in "US English"
I can speak as: Lucia in "Castilian Spanish"
I can speak as: Cristiano in "Portuguese"
I can speak as: Astrid in "Swedish"
I can speak as: Vicki in "German"
I can speak as: Mia in "Mexican Spanish"
I can speak as: Bianca in "Italian"
I can speak as: Vitória in "Brazilian Portuguese"
I can speak as: Raveena in "Indian English"
I can speak as: Chantal in "Canadian French"
I can speak as: Amy in "British English"
I can speak as: Brian in "British English"
I can speak as: Kevin in "US English"
I can speak as: Russell in "Australian English"
I can speak as: Aditi in "Indian English"
I can speak as: Matthew in "US English"
I can speak as: Dóra in "Icelandic"
I can speak as: Enrique in "Castilian Spanish"
I can speak as: Hans in "German"
I can speak as: Carmen in "Romanian"
I can speak as: Ivy in "US English"
I can speak as: Ewa in "Polish"
I can speak as: Maja in "Polish"
I can speak as: Nicole in "Australian English"
I can speak as: Gabrielle in "Canadian French"
I can speak as: Camila in "Brazilian Portuguese"
I can speak as: Filiz in "Turkish"
I can speak as: Jacek in "Polish"
I can speak as: Justin in "US English"
I can speak as: Céline in "French"
I can speak as: Kendra in "US English"
I can speak as: Ricardo in "Brazilian Portuguese"
I can speak as: Mads in "Danish"
I can speak as: Mathieu in "French"
I can speak as: Léa in "French"
I can speak as: Naja in "Danish"
I can speak as: Penélope in "US Spanish"
I can speak as: Tatyana in "Russian"
I can speak as: Olivia in "Australian English"
I can speak as: Ruben in "Dutch"
I can speak as: Mizuki in "Japanese"
I can speak as: Takumi in "Japanese"
I can speak as: Conchita in "Castilian Spanish"
I can speak as: Carla in "Italian"
I can speak as: Kimberly in "US English"
I can speak as: Jan in "Polish"
I can speak as: Liv in "Norwegian"
I can speak as: Joey in "US English"
I can speak as: Lupe in "US Spanish"
I can speak as: Seoyeon in "Korean"
I can speak as: Emma in "British English"
Voices supporting a neural engine: [Salli, Joanna, Vicki, Amy, Brian, Kevin, Matthew, Ivy, Unknown("Gabrielle"), Camila, Justin, Kendra, Olivia, Kimberly, Joey, Lupe, Seoyeon, Emma]
test tts::tests::test_describe ... ok

まずお試し。いい感じ。

tkt

Rust 1.49 だかでやってるとエラーが出て、1.51 くらいに上げると改善した

参考

tkt

実行はこのあたり

tkt

環境変数で鍵を渡してる

tkt

実行したのはこれ

tkt

つぎこれ。再生ファイル生成できそう

tkt

#[derive(Debug, StructOpt)]
struct Opt {
    /// The region
    #[structopt(short, long)]
    region: Option<String>,

    /// The file containing the text to synthesize
    #[structopt(short, long)]
    filename: String,

    /// Whether to show additional output
    #[structopt(short, long)]
    verbose: bool,
}

このあたりとか、CLI用に調整されてたり、ファイルから読むようになってたりするので調整する

tkt

CLI で実行して先に確認したほうが良いのではと思ったけど、動きそうなのでこのまま進める

tkt

mp3 のダウンロードに成功

tkt

SubscriberBuilder::default()
    .with_env_filter("info")
    .with_span_events(FmtSpan::CLOSE)
    .init();

これを複数回呼び出すと

thread 'tokio-runtime-worker' panicked at 'Unable to install global subscriber: SetGlobalDefaultError { _no_construct: () }', /usr/local/cargo/registry/src/github.com-1ecc6299db9ec823/tracing-subscriber-0.2.18/src/fmt/mod.rs:571:14
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

怒られるので、singletonにするとかの対策をしないといけなそう
ただ実行するだけなら使わなくても大丈夫

tkt

ローカルで動いた

tkt

ローカルで動いた

tkt

herokuに上げる

$ heroku login
$ heroku container:login
$ heroku create [<APPLICATION NAME>]
$ heroku config:set 環境変数
$ heroku stack:set container
$ heroku ps:scale worker=1

このあたりで動くようになった？

tkt

まだ動くところまでは見えてないので様子見

2021-06-26T11:07:04.190523+00:00 app[app.1]: Compiling universal-hash v0.4.0
2021-06-26T11:07:04.272565+00:00 app[app.1]: Compiling aead v0.3.2
2021-06-26T11:07:05.289393+00:00 heroku[app.1]: Process running mem=614M(116.2%)
2021-06-26T11:07:05.291635+00:00 heroku[app.1]: Error R14 (Memory quota exceeded)
2021-06-26T11:07:06.876697+00:00 app[app.1]: Compiling syntex_syntax v0.42.0
2021-06-26T11:07:08.434419+00:00 app[app.1]: Compiling url v2.2.2
2021-06-26T11:07:08.438687+00:00 app[app.1]: Compiling rand_chacha v0.3.1
2021-06-26T11:07:20.802917+00:00 app[app.1]: Compiling rand_chacha v0.2.2
2021-06-26T11:07:24.604586+00:00 app[app.1]: Compiling audiopus v0.2.0
2021-06-26T11:07:26.418395+00:00 app[app.1]: Compiling webpki v0.21.4
2021-06-26T11:07:27.570347+00:00 app[app.1]: Compiling sct v0.6.1
2021-06-26T11:07:28.299319+00:00 app[app.1]: Compiling aws-types v0.1.0 (https://github.com/awslabs/aws-sdk-rust?tag=v0.0.7-alpha#1b923f86)
2021-06-26T11:07:29.876635+00:00 app[app.1]: Compiling enum_primitive v0.1.1
2021-06-26T11:07:30.837048+00:00 app[app.1]: Compiling sha-1 v0.9.6
2021-06-26T11:07:31.236532+00:00 app[app.1]: Compiling salsa20 v0.7.2
2021-06-26T11:07:31.411752+00:00 app[app.1]: Compiling poly1305 v0.6.2
2021-06-26T11:07:31.988065+00:00 heroku[app.1]: Process running mem=807M(151.9%)
2021-06-26T11:07:31.990173+00:00 heroku[app.1]: Error R14 (Memory quota exceeded)
2021-06-26T11:07:33.580496+00:00 app[app.1]: Compiling rand v0.8.4
2021-06-26T11:07:35.593107+00:00 app[app.1]: Compiling rand v0.7.3
2021-06-26T11:07:37.151300+00:00 app[app.1]: Compiling webpki-roots v0.21.1
2021-06-26T11:07:38.164274+00:00 app[app.1]: Compiling webpki-roots v0.20.0
2021-06-26T11:07:39.412978+00:00 app[app.1]: Compiling rustls v0.19.1
2021-06-26T11:07:41.941184+00:00 app[app.1]: Compiling ct-logs v0.8.0
2021-06-26T11:07:43.352893+00:00 app[app.1]: Compiling xsalsa20poly1305 v0.6.0
2021-06-26T11:07:44.047019+00:00 app[app.1]: Compiling tungstenite v0.11.1
2021-06-26T11:07:53.310837+00:00 heroku[app.1]: Process running mem=736M(138.0%)
2021-06-26T11:07:53.312790+00:00 heroku[app.1]: Error R14 (Memory quota exceeded

メモリが悲鳴を上げている

tkt

021-06-26T11:09:00.122097+00:00 app[app.1]: Compiling syntex v0.42.2
2021-06-26T11:09:00.506887+00:00 heroku[app.1]: Process running mem=1072M(207.2%)
2021-06-26T11:09:00.508660+00:00 heroku[app.1]: Error R15 (Memory quota vastly exceeded)
2021-06-26T11:09:00.510568+00:00 heroku[app.1]: Stopping process with SIGKILL
2021-06-26T11:09:00.781304+00:00 heroku[app.1]: Process exited with status 0
2021-06-26T11:09:00.901447+00:00 heroku[app.1]: State changed from up to crashed
2021-06-26T11:09:00.917995+00:00 heroku[app.1]: State changed from crashed to starting
2021-06-26T11:09:24.385858+00:00 heroku[app.1]: Starting process with command `/bin/sh -c cargo\ run`
2021-06-26T11:09:25.156783+00:00 heroku[app.1]: State changed from starting to up
2021-06-26T11:09:29.425477+00:00 app[app.1]: Updating git repository `https://github.com/awslabs/aws-sdk-rust`

限界？

tkt

2021-06-26T11:09:00.122097+00:00 app[app.1]: Compiling syntex v0.42.2
2021-06-26T11:09:00.506887+00:00 heroku[app.1]: Process running mem=1072M(207.2%)
2021-06-26T11:09:00.508660+00:00 heroku[app.1]: Error R15 (Memory quota vastly exceeded)
2021-06-26T11:09:00.510568+00:00 heroku[app.1]: Stopping process with SIGKILL
2021-06-26T11:09:00.781304+00:00 heroku[app.1]: Process exited with status 0
2021-06-26T11:09:00.901447+00:00 heroku[app.1]: State changed from up to crashed
2021-06-26T11:09:00.917995+00:00 heroku[app.1]: State changed from crashed to starting
2021-06-26T11:09:24.385858+00:00 heroku[app.1]: Starting process with command `/bin/sh -c cargo\ run`
2021-06-26T11:09:25.156783+00:00 heroku[app.1]: State changed from starting to up
2021-06-26T11:09:29.425477+00:00 app[app.1]: Updating git repository `https://github.com/awslabs/aws-sdk-rust`
2021-06-26T11:09:33.123622+00:00 app[app.1]: Updating crates.io index
2021-06-26T11:10:13.639242+00:00 app[app.1]: Updating git repository `https://github.com/rcoh/sigv4`
2021-06-26T11:10:14.121754+00:00 app[app.1]: Downloading crates ...
2021-06-26T11:10:14.332188+00:00 app[app.1]: Downloaded base64 v0.12.3
2021-06-26T11:10:14.347448+00:00 app[app.1]: Downloaded audiopus v0.2.0
2021-06-26T11:10:14.358910+00:00 app[app.1]: Downloaded async-tungstenite v0.13.1
2021-06-26T11:10:14.376605+00:00 app[app.1]: Downloaded pretty_assertions v0.7.2
2021-06-26T11:10:14.386350+00:00 app[app.1]: Downloaded proc-macro2 v1.0.27
2021-06-26T11:10:14.394338+00:00 app[app.1]: Downloaded quote v1.0.9

再始動してからは元気っぽい

tkt

メモリエラー出だした

tkt

コンパイルはイメージ作るときにやるべきか、

tkt

Dockerfile 修正してから

heroku container:push app -a <APPLICATION NAME>
heroku container:release app -a <APPLICATION NAME>
heroku ps:scale app=1 -a <APPLICATION NAME>

tkt

heroku logs --tail

これで起動時ログが見れた

tkt

heroku コマンドも docker container 上でやってる

tkt

docker-compose.yml

version: "3.7"
services:
  app:
    build:
      context: .
      dockerfile: ./Dockerfile
      # target: build-env
    working_dir: /workspace
    environment:
      - USER=$USER
      - DISCORD_TOKEN=$DISCORD_TOKEN
      - AWS_ACCESS_KEY_ID=$AWS_ACCESS_KEY_ID
      - AWS_SECRET_ACCESS_KEY=$AWS_SECRET_ACCESS_KEY
      - AWS_REGION=$AWS_REGION
    volumes:
      - .:/workspace:cached
      - target:/workspace/target:cached
    tty: true
  heroku:
    # run `apk add docker` before running heroku commands
    image: sue445/heroku-cli
    volumes:
      - .:/workspace:cached
      - /var/run/docker.sock:/var/run/docker.sock:rw
volumes:
  target:

FROM rust:1.53-slim-buster AS build-env

RUN apt-get update && \
    apt-get install -y \
    libopus-dev \
    build-essential \
    libssl-dev \
    pkg-config \
    autoconf \
    automake \
    libtool \
    m4 \
    ffmpeg \
    curl \
    python \
 && apt-get -y clean \
 && rm -rf /var/lib/apt/lists/*

RUN curl -L https://yt-dl.org/downloads/latest/youtube-dl -o /usr/bin/youtube-dl && \
    chmod a+rx /usr/bin/youtube-dl

ENV LC_ALL=C.UTF-8

COPY Cargo.lock Cargo.lock
COPY Cargo.toml Cargo.toml

COPY  . .

RUN cargo build --release

FROM debian:buster-20210621-slim

RUN apt-get update && \
    apt-get install -y \
    libopus-dev \
    build-essential \
    libssl-dev \
    pkg-config \
    autoconf \
    automake \
    libtool \
    m4 \
    ffmpeg \
    curl \
    python \
 && apt-get -y clean \
 && rm -rf /var/lib/apt/lists/*

RUN curl -L https://yt-dl.org/downloads/latest/youtube-dl -o /usr/bin/youtube-dl && \
    chmod a+rx /usr/bin/youtube-dl

ENV LC_ALL=C.UTF-8

COPY --from=build-env /target/release/app /bin/app

CMD [ "/bin/sh",  "-c", "app" ]

build stage のインストールもう少し削れそう

tkt

2021-06-26T12:46:22.972828+00:00 app[app.1]: thread 'tokio-runtime-worker' panicked at 'failed to create file: Os { code: 2, kind: NotFound, message: "No such file or directory" }', src/tts/mod.rs:75:10
2021-06-26T12:46:31.586284+00:00 app[app.1]: thread 'tokio-runtime-worker' panicked at 'failed to create file: Os { code: 2, kind: NotFound, message: "No such file or directory" }', src/tts/mod.rs:75:10
2021-06-26T12:46:34.149492+00:00 app[app.1]: thread 'tokio-runtime-worker' panicked at 'failed to create file: Os { code: 2, kind: NotFound, message: "No such file or directory" }', src/tts/mod.rs:75:10
2021-06-26T12:47:01.890048+00:00 app[app.1]: thread 'tokio-runtime-worker' panicked at 'failed to create file: Os { code: 2, kind: NotFound, message: "No such file or directory" }', src/tts/mod.rs:75:10

panic してた。

let root = env!("CARGO_MANIFEST_DIR");
let path = Path::new(root);

このあたりが原因な気がする

tkt

let root = env!("CARGO_MANIFEST_DIR");
println!("CARGO_MANIFEST_DIR: {}", root);
let path = Path::new(root);
println!("path: {}", path.display());
let file_path = path.join("binaries").join("tts");
println!("file_path: {}", file_path.display());

CARGO_MANIFEST_DIR: /
path: /
file_path: /binaries/tts

tkt

tokio::fs::File::create

これではディレクトリの作成はやってくれないらしく、存在しない dir を渡すと

Os { code: 2, kind: NotFound, message: "No such file or directory" }

と怒られる

tkt

Dockerfile でディレクトリを作っておく

tkt

通った

tkt

heroku にあげなおした
けど、しゃべらない
エラーログもでてない

tkt

これは謎

tkt

cargo run でも喋らなくなってた
でぐれ？

tkt

test は通る
そんなに数書いてないけど

tkt

デバッグしてるサーバー変えてbotの権限が変わったのが問題かも

tkt

dekita-

tkt

だめだったサーバーでも再招待で動くようになった

tkt

二回目以降の heroku 操作

docker-compose run --rm heroku /bin/bash

heroku login
apk add docker
heroku container:login
heroku container:push app -a <APP_NAME>
heroku container:release app -a <APP_NAME>
heroku ps:scale app=1 -a <APP_NAME>

tkt

// 声質の変更と最大再生秒数の設定
// https://docs.aws.amazon.com/ja_jp/polly/latest/dg/supportedtags.html
ssml_text.push_str("<speak><prosody pitch=\"+200%\"><amazon:effect phonation=\"soft\"><amazon:effect vocal-tract-length=\"-15%\"><prosody amazon:max-duration=\"5s\">");
ssml_text.push_str(&content);
ssml_text.push_str("</prosody></amazon:effect></amazon:effect></prosody></speak>");

tkt

一旦できたのでClose

tkt

terraform の環境使わなくなったから別で書いといた

tkt

今やるならこの機構をもとにやれば楽そう
https://github.com/aquelemiguel/parrot

tkt

スラッシュコマンド対応とかなんやかんやでまたたくさんcommitしてる