Merge pull request #17 from femshima/model-inside-container

Model inside container

Dockerfile

@@ -15,10 +15,21 @@ RUN npm ci
 COPY ./src/ ./src/
 RUN npm run build
 
+FROM node:18 AS model-fetch
+
+WORKDIR /app
+RUN wget https://github.com/jpreprocess/jpreprocess/releases/download/v0.6.1/naist-jdic-jpreprocess.tar.gz \
+    && tar xzf naist-jdic-jpreprocess.tar.gz \
+    && rm naist-jdic-jpreprocess.tar.gz
+RUN wget http://downloads.sourceforge.net/open-jtalk/hts_voice_nitech_jp_atr503_m001-1.05.tar.gz \
+    && tar xzf hts_voice_nitech_jp_atr503_m001-1.05.tar.gz \
+    && rm hts_voice_nitech_jp_atr503_m001-1.05.tar.gz
+
 FROM gcr.io/distroless/nodejs18-debian12:nonroot AS runner
 WORKDIR /app
 ENV NODE_ENV=production
 COPY ./package.json ./
 COPY --from=builder /app/dist/ ./dist/
 COPY --from=deps /app/node_modules/ ./node_modules/
+COPY --from=model-fetch /app/ ./model/
 CMD ["dist/main.js"]

README.md

@@ -8,39 +8,34 @@ Discord.js Japan User Group向けの読み上げボットです．
 `DISCORD_TOKEN=`の後に続けて，[Discord Developer Portal](https://discord.com/developers/applications)
 から取得したトークンを記述してください．
 
-次に，`model`ディレクトリを作成し，ダウンロードした辞書と音声モデルを置いてください．
-辞書は
-[naist-jdic-jpreprocess.tar.gz](https://github.com/jpreprocess/jpreprocess/releases/download/v0.6.1/naist-jdic-jpreprocess.tar.gz)，
-モデルは
-[hts_voice_nitech_jp_atr503_m001-1.05.tar.gz](http://downloads.sourceforge.net/open-jtalk/hts_voice_nitech_jp_atr503_m001-1.05.tar.gz)
-をお勧めします．
-いずれもダウンロードした後解凍が必要です．
-
-それぞれ次の図のように，パスが`model/naist-jdic`，`model/nitech_jp_atr503_m001.htsvoice`となるように配置してください．
-```
-(project-root)
-┣ model
-┃  ┣ naist-jdic
-┃  ┃  ┣ char_def.bin
-┃  ┃  ┣ dict.da
-┃  ┃  ┣ dict.vals
-┃  ┃  ┣ dict.words
-┃  ┃  ┣ dict.wordsidx
-┃  ┃  ┣ matrix.mtx
-┃  ┃  ┗ unk.bin
-┃  ┗ nitech_jp_atr503_m001.htsvoice
-┗ .env
-```
-
-最後に，次のコマンドを実行すると，ボットが起動します（Dockerがインストールされていることが必要です）．
+次のコマンドを実行すると，ボットが起動します（Dockerがインストールされていることが必要です）．
 
 ```bash
-docker run \
-  --rm \
-  -d \
-  -v ./model:/app/model \
-  --env-file .env \
-  -e DICTIONARY=model/naist-jdic \
-  -e MODEL=model/nitech_jp_atr503_m001.htsvoice \
-  ghcr.io/discordjs-japan/om:latest
+docker run --rm -d --env-file .env ghcr.io/discordjs-japan/om:latest
 ```
+
+## Copyright Notice
+
+For copyright of dependent packages, please see package.json.
+
+### [HTS Voice "NIT ATR503 M001" version 1.05](http://downloads.sourceforge.net/open-jtalk/hts_voice_nitech_jp_atr503_m001-1.05.tar.gz)
+
+> [!NOTE]
+> HTS Voice is only included in Docker container
+
+Creative Commons Attribution 3.0
+
+- Copyright (c) 2003-2012  Nagoya Institute of Technology Department of Computer Science
+- Copyright (c) 2003-2008  Tokyo Institute of Technology Interdisciplinary Graduate School of Science and Engineering
+
+### [naist-jdic](https://github.com/jpreprocess/jpreprocess/releases/download/v0.6.1/naist-jdic-jpreprocess.tar.gz)
+
+> [!NOTE]
+> naist-jdic is only included in Docker container
+
+BSD 3-Clause License
+
+- Copyright (c) 2009, Nara Institute of Science and Technology, Japan.
+- Copyright (c) 2011-2017, The UniDic Consortium
+- Copyright (c) 2008-2016 Nagoya Institute of Technology Department of Computer Science
+- Copyright (c) 2023, JPreprocess Team

src/synthesis/index.ts

@@ -1,10 +1,8 @@
 import { Synthesizer } from "./synthesizer";
 import WorkerSynthesizer from "./worker-synthesizer";
 
-if (!process.env.DICTIONARY || !process.env.MODEL) {
-  throw new Error("Dictionary and model must be specified.");
-}
 export const synthesizer: Synthesizer = new WorkerSynthesizer(
-  process.env.DICTIONARY,
-  process.env.MODEL,
+  process.env.DICTIONARY ?? "/app/model/naist-jdic",
+  process.env.MODEL ??
+    "/app/model/hts_voice_nitech_jp_atr503_m001-1.05/nitech_jp_atr503_m001.htsvoice",
 );