Slack Translate Bot

This part of the repo provides and implementation of a Slackbot which can perform either one-to-one translation (e.g. Japanese <-> English), or many-to-one translation (e.g. anything -> English).

The target language of the bot (along with some other variables) is set in the environment, rather than the code or a static file, to make it easier to deploy multiple bots with different languages.

Screenshots

Example of a many-to-one translation bot for English:

Example of a bilingual one-to-one translation bot for Japanese and English:

Setup and Configuration

Create the Slack bot user

First, create a bot user for your team and configure the name, description, photo, etc as you wish. See the Slack Bot Documentation if you do not know how to do this. You will need the token on this page to run your bot.

env variables

The bot depends on a few environment variables:

• BOT_TOKEN: (required) Slack bot token obtained when creating the bot user
• TARGET_LANG: (required) official 2-letter code for the target language
• SOURCE_LANG: (optional) official 2-letter code for the source language. Defining this variable will result in a one-to-one translation bot, not defining it will result in a many-to-one translation bot.
• GOOGLE_CREDENTIALS: (optional) path to GCP credentials JSON file (e.g. service account). If deploying in Google Cloud, you don't need to supply this unless you have specific requirements for it. When running locally, failing to supply this mean you are using the public-facing Translate API and will be cut off after a small number of requests.
• GREETING: (optional) greeting message to display when the bot joins a channel
• APOLOGY: (optional) apology message to display when the bot encounters an error
• ILLITERATE: (optional) message to display when a one-to-one bot is asked to translate a language it does not speak

Save these values in typical bash syntax inside of a file (see .sample-env.one2one and .sample-env.many2one in this repo).

Node

The env values we defined are obtained through the dotenv package when running directly in node:

node -r dotenv/config index.js dotenv_config_path=my-env-file

Docker

You don't have to run the bot in Docker, but there are some nice benefits of doing so.

You can build the Docker image in the root of this repo like this:

docker build -t slack-translate-bot .

It is best practice to never include sensitive data like credentials or API keys inside of a Docker image. The .dockerignore supplied in this repo will ignore a credentials/ directory, which is where you can place your Google service account JSON. Since the env file you created specifies a path to a credentials JSON, you must mount the directory as a volume in your Docker container, so it can be found at runtime (that means you may need to alter your GOOGLE_CREDENTIALS variable to match your desired path inside the container);

e.g.:

docker run -d \
    --name bot-en
    --env-file my-env-file
    -v $(pwd)/credentials/:/credentials
    slack-translate-bot

Assuming I set GOOGLE_CREDENTIALS=/credentials/credentials.json and I have a file ./credentials/credentials.json, then the container ran in the command above should have no trouble.

Deploy in Google Cloud

Note that if you are reploying this in Google Container Engine (GKE), then you don't even need to specify or provide credentials at all, as the defaults will automatically be used.

If you still want to use a specific service account, then you should probably supply it as a Kubernetes secret.

TODO: add instructions for GKE deployment