Redis Backplane for Hubots

hubot-redis-backplane leverages Redis Streams to create a distributed backplane for Hubot.

This module is a Hubot Adapter. If you npm install it, you would start a Hubot instance with -a @hubot-friends/hubot-redis-backplane and it will connect to Redis, create a Consumer Group and register itself as a Consumer to the specified Redis Stream (e.g. inbox) and "listen" for incoming messages. The term "Adapter" here is just the fact that this connection utlizes Hubot's Adapter design – technically speaking. So in this part of the setup, Redis Streams is the message input and the Hubot Consumer connects to a stream as it's source of messages.

At this point, it does nothing until messages are added to the inbox stream. Note. You could add messages manually via the Redis CLI if you wanted and the Hubot Consumer would process those messages and add it's response to a Redis Stream – outbox. Referring to the right side of the diagram below, we've started a Hubot Consumer # 1, part of a Redis Consumer Group. There can be multiple Consumer Groups. I think of a Consumer Group as providing a set of functionality. You'll typlically just have a single Consumer Group where all the Hubot Consumers come from a single codebase, but have different "consumer ids". I'm sure we could go crazy here with scaling functionality out with multiple Consumer Groups. But let's try to keep things simple first before we jump to Hyperspace.

The next part of the design is the left side of the diagram above – the input.

This Hubot instance's role is the adapter to the chat platoform (e.g. Slack, MS Teams, Discord). You'll still start this instance with something like hubot -a @hubot-friends/hubot-discord because the chat platform is how users interact with Hubot – the input. But now, this Hubot instance won't have all you're command handling scripts; it'll have the inbox Hubot script located in this package.

The steps to manually install this script is:

• npm i @hubot-friends/hubot-redis-backplane

• Add hubot-redis-backplane/inbox.mjs to your external-scripts.json file

  [ "hubot-redis-backplane/inbox.mjs" ]

If you want to start from scratch:

• cd folder-to-have-mybot Probably use a different folder name

• npx hubot --create myhubot --adapter @hubot-friends/hubot-discord Pick your adapter. Just using Discord as an example here.

• cd myhubot

• npm i @hubot-friends/hubot-redis-backplane

• Add hubot-redis-backplane/inbox.mjs to your external-scripts.json file

  [ "hubot-redis-backplane/inbox.mjs" ]

Now when you start the Hubot Adapter instance as shown in the diagram above, incoming chat messages will be pushed into a Redis inbox stream, Hubot Redis Consumers get and process these messages, and responses are written back to an outbox Redis stream for delivery back to the chat. This lets you scale Hubot horizontally across processes or servers while keeping the UX seamless.

Don't forget, you need a Redis server.

My assumption is that this is a Git repo that contains all of your Hubot scripts (i.e. in a folder called scripts).

NAME_OF_YOUR_HUBOT is the name you gave your Hubot instance that is connected to the chat app.

Say you're building a Discord bot with Hubot. You have a start task in your package.json file like so:

Starting the instance with:

sets the name of your Hubot. So when you're on Discord and you want to send your Hubot a message, you'd do it like (where name is mybot in this example), @mybot help.

You might have the name set in the start script like so:

In which case, you wouldn't need to include the -- --name ${NAME_OF_YOUR_HUBOT} becasue it's already set in the start task.

• HUBOT_REDIS_INBOX_URL (default: redis://localhost:6378)
• HUBOT_REDIS_INBOX_STREAM_NAME (default: hubot-inbox)
• HUBOT_REDIS_OUTBOX_STREAM_NAME (default: hubot-outbox)
• HUBOT_REDIS_OUTBOX_GROUP_NAME (default: hubot-group)
• HUBOT_REDIS_OUTBOX_CONSUMER_NAME (default: consumer-1)

Where each consumer instance has a unique value for it's consumer name.