Been looking for a way to use Dialogflow to make a Twitch bot? Look no further! Here's a quick and easy way to link up a Dialogflow agent to a Twitch account and channel.
The idea is that you should be able to allow your Dialogflow bot to respond to messages in chat, both from direct @ mentions and also just joining in the overall conversation when appropriate. The latter is a bit of a different use that Dialogflow isn't typically used for, so we've got a bit of additional settings for that to make sure the bot doesn't go a bit wild and respond to everything!.
It's still early days, so if you spot issues, definitely raise them here.
Want to give it a go? Do the following:
- Clone this repo somewhere where you'll want to run a Node server (I've typically just run it locally on my PC/Mac). Cloning should be as simple as:
git clone https://github.com/patcat/DialogflowTwitchBotBridge.git
- Open up
server.jsand changeprojectIdto your own access token from Dialogflow (you can get that from https://dialogflow.com/docs/agents#settings).
projectId = "YOURTOKENHERE",
- Change
dialogflowSessionIDto any random series of characters you'd like to use (all one word or separated with-should be fine):
dialogflowSessionID = 'twitchbotbridge-session-id',
- Change
dialogflowLanguageCodeif you're using a different language than'en-US'(I haven't ever tried an alternative language, so if you try this, let me know how it goes!),
dialogflowLanguageCode = 'en-US',
- Change
genericIntentsto list which intents in your Dialogflow agent should apply to the general chat in the room (e.g. run them even if your bot isn't mentioned directly in the message):
genericIntents = [{
intent: "smalltalk.greetings.hello",
freq: 0.5
},
{
intent: "smalltalk.greetings.bye",
freq: 0.5
}],
You can change the frequency they're run too. For things like hello and goodbye... the bot will be watching out for those all the time — so if everyone is saying the word "Hello", your bot would theoretically try to respond to that every time. freq: 0.5 adds a bit of randomness to it and sets up a 50% chance it'll actually respond instead. Use this to prevent your bot from constantly triggering generic responses, while also letting it get involved in chat occasionally.
- Add in your Twitch channel you'd like the bot to be chatting in:
// Twitch settings
twitchChannel = '#devdiner',
- Add in the bot's name (needs to be an actual account you've got access to!):
twitchBotName = 'TheGreatIntelligenceBot',
- You'll need a token that'll allow you to post on that account's behalf. Grab one from https://twitchapps.com/tmi/ if you don't already have one. Then put it into the
twitchPasswordfield:
twitchPassword = 'oauth:YOUROAUTHHERE',
- When it comes to training up your bot, it can be a bit awkward for the messages in Dialogflow to all have
@YourBotUsernameconstantly. That's not really pronouncable or what Dialogflow expects. So instead, we set a name that isn't the username to replace any instances of the username with (e.g. "Dr" for@DrArnoldBernard, you could use say "Barry" for@BarryAllenRunBarryRunas another example):
genericName = 'Bot',
- You could want your bot to respond instantly to anything, but that can be offputting. People don't respond instantly, they take time to read a message. Let's not show just how speedy and superior our bot is to its fellow Twitch humans by adding in a delay to mimic them reading the message first. The delay is in milliseconds:
typicalDelayBeforeResponding = 1000,
- Humans also take time to type a message. That time is different depending on how long that message was. The Dialogflow Twitch Bot Bridge can mimic that too! You can decide how many characters per minute your bot can type and we'll delay its responses accordingly:
charactersPerMinute = 300,
- Finally, Dialogflow has a fallback for when it doesn't understand something. If someone @ mentions your bot, you can decide whether it should respond or not if it doesn't know what they mean. If your bot isn't quite smart enough yet to respond to everything — and you don't want your chat spammed with "I didn't understand" type messages — you can tell Dialogflow Twitch Bot Bridge to just not respond if it doesn't understand instead.
keepQuietOnFallback = true;
- Once all is updated, run the following from root of your cloned folder to install all dependencies:
npm install
-
You'll also need a Google Cloud Platform project and the assorted steps ready as per the instructions from the (you should have a file in your environment providing Google Application Credentials). The steps from the Dialogflow Node.js Client Library provide the following (from 15 - 19)
-
Select or create a Cloud Platform project.
-
Enable billing for your project.
-
Enable the Dialogflow API.
-
Set up authentication with a service account so you can access the API from your local workstation.
-
Then, once everything is all good to go Google Cloud Platform wise, run it like so:
node server.js
I'm releasing an online course on Dialogflow and I'll have a guide that'll explain a lot more of the process of building a Dialogflow bot there! If you're interested, register your interest for the Dialogflow content here!
This API bridge is entirely free to use with a glorious MIT license — but by downloading it, you agree to be responsible with it. With great power comes great responsibility! Go forth and build cool things! Play nice.
Please check Twitch terms and conditions, along with other relevant rules and regulations, before using this Dialogflow Twitch Bot Bridge. You are responsible for anything that goes wrong while you use this, please use it with care!