Twitter Vision Bot

A Twitter bot that uses the Google Vision API to understand the content of an image. Machine learning in bot form. Not a useful bot, it has to be said. There's 2 different functions of this bot; identification of what the image is of, and face detection.

Identify Bot

This will wait for an image to be posted and attempt to identify what is in it.

Pixelate Face Bot

This will wait for an image to be posted and attempt to find faces in the image. Any found will be pixelated and posted back in reply.

bots.monitor(@someone/#something, visionId);
bots.monitor(@someone/#something, visionFace);

I'd suggest monitoring a user as Twitter does not like bots to go all rampant on them. It could get ugly. And a free heroku account might not last long if you monitor the twitter stream for @bieber

Live

Monitoring @b_o_t_1. Send it a picture of a face. It is only doing the pixelate trick at the moment. Being a bot is tiring work.

Run Locally

node index.js

Requirements & Setup

Twitter Apps

Create an app at https://apps.twitter.com/ Go to Keys and Access Tokens Note down:

Consumer Key (API Key)
Consumer Secret (API Secret)
Access Token
Access Token Secret

Copy to the file .env.sample, renaming it to .env Double check .env is in .gitignore - you never want these keys in your repo.

Google Vision API

Before You Begin

Create a new project or select an existing one at Google Developers Console In the API Manager:

Turn on the Google Cloud Vision API
Go to Credentials and create a new Service Account. Create and download a json key to authenticate your bot to Google. Keep it really safe, really private! Do not add it to github. Ever.

The usual way to use the json keyfile is to provide a path to this file:

const vision = require('@google-cloud/vision')({
  projectId: 'your-bot',
  keyFilename: '/path/to/keyfile.json'
});

However, for the live version - at least when dealing with 3rd party servers like Heroku - that would mean adding the keyfile to the git repo and pushing to the server. Again - very bad.

The alternative way (which was a nightmare to figure out, thanks docs :-|) is instead to add another environment variable to .env. Name it VISION_KEYFILE_JSON and set the value to the contents of your keyfile.json. Remove whitespace, paste it in, with no surrounding quotes.

CONSUMER_KEY=ABC123
CONSUMER_SECRET=DEF456
ACCESS_TOKEN=123ABC
ACCESS_TOKEN_SECRET=456DEF
VISION_KEYFILE_JSON={"type":"service_account","project_id":"your-bot","and_all":"the_rest_form_keyfile"}

You can then init google-cloud/vision with:

let credentials;
try {
  credentials = JSON.parse(process.env.VISION_KEYFILE_JSON);
} catch(error) {
  throw error;
}

const vision = require('@google-cloud/vision')({
  projectId: 'your-bot',
  credentials
});

For the live version, remember to add all 5 environment variables. In Heroku, do that under Settings > Config vars.

You also want the bot running as a worker, not a web app. For Heroku: Add a Procfile, go to Resources and turn on that worker, and off the web app.

mikeybyker / twitter-vision-bot Goto Github PK