jtheletter/TweetsLineByLine

Written works in tweets line by line.

Tweets Line by Line

Written works in tweets line by line.

Code Dependencies:

• NPM Twit for Twitter authentication and posting.
• NPM AWS-SDK for index storage and retrieval.
• AWS CLI for CLI deployment. Installed globally and directly from Amazon. Version 2 (latest) used.

Platform Dependencies:

• Twitter.
• Twitter Developer Platform.
• AWS IAM for lambda access.
• AWS Systems Manager (SSM) for index storage.
• AWS Lambda for code storage and execution.
• AWS CloudWatch for scheduling and logs.
• Optional: AWS Simple Notification Service (SNS) for emailed errors.

History:

• This bot initially served only one written work.
• It was first written for Heroku. Heroku reboots worker dynos every 24 hours, which reset the index to zero and prevented the sequence of tweets from continuing in order.
• It was next adjusted to run on AWS EC2. This proved more expensive than desired.
• It was then configured to run on AWS Lambda, with its index saved in AWS Systems Manager (SSM)’s Parameter Store.
• Finally, the bot can now serve multiple works.

Instructions:

For all written works:

• Install AWS CLI. Access keys are stored in ~/.aws/credentials.
• Create a Twitter dev app.
• Create an AWS account.
• Create an AWS IAM role.
  • Set type to AWS service.
  • Set use case to lambda.
  • Set permissions to Lambda Full Access and SSM Full Access.
• Optional: Create an AWS Lambda function to email logged errors automatically (see logEmailer.js).

For each written work:

• Save the work’s lines to the lines directory, and add its name and handle to the works.js list.
• Create a Twitter account.
• Generate tokens/keys to authorize the Twitter app to post to the Twitter account:
  • Visit this URL on the dev account’s browser to get new temporary/one-use tokens:
    • https://twitter.com/oauth/request_token?oauth_consumer_key=API_KEY_FROM_TWITTER_DEVELOPER_PORTAL&oauth_callback=oob
  • Visit this URL on the bot account’s browser, with the oauth_token from DOM output of the previous step, and click “Authorize app”:
    • https://twitter.com/oauth/authenticate?oauth_token=OAUTH_TOKEN_FROM_PREVIOUS_STEP
  • Run this curl POST in Terminal, with the oauth_token and oauth_verifier PIN from the DOM outputs of the previous steps:
    • curl --request POST --url 'https://twitter.com/oauth/access_token?oauth_token=OAUTH_TOKEN_FROM_PREVIOUS_STEP&oauth_verifier=PIN_FROM_PREVIOUS_STEP' && print ;
  • Save the output oauth_token for the access_token, and the oauth_token_secret for the access_token_secret in the Twitter config:
    • oauth_token=ACCESS_TOKEN&oauth_token_secret=ACCESS_TOKEN_SECRET&user_id=USER_ID&screen_name=SCREEN_NAME
• Save tokens/keys to config/twitter/<work>/dev.js.
• Create an AWS SSM Parameter Store named /<work>/index with string/text value 0. Save its region in config/aws/<work>/.
• Create an AWS lambda function.
  • Author from scratch.
  • Set the function name to the work’s name.
  • Change the default execution role to the IAM role.
  • Configure general timeout to six seconds (each execution gets an index from AWS, posts text to Twitter, and saves a new index to AWS; together this averages three seconds in CloudWatch).
  • Configure environmental variables:
    • Add NODE_ENV as production.
    • Add WORK as the work’s name.
    • Add per Twitter config:
      • ACCESS_TOKEN (work bot’s oauth_token).
      • ACCESS_TOKEN_SECRET (work bot’s oauth_token_secret).
      • CONSUMER_KEY (dev app’s API Key).
      • CONSUMER_SECRET (dev app’s API Secret Key).
• Create an AWS CloudWatch event rule.
  • Schedule a fixed rate to the desired minutes.
  • Add the corresponding lambda as its target.
  • Name it the same as the work.
  • Toggle its state to enabled when ready to begin automated executions.
  • NB: AWS will not execute upon enablement during creation, but it will execute upon enablement after creation or upon saving edits to the rule.
• View Log groups in AWS CloudWatch. Reduce retention settings from the “never expire” default.
• Optional: Add Log group as a trigger to logEmailer.js Lambda function, filter pattern ERROR, for automated error emails (filter name is email subject).
• Command-line scripts:
  • Break paragraphs into sentences (overwrites the work’s “lines” file):
    • WORK=<work> npm run break-paragraphs
  • Check lines meet Twitter limits:
    • WORK=<work> npm run check-lines
  • Break lines to meet Twitter limits (overwrites the work’s “lines” file):
    • WORK=<work> npm run break-long-lines "<breaker>"
  • Confirm Twitter credentials:
    • WORK=<work> npm run confirm-twitter
  • Get index from AWS storage:
    • WORK=<work> npm run get-index
  • Set index in AWS storage (use -- before index if -1):
    • WORK=<work> npm run set-index <index>
  • Zip code for deploy:
    • npm run zip
  • Deploy the zip (increase timeout in script if needed):
    • WORK=<work> npm run deploy
  • Execute handler manually:
    • WORK=<work> npm run execute
  • Tweet a special announcement without iterating index:
    • WORK=<work> npm run psa <announcement>