Publish your Twitter likes in a cheap serverless manner. Uses AWS Lambda and Amazon S3 for collection and storage.

• Example: https://like.sorah.jp

• AWS account
• Terraform
• jrsonnet
• Node.JS (npm)
• Ruby

Deploy infrastructure using Terraform. This repository works as a terraform module:

module "prd" {
  source = "github.com/sorah/publikes"

  iam_role_prefix = "PublikesPrd"
  name_prefix     = "publikes-prd"
  s3_bucket_name  = "..."
  app_domain      = "like.example.com"

  certificate_arn       = data.aws_acm_certificate.my-certificate.arn
  cloudfront_log_bucket = "example.s3.amazonaws.com"
  cloudfront_log_prefix = "like.example.com/"
}

# outputs:
# - module.prd.cloudfront_distribution_domain_name to create a DNS record
# - module.prd.cloudfront_distribution_id for deploy.rb (see below)
# - module.prd.lambda_function_url for ingestion webhook (see below)

cd ui/
npm i
vim .env.production.local
npm run build
ruby deploy.rb $S3_BUCKET_NAME $CLOUDFRONT_DISTRIBUTION_ID

Refer to ui/.env for available environment variables.

You need to manually setup AWS Secrets Manager Secret {name_prefix}/secret with following key-value values:

• ingest_secret: Secret string used for webhook. Send as x-secret in webhook requests to verify its authenticity.

To avoid paying 100 USD/mo for Twitter API, this system uses IFTTT to feed likes information to store. You can use [New liked tweet by you] trigger with [Make a web request] action.

• URL: {lambda_function_uri}/publikes-ingest
• Content-Type: application/json
• Headers: x-secret: {secret}
• Content: {"url": "<<<{{LinkToTweet}}>>>"}

The lambda function behind Function URL enqueues incoming tweet URL to the SQS queue. Then the messages will be consumed by an another Lambda function, and the included Tweet URLs are inserted into the latest page.

The page of tweet IDs are rotated every N tweets (MAX_ITEMS_IN_HEAD) or after 6 hours using [rotate-batch] Step Functions State Machine. This system carefully uses S3 for ingestion not to happen any data loss by ingesting into the single object (such as race condition and conflicts). Partial data are eventually completed, especially during the merge process of [rotate-batch] state machine run.

MIT License