AzTwitterSAR -- tweet analysis and notification

This is an Azure Function that polls Twitter's API to obtain new tweets from a given account. Then it scores the new tweetby the presence of specific words. If this indicates that the tweet may be of interest, then a machine learning-based filter function (see the sibling project TwitterSARai) is called. If that, too, indicates relevance, then the tweet is posted to a slack channel through a webhook.

Prerequisites

• An Azure account and decent knowledge of deploying and running a serverless solution. ARM templates are included, i.e. the Azure infrastructure can be set up mostly automatically.
• Twitter API key + secret.
• The environment variable AZTWITTERSAR_SLACKHOOK must be set with a link to a web hook that should be used to post the notification. For local testing the definition must be in local.settings.json, on Azure it must be added in the deployment template parameters so that it can be written to the key vault.

Workflow when redeploying the code

• Send a POST request to the terminatePostUri (obtained when starting the workflow).
• Send a POST request to the purgeHistoryDeleteUri.
• Stop the Azure function.
• Optionally: download the tweets from the storage account (can be done at any time).
• Optionally: delete or expunge the tables in the storage account.
• Run infrastructure and/or code deployment.
• Start the Azure function.
• Start the workflow by triggering the starter function from the Azure portal with the id of the last treated tweet as argument. Save the response with the durable function management webhooks.