This repo contains the code and detailed instruction to reproduce the intelligent Slackbot I built in this blog post. Powered by the torchMoji model, the bot automatically reacts to Slack messages with emoji.
Before proceeding, make sure you have installed python3 (the repo is tested with python3.6), wget, and zip.
Also, install and configure AWS CLI.
If you are deploying the model on SageMaker, you also need docker (installation guide here).
If you are deploying the model on Algorithmia, you also need pip3 (the repo is tested with pip3.6).
Overall, there are 5 steps:
- Set up the Slackbot by completing steps 1-5 here.
- Deploy the torchMoji model on SageMaker or Algorithmia (or both!). See here for SageMaker deployment and here for Algorithmia deployment.
- Deploy the AWS Lambda function that performs request authentication and routes traffic between Slack and the prediction service. Instruction here.
- Set up the API gateway that routes traffic to the Lambda function and exposes a REST endpoint to Slack. Instruction here.
- Finish setting up the Slackbot by completing step 6 here.
After completing all the steps above, send a direct message "this is great!" to your bot (not to yourself!) in the Slack workspace. You should see the following.
You can even add the bot to channels by doing the following:
- Click "Apps" on your Slack workspace (upper-left) and click the name of your bot.
- Click "Details" on the upper-right corner, click the "..." icon that says "More", and click "Add this app to a channel".
- Select the channel you want the bot to join and click "Add".
In my case, I have added the bot to the "test" channel, and here we go:
When you are done having fun, remember to delete all the AWS resources you created to save $$$!