Rickybot is a Bluesky (social media) automation routine intended to manage account growth.
- it browses feeds (lists of posts) and uses AI to identify posts with media that aligns with your interests
- (so long as those interests are cats. It uses Google's Visual Transformer (ViT) to create labels for images in posts, and I've written an assessment algorithm to see if they are 'real' cat posts. Real being not memes and not drawings. I plan on trying to make this more flexible in the future, but might only be feasible to tune it to a specific use-case basis.)
- it follows users with a high likelihood of reciprocating follows and improving your engagement rate
- (when a cat post is identified it likes the post, follows the user, and follows users that liked that post if the likes are above a certain threshold)
- performs follower management
- (after a week the last week's follows are evaluated, and any users that did not follow back are pruned so that your follow:follower ratio does not get out of control)
- automates its tasks using AWS services
To run this program you need to have:
- AWS account with SecretsManager, S3, and DynamoDB set up.
- A huggingface login (not necessary since ViT is public but it helps).
- A github repository set up if you want to use the logging features.
- Your own BlueSky account
Downloading and running these colabs should not allow you to actually run the code without plugging in colab secrets and/or aws secretmanager secrets where necessary.
To run the lambdaized version of the Add Followers you need to go to the progress files and use the local_download.py to download the ViT files into a folder. config.json, model.safetensors, and preprocessor_config.json