mastodon-gpt is an open source serverless conversational chatbot that checks every X minutes
to see if it was mentioned on Mastodon. When the AWS Lambda function detects a mention, it uses an OpenAI language model to generate a response based on the context of the mention (and the prior message if it exists) and then posts the response back to Mastodon as a reply.
The first deployed instance of this bot is @gpt@mastodon.solar.
Each mention (along with the prior message if present) is checked against the OpenAI Moderations endpoint.
If the bot is asked to reply to content that fails any moderation check by OpenAI, then the bot will simply ignore the mention.
Infrastructure is deployed to AWS via Terraform from the ./terraform folder. The ./terraform/01-bootstrap layer must be applied first. See each layer's folder for deployment and config details.
