A simple GPT bot which can generate a shell script to perform the required task. The generated script is run within a docker container, with any files or directories it requires from the host mounted as read-only volumes.
genbot.mov
make install
This copies the bot's config as well as the bot script to the
default config directory: ~/botmand-engines.
Edit ~/botmand-engines/genbot.yaml and set the OPENAI_API_KEY variable.
Check OpenAI documentation on signing up
for the API and retrieving your API
key.
If botmand is already running, make it reload its engines:
pkill -HUP botmand
The bot only listens to messages directed at it in specific channels
(by default, #genbot). See ~/botmand-engines/genbot.yaml after installation
to configure this.
When issued an instruction to perform a specific task (eg., to monitor a specific file and print an alert when its contents change), the bot generates a shell script which it then runs by the user for confirmation. Once the user accepts the script, it generates a genbot (a docker image with the script) and runs it.
Currently running genbots can be listed by the command "genbot ls", which prints a list of running genbots along with their IDs. Any running genbot can be stopped by entering the command: "genbot stop ".