Skip to content

qbart/gaia

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

33 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

GΛIΛ

GAIA is golang based implementation of Ralph Loop (basically "AI continous delivery" with pushes to current branch). AI instructions and rules are fed from issues so everything can be controlled remotely via tasks with appropriate labels.

Installation

  1. Download project
  2. make build (ensure go toolchain is installed)
  3. make install (it will copy binary to ~/bin/gaia)

Run

export PAT=... # github token 
gaia --god --project qbart/gaia --model sonnet
  • --god - enters skip permission mode (default permissions are auto, usually you want sandboxed environment and god mode on)
  • --provider - "github" or "trello"
  • --project - source of tasks (trello "board id" or github "owner/repo")
  • --model - overwrite the default claude model (default opus)
  • --wait - wait time if no tasks (before next fetch, default 30s)
  • --env-file - path to envs (needed for providers, PAT for Github, or TRELLO_KEY/TRELLO_TOKEN)

Following lables must exist in repo (they are created when gaia starts):

  • docs (instructions for AI)
  • todo (tasks to be picked up by claude)
  • doing (claude picked it up and it is currnently working on it)
  • review (claude finished and marked the issue for review by us, it will not be picked unless rejected)
  • rejected (review is rejected and claude will pick it up again and apply feedback from comments)
  • done (review is approved and we can close the task)
  • brainstorm (instructions for AI to come up with new tasks when nothing to work on, no brainstorm = no brainstorming)

Tasks are implemented in the following order: doing, rejected, todo.

When starting a task, all docs are concatenated into single prompt followed by task name and description. When rate limit is hit, wait step switches to 5 minute window.

Happy burning tokens! 🔥

⚠️ API based usage is not recommended as it might generate huge cost!

About

AI continous delivery and self-improvement

Resources

Stars

Watchers

Forks

Contributors