RIFT DevOps Agent

Agents that clone a repo, find what breaks, fix it and open a pull request.

RIFT agent log streaming a run: environment detected, two failures found, fixes applied for an unhandled promise rejection and a missing dependency, iteration two starting.

The problem

When CI fails, someone has to read the logs, find the cause and push a fix before anything else can merge. Much of that loop is mechanical, and it is the kind of work an agent can take on if it runs safely.

How it works

Four agents run in sequence: one clones and prepares the repository, one runs static analysis and collects failures, a healing agent asks Gemini for fixes and applies them, and a git agent commits and opens a pull request through the GitHub API. The pipeline repeats until the checks pass or an iteration limit is reached.

Each run executes inside a Docker worker container, so the repository's code runs isolated from the server. A FastAPI server with Socket.IO streams logs to a Next.js dashboard, and the backend was deployed on Azure.

Highlights

  • Repo prep, analysis, healing and git agents in one loop
  • Sandboxed execution in Docker
  • Live log streaming over Socket.IO
  • Pull requests opened automatically with the fixes

Screens

RIFT agent start screen with fields for repository URL, team name and team leader, and a Deploy Agent button.
Start a run with a repository URL.