Rufler

# Rufler

**One command. One YAML file. A whole autonomous AI swarm.**

`rufler` is a Python wrapper around [`ruflo`](https://github.com/ruvnet/ruflo) (the TypeScript AI agent orchestration CLI from the `claude-flow` family) that turns a dozen bootstrap commands into a single, declarative `rufler_flow.yml` file.

Instead of memorizing `ruflo init → daemon start → memory init → swarm init → hive-mind init → hive-mind spawn --claude --objective=...`, you write a YAML file that describes your project, your agent team and your task, and run:

rufler run

rufler handles preflight checks, daemon lifecycle, swarm/memory/hive-mind init, objective composition, logging, multi-task sequencing, AI-driven task decomposition, background supervision and live log following.

---

Why rufler and not ruflo directly?

`ruflo` is powerful but low-level. It exposes ~26 commands and 140+ subcommands that all have to be called in the right order with the right flags. For day-to-day driving of an autonomous Claude-Code swarm, that's a lot of ceremony.

Problem with raw ruflo	How rufler fixes it
6–10 shell commands to boot a project (init → daemon start → memory init → swarm init → hive-mind init → hive-mind spawn --claude --objective=...)	One command: rufler run reads your YAML and runs the whole pipeline.
Objective prompt must be assembled by hand — project name, task body, every agent's role and prompt, autonomy flags	Auto-composed objective from rufler_flow.yml — agents are sorted lead → senior → junior and injected into a single coherent prompt.
No native multi-task support — you spawn one hive per task manually	Multi-task mode (task.multi: true) — run a list of subtasks sequentially or in parallel from one YAML.
No task decomposition	**AI