Skip to content

README.md

Latest commit

 

History

History
104 lines (70 loc) · 3.4 KB

README.md

File metadata and controls

104 lines (70 loc) · 3.4 KB

Claude Code Configuration

This directory contains configuration for using this project with Claude Code, particularly in the online environment.

SessionStart Hook

The hooks/SessionStart script automatically runs when a new Claude Code session begins. It:

  1. Installs pixi (if not already installed) - The package and environment manager
  2. Installs project dependencies - All Python packages and tools defined in pyproject.toml
  3. Sets up prek hooks - For automatic code quality checks
  4. Configures git - Sets up merge drivers for lockfiles

Usage with Claude Code Online

When you start a Claude Code online session, the SessionStart hook will automatically run and set up your environment. You'll see output indicating the setup progress.

Available Commands

Once setup is complete, Claude can use these pixi tasks:

Testing & Quality

  • pixi run test - Run pytest tests
  • pixi run coverage - Run tests with coverage report
  • pixi run lint - Run ruff and pylint linters
  • pixi run format - Format code with ruff

CI Tasks

  • pixi run ci - Run full CI pipeline (format, lint, test, coverage)
  • pixi run fix - Auto-fix common issues (update lock, format, lint)

Prek Hooks

  • pixi run prek - Run all prek hooks
  • pixi run prek-update - Update prek hook versions

How It Works

  1. First session: The hook installs pixi and all dependencies (may take a few minutes)
  2. Subsequent sessions: The hook verifies the environment is ready (much faster)
  3. Environment activated: All pixi commands are available to Claude

Troubleshooting

If you encounter issues:

  1. Dependencies not found: Run pixi install manually
  2. Lockfile conflicts: Run pixi run update-lock
  3. Prek issues: Run pixi run prek-update
  4. Full reset: Delete .pixi directory and restart session

Customization

You can modify hooks/SessionStart to:

  • Install additional tools
  • Run custom setup scripts
  • Configure environment variables
  • Set up external services

Ralph Orchestrator

This project includes ralph-orchestrator, an autonomous agent orchestrator for managing complex multi-step tasks.

Installation

Ralph is installed as a pixi dependency. After running pixi install, it's available in the environment.

Basic Usage

# Initialize ralph (creates ralph.yml and PROMPT.md if not present)
pixi run ralph init

# Plan a task (creates a spec file)
pixi run ralph plan "Implement feature X"

# Run ralph with the current PROMPT.md
pixi run ralph run

# Check ralph version
pixi run ralph --version

Configuration Files

  • ralph.yml - Main configuration (event loop, hats, backends)
  • PROMPT.md - Task description for ralph to execute
  • specs/ - Directory for task specifications

Customization

Edit ralph.yml to customize:

  • event_loop.max_iterations - Maximum iterations before stopping
  • event_loop.max_runtime_seconds - Maximum runtime
  • cli.backend - AI backend (claude, gemini, codex, etc.)
  • hats - Define custom personas for multi-agent workflows

See the ralph-orchestrator documentation for more details.

Learn More