User Guide

Git-Iris is an AI-powered Git workflow assistant that uses Iris, an intelligent agent built on the Rig framework, to help you write commit messages, review code, generate PR descriptions, and maintain changelogs.

Core Capabilities

Feature	Command	What It Does
Commit Messages	git-iris gen	Generates conventional commit messages with optional emoji
Code Reviews	git-iris review	Performs multi-dimensional code analysis across 10+ dimensions
PR Descriptions	git-iris pr	Creates comprehensive pull request descriptions
Changelogs	git-iris changelog	Generates Keep a Changelog-formatted entries
Release Notes	git-iris release-notes	Creates release documentation
Studio TUI	git-iris studio	Unified terminal interface for all operations

Quick Start

# Generate a commit message for staged changes
git add .
git-iris gen

# Auto-commit with generated message
git-iris gen --auto-commit

# Review staged changes
git-iris review

# Generate PR description comparing branches
git-iris pr --from main --to feature-branch

# Create changelog between versions
git-iris changelog --from v1.0.0 --to v2.0.0

Global Flags

Available across all commands:

Flag	Description
--provider <name>	Override LLM provider (openai, anthropic, google)
-i, --instructions "text"	Add custom instructions for this operation
--preset <name>	Use instruction preset (see Presets)
--debug	Enable color-coded agent execution visualization
--quiet	Suppress non-essential output
--theme <name>	Override theme for this session

Configuration

Set your API key and preferences:

# Configure provider
git-iris config --provider anthropic --api-key YOUR_KEY

# Set default model
git-iris config --model claude-sonnet-4-5-20250929

# List available presets
git-iris list-presets

# List available themes
git-iris themes

Next Steps

• Commit Messages - Generate intelligent commit messages
• Code Reviews - AI-powered code analysis
• Pull Requests - Professional PR descriptions
• Changelogs - Structured changelog generation
• Release Notes - Release documentation
• Presets - Customize Iris's behavior

How Iris Works

Iris is an LLM-driven agent that dynamically explores your codebase using tool calls. Instead of dumping all context upfront, she:

1. Analyzes your request and repository state
2. Calls tools to gather precise information (git diff, file content, commit history)
3. Adapts strategy based on changeset size (full context, relevance scoring, or parallel analysis)
4. Generates structured output tailored to your workflow

This tool-based approach ensures Iris sees exactly what she needs without overwhelming her context window.