RXDev

AI-powered developer workflow tool — commit messages, code review, PR automation, analytics, and more.

Documentation · Report Bug · Русский · 中文 · Deutsch · Français

Table of Contents

• What's New in v4.1
• What's New in v4.0
• Features
• Installation
• Quick Start
• Commands
• Providers & Models
• Configuration
• GitHub Action
• Documentation
• License

What's New in v4.1

• Full Interface Localization — entire interface in your language (en/ru/de/fr/zh)
• Token Management — track usage, set daily/monthly limits, get warnings
• Command Suggestions — "Did you mean 'stats'?" for typos
• Improved File Picker — shows file sizes, better navigation
• New Command — rxdev tokens for token usage statistics

What's New in v4.0

• Renamed from rxcommit to rxdev — broader scope beyond just commits
• AI Code Review (rxdev review) — review staged changes or PRs before committing
• Contextual Commits — uses branch name, recent commits, and GitHub Issues for better messages
• Analytics (rxdev analytics) — commit statistics and bad practice detection
• Smart Split — auto-detects file types (tests, docs, config) for intelligent grouping
• Config as Code — rxdev.yml in your project root for team-shared settings
• GitHub Action — automatic AI review on every PR
• Improved Diff System — smart truncation (16K chars), never cuts files mid-way
• Cross-platform CI — tested on Windows, macOS, and Ubuntu

Features

QuickFlow (rxdev go)

The flagship workflow. Stage, generate, review, push in one command:

$ rxdev go

1. Stage — interactively choose files to include
2. Generate — AI analyses the diff and streams a Conventional Commit message in real-time
3. Review — see the message, choose what to do next
4. Push — accept and push, or edit, regenerate, or cancel

AI Code Review (rxdev review)

Get instant feedback on your staged changes:

$ rxdev review

• Analyzes code for bugs, security issues, and performance problems
• Provides severity ratings (critical/warning/suggestion)
• Works on staged changes or PR diffs

Contextual Commits

Commit messages that understand your workflow:

• Branch context — knows you're on feature/user-auth
• Recent history — considers the last 3 commits
• GitHub Issues — auto-links to issues from branch names (fix/123, feature/PROJ-456)

Smart Split (rxdev split)

Split large changes into logical commits:

$ rxdev split

• Auto-detects test files → test: commits
• Auto-detects docs → docs: commits
• Auto-detects config files → chore: commits
• LLM handles the rest for intelligent grouping

Analytics (rxdev analytics)

Understand your commit patterns:

$ rxdev analytics

• Commit statistics by type
• Average message length
• Breaking change detection
• Bad practice warnings (too long, non-conventional, etc.)

GitHub Action

Automatic AI review on every PR — uses your existing provider key:

name: AI Review
on:
  pull_request:
    types: [opened, synchronize]

jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: rxgodev/rxdev-review@v1
        env:
          GROQ_API_KEY: ${{ secrets.GROQ_API_KEY }}
          # or OPENAI_API_KEY, OPENROUTER_API_KEY — whichever you use

Other Features

• Auto-bump version — discovers 15+ manifest formats (package.json, Cargo.toml, pyproject.toml, etc.)
• Secret scanning — detects leaked credentials in staged changes
• PR generation — creates PR titles and descriptions from commits
• GitHub Releases — creates releases with tags from conventional commits
• History rewriting — interactive git-filter-repo wrapper
• Multi-project — manage hooks across multiple repositories

Installation

Prerequisites

• Node.js >= 18
• An API key from one of the supported providers (see Providers & Models)

1. Install globally

npm:

npm install -g rxdev@latest

pnpm:

pnpm add -g rxdev@latest

yarn:

yarn global add rxdev@latest

2. Configure your API key

rxdev config

Navigate to API key in the menu and paste your key.

Quick Start

After installing, get your first AI-powered commit in seconds:

# Install the hook (one time per repo)
rxdev init

# QuickFlow — stage, generate, review, push
rxdev go

That's it. rxdev go guides you through every step:

1. Choose files to stage — or press Enter to stage all
2. AI generates the message — streamed live to your terminal
3. Review loop — decide what to do:
   • Push — accept and push immediately
   • Edit message — open $EDITOR and amend
   • Regenerate — ask the model for a new suggestion
   • Cancel — soft-reset and abort
4. Push — select remote and branch interactively from lists

Commands

| Command | Description | |---------|-------------| | rxdev go | QuickFlow — stage → scan → generate → review → push | | rxdev review | AI code review of staged changes | | rxdev split | Split staged changes into multiple logical commits | | rxdev scan | Scan staged changes for leaked secrets | | rxdev stats | Show commit statistics and bad practices | | rxdev tokens | Show token usage statistics | | rxdev pr | Generate a pull request title + description | | rxdev release | Create GitHub Release with tag | | rxdev init | Install the AI commit hook | | rxdev config | Configure provider, model, language, and more | | rxdev status | Show integration status | | rxdev doctor | Diagnose your setup | | rxdev filter | Rewrite git history | | rxdev uninstall | Remove the hook | | rxdev version | Show version number | | rxdev update | Update to latest version |

Providers & Models

Groq (default, free tier)

| Model | Speed | Context | Best for | |-------|-------|---------|----------| | Llama 3.3 70B Versatile | ~280 t/s | 128K | High quality, complex changes | | Llama 3.1 8B Instant | ~560 t/s | 128K | Fast, simple changes (default) | | Mixtral 8x7B | ~500 t/s | 32K | Balanced speed/quality |

OpenAI

| Model | Context | Best for | |-------|---------|----------| | GPT-4o | 128K | Best quality | | GPT-4o-mini | 128K | Fast, cost-effective (default) | | GPT-4 Turbo | 128K | Legacy support |

OpenRouter

Access to 100+ models through OpenRouter:

• openai/gpt-4o — GPT-4o via OpenRouter
• anthropic/claude-3.5-sonnet — Claude 3.5
• meta-llama/llama-3.3-70b — Llama 3.3
• Any model available on openrouter.ai

Ollama (local, private)

Run any model locally — your code never leaves your machine:

• llama3.1 — Meta's Llama 3.1
• codellama — Code-specialized Llama
• mistral — Mistral 7B
• Any model supported by Ollama

Custom Endpoint

Use any OpenAI-compatible API:

rxdev config
# Set Provider → custom
# Set API URL → https://your-api.com/v1/chat/completions

Configuration

Interactive Configuration

rxdev config

Settings

| Setting | Description | |---------|-------------| | Provider | Choose LLM backend: Groq, OpenAI, OpenRouter, Ollama, or custom | | Model | Pick a model for the chosen provider | | Language | Choose language for commit body: English, Russian, German, French, Chinese | | Custom prompt | Override system prompt. Use {types} placeholder | | Custom types | Add extra Conventional Commits types | | API key | Set provider API key (or use env vars) | | Co-author | Toggle Co-authored-by trailer | | Auto-bump | Toggle automatic version bumps |

Environment Variables

| Variable | Description | |----------|-------------| | GROQ_API_KEY | Groq API key (default provider) | | OPENAI_API_KEY | OpenAI API key | | OPENROUTER_API_KEY | OpenRouter API key |

Config as Code (rxdev.yml)

Create rxdev.yml in your project root for team-shared settings:

provider: groq
model: llama-3.3-70b-versatile
language: en
maxDiffLength: 16000
coauthor: true
bumpVersion: false
validTypes:
  - feat
  - fix
  - docs
  - refactor
  - perf
  - test
  - chore

Priority: rxdev.yml → CLI flags → ~/.config/rxdev/config.json → defaults

.commitignore

Edit .commitignore (follows .gitignore syntax) to exclude files from the diff sent to the model.

GitHub Action

Automatic AI review on every PR — uses your existing provider key:

name: AI Review
on:
  pull_request:
    types: [opened, synchronize]

jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: rxgodev/rxdev-review@v1
        env:
          GROQ_API_KEY: ${{ secrets.GROQ_API_KEY }}
          # or OPENAI_API_KEY, OPENROUTER_API_KEY — whichever you use

Documentation

| Document | Description | |----------|-------------| | Architecture | High-level overview of the CLI + git-hook architecture | | Auto-bump | Smart version bumping for 15+ manifest formats | | Templates & Projects | Multi-repository management and shared templates |

License

MIT — see LICENSE for details.