rails-ai

内容来源：README.md（说明文档） · 原始地址 · 查看安装指南

原始内容

# rails-ai 🚂

> **⚠️ EXPERIMENTAL:** Under active development. APIs and architecture may change. Phase 2 of 4.

Rails domain expertise for Claude Code. Built on [Superpowers](https://github.com/obra/superpowers) workflows.

## The Vision

Build Rails apps so autonomously they leave today's Next.js and React generators in the dust. We're not playing catch-up—we're setting the pace and showing what LLM-driven development really looks like when Rails is in the driver's seat.

## The Problem

Right now, LLMs excel at Next.js and Node code. Ask them to build a Rails app and you get... okay results. Ask for a Next.js app and you get production-ready code.

**Why?** Training data. JavaScript/TypeScript dominates the training sets. Rails, despite being more mature and productive, is underrepresented.

**The irony?** Rails is actually better suited for AI code generation:

- **Convention over Configuration** — Less decision-making, more consistent patterns
- **Strong opinions** — Clear right/wrong answers
- **Battle-tested** — 20+ years of best practices
- **Monolithic by default** — Simpler mental model
- **Solid Stack** — Rails 8's batteries-included approach

Rails was built to make developers productive by removing decisions. That same philosophy makes it perfect for LLMs — fewer choices, more consistency, clearer patterns.

## What It Does

Adds Rails-specific skills to Claude Code: ActiveRecord patterns, Hotwire, Minitest/TDD, security, Solid Stack (Rails 8), and 20+ team conventions.

The `/rails-ai:architect` command loads Superpowers workflows (process) and Rails-AI skills (domain knowledge), then coordinates general-purpose agents to build features end-to-end.

## Installation

```bash
# 1. Install Superpowers (required)
claude
/plugin marketplace add obra/superpowers
/plugin install superpowers

# 2. Install rails-ai
/plugin marketplace add zerobearing2/rails-ai
/plugin install rails-ai
```

### Local Development Install

To contribute or test changes locally:

```bash
# 1. Clone the repository
git clone https://github.com/zerobearing2/rails-ai.git
cd rails-ai

# 2. Install Superpowers (required dependency)
claude
/plugin marketplace add obra/superpowers
/plugin install superpowers

# 3. Add local directory as a marketplace and install
/plugin marketplace add /path/to/rails-ai
/plugin install rails-ai@rails-ai

# 4. Restart Claude Code to load the plugin
```

**Verify installation:**

```bash
/plugin
# Opens plugin management interface - verify rails-ai is listed
```

**Run tests before submitting changes:**

```bash
bin/ci
```

## Usage

```bash
/rails-ai:architect add user authentication
/rails-ai:architect debug failing test
/rails-ai:architect refactor UserController
```

**Local Development Note:** If you're developing rails-ai locally (installed from a local directory), SessionStart hooks don't execute due to a [Claude Code limitation](https://github.com/anthropics/claude-code/issues/11939). Manually use the skill once per session:

```text
Use the skill: using-rails-ai
```

This uses the full protocol including Superpowers integration and skill-loading enforcement. Not needed when installed from GitHub.

## Architecture

**Two-layer system:**

- **Superpowers** = HOW to work (brainstorming, planning, TDD, debugging, code review)
- **Rails-AI** = WHAT you're building (12 Rails domain skills + team rules)

The `/rails-ai:architect` slash command loads both layers, then dispatches general-purpose agents as workers to implement features.

## Philosophy

Opinionated Rails development:
- 37signals philosophy (simple, pragmatic)
- Rails conventions (REST-only)
- Solid Stack (SolidQueue, SolidCache, SolidCable)
- Minitest, not RSpec
- TDD always

## Roadmap

We're turning that gap into a roadmap instead of a wish list.

**Phase 1 — Architect Coordinator** ✅ Complete
Clean architecture with `/rails-ai:architect` slash command that coordinates development: loads Superpowers workflows (HOW) and Rails-AI skills (WHAT), then dispatches general-purpose workers to implement. No complexity, clear separation between coordination and execution.

**Phase 2 — Domain Skills & Team Rules** 🚧 Current
12 focused domain skills, 20+ team conventions, and custom RuboCop cops. Every decision backed by Rails best practices in reusable, testable modules.

**Phase 3 — Memory & Context** 🔬 Next
Index thousands of production Rails patterns, wire them into local memory + knowledge graph, and retrieve the right snippet at the right time with RAG + SQLite vector search.

**Phase 4 — Fully Autonomous Rails** 🚀 Goal
Ship end-to-end features—auth, payments, background jobs, admin, APIs—complete with tests and security. Target: 85%+ first-pass success rate.

**What already works:**
- `/rails-ai:architect` coordinates features end-to-end (planning → worker dispatch → review)
- 12 domain skills cover models, controllers, views, Hotwire, security, testing, and more
- Superpowers workflows provide battle-tested TDD, debugging, and code review
- 20+ team rules enforce Rails conventions automatically
- Clean separation: coordinator loads skills, workers implement features

**What we're building next:**
- Phase 3 RAG pipeline with SQLite vector search
- Library of thousands of indexed production Rails patterns
- 30–50% accuracy gains from better retrieval and context
- Path to 85%+ first-try pass rates and autonomous feature delivery

## Join Us

This is bigger than a plugin. We're proving Rails can lead in the AI era.

**If you're a Rails developer:** Try the architect. Break it. Tell us what's missing. Your real-world usage drives what we build.

**If you're into AI/ML:** We need help with Phase 3 (RAG) and Phase 4 (autonomous delivery). Indexing Rails codebases, building evaluation frameworks, improving retrieval quality.

**If you care about Rails:** Star the repo. Spread the word. Rails deserves world-class AI tooling. Help us build it.

The goal isn't just good—it's to make Rails the obvious choice for AI-assisted development. Help us get there.

## Contributing

See [CONTRIBUTING.md](https://github.com/zerobearing2/rails-ai/blob/HEAD/CONTRIBUTING.md) and [TESTING.md](https://github.com/zerobearing2/rails-ai/blob/HEAD/TESTING.md).

## Credits

Built on [Superpowers](https://github.com/obra/superpowers) by [@obra](https://github.com/obra).

Planning approach inspired by [Specification Pyramid](https://signalthinking.com/p/why-your-prd-isnt-working-for-ai-083) by [Robert Evans](https://github.com/revans).

## License

MIT License - see [LICENSE](https://github.com/zerobearing2/rails-ai/tree/HEAD/LICENSE).

rails-ai 🚂

⚠️ EXPERIMENTAL: Under active development. APIs and architecture may change. Phase 2 of 4.

Rails domain expertise for Claude Code. Built on Superpowers workflows.

The Vision

Build Rails apps so autonomously they leave today's Next.js and React generators in the dust. We're not playing catch-up—we're setting the pace and showing what LLM-driven development really looks like when Rails is in the driver's seat.

The Problem

Right now, LLMs excel at Next.js and Node code. Ask them to build a Rails app and you get... okay results. Ask for a Next.js app and you get production-ready code.

Why? Training data. JavaScript/TypeScript dominates the training sets. Rails, despite being more mature and productive, is underrepresented.

The irony? Rails is actually better suited for AI code generation:

Convention over Configuration — Less decision-making, more consistent patterns
Strong opinions — Clear right/wrong answers
Battle-tested — 20+ years of best practices
Monolithic by default — Simpler mental model
Solid Stack — Rails 8's batteries-included approach

Rails was built to make developers productive by removing decisions. That same philosophy makes it perfect for LLMs — fewer choices, more consistency, clearer patterns.

What It Does

Adds Rails-specific skills to Claude Code: ActiveRecord patterns, Hotwire, Minitest/TDD, security, Solid Stack (Rails 8), and 20+ team conventions.

The /rails-ai:architect command loads Superpowers workflows (process) and Rails-AI skills (domain knowledge), then coordinates general-purpose agents to build features end-to-end.

Installation

# 1. Install Superpowers (required)
claude
/plugin marketplace add obra/superpowers
/plugin install superpowers

# 2. Install rails-ai
/plugin marketplace add zerobearing2/rails-ai
/plugin install rails-ai

Local Development Install

To contribute or test changes locally:

# 1. Clone the repository
git clone https://github.com/zerobearing2/rails-ai.git
cd rails-ai

# 2. Install Superpowers (required dependency)
claude
/plugin marketplace add obra/superpowers
/plugin install superpowers

# 3. Add local directory as a marketplace and install
/plugin marketplace add /path/to/rails-ai
/plugin install rails-ai@rails-ai

# 4. Restart Claude Code to load the plugin

Verify installation:

/plugin
# Opens plugin management interface - verify rails-ai is listed

Run tests before submitting changes:

bin/ci

Usage

/rails-ai:architect add user authentication
/rails-ai:architect debug failing test
/rails-ai:architect refactor UserController

Local Development Note: If you're developing rails-ai locally (installed from a local directory), SessionStart hooks don't execute due to a Claude Code limitation. Manually use the skill once per session:

Use the skill: using-rails-ai

This uses the full protocol including Superpowers integration and skill-loading enforcement. Not needed when installed from GitHub.

Architecture

Two-layer system:

Superpowers = HOW to work (brainstorming, planning, TDD, debugging, code review)
Rails-AI = WHAT you're building (12 Rails domain skills + team rules)

The /rails-ai:architect slash command loads both layers, then dispatches general-purpose agents as workers to implement features.

Philosophy

Opinionated Rails development:

37signals philosophy (simple, pragmatic)
Rails conventions (REST-only)
Solid Stack (SolidQueue, SolidCache, SolidCable)
Minitest, not RSpec
TDD always

Roadmap

We're turning that gap into a roadmap instead of a wish list.

Phase 1 — Architect Coordinator ✅ Complete Clean architecture with /rails-ai:architect slash command that coordinates development: loads Superpowers workflows (HOW) and Rails-AI skills (WHAT), then dispatches general-purpose workers to implement. No complexity, clear separation between coordination and execution.

Phase 2 — Domain Skills & Team Rules 🚧 Current 12 focused domain skills, 20+ team conventions, and custom RuboCop cops. Every decision backed by Rails best practices in reusable, testable modules.

Phase 3 — Memory & Context 🔬 Next Index thousands of production Rails patterns, wire them into local memory + knowledge graph, and retrieve the right snippet at the right time with RAG + SQLite vector search.

Phase 4 — Fully Autonomous Rails 🚀 Goal Ship end-to-end features—auth, payments, background jobs, admin, APIs—complete with tests and security. Target: 85%+ first-pass success rate.

What already works:

/rails-ai:architect coordinates features end-to-end (planning → worker dispatch → review)
12 domain skills cover models, controllers, views, Hotwire, security, testing, and more
Superpowers workflows provide battle-tested TDD, debugging, and code review
20+ team rules enforce Rails conventions automatically
Clean separation: coordinator loads skills, workers implement features

What we're building next:

Phase 3 RAG pipeline with SQLite vector search
Library of thousands of indexed production Rails patterns
30–50% accuracy gains from better retrieval and context
Path to 85%+ first-try pass rates and autonomous feature delivery

Join Us

This is bigger than a plugin. We're proving Rails can lead in the AI era.

If you're a Rails developer: Try the architect. Break it. Tell us what's missing. Your real-world usage drives what we build.

If you're into AI/ML: We need help with Phase 3 (RAG) and Phase 4 (autonomous delivery). Indexing Rails codebases, building evaluation frameworks, improving retrieval quality.

If you care about Rails: Star the repo. Spread the word. Rails deserves world-class AI tooling. Help us build it.

The goal isn't just good—it's to make Rails the obvious choice for AI-assisted development. Help us get there.

Contributing

See CONTRIBUTING.md and TESTING.md.

Credits

Built on Superpowers by @obra.

Planning approach inspired by Specification Pyramid by Robert Evans.

License

MIT License - see LICENSE.