Ralph

Quick Start - Ralph Multi-Agent

Get Ralph running in 5 minutes with any AI agent.

Choose Your Agent

If you want…	Use	Setup
Free + huge context	gemini	npm install -g @google/gemini-cli
Best code quality	openai	Set OPENAI_API_KEY
Privacy / offline	ollama	winget install Ollama.Ollama
GUI for local	lmstudio	Download from lmstudio.ai
Enterprise	azure	Configure endpoint

Quick Setup

Option A: Gemini (Default, Free)

# 1. Install
npm install -g @google/gemini-cli
gemini auth login

# 2. Run
.\ralph.bat

Option B: OpenAI

# 1. Set API key
$env:OPENAI_API_KEY = "sk-..."

# 2. Run
.\ralph.bat openai

Option C: Ollama (Local)

# 1. Install
winget install Ollama.Ollama

# 2. Pull a model
ollama pull codellama:13b

# 3. Run
.\ralph.bat ollama

Option D: Network Model

# Point to any OpenAI-compatible API
.\ralph.bat network -Endpoint http://192.168.1.100:8080/v1/chat/completions

Define Your Task

Edit RALPH_TASK.md:

---
task: Build REST API
test_command: npm test
---

# Task: REST API

## Success Criteria
1. [ ] GET /health returns 200
2. [ ] POST /users works
3. [ ] Tests pass

Run & Monitor

# Run
.\ralph.bat

# Watch progress
.\ralph.bat watch

When Things Go Wrong

# Check errors
Get-Content .ralph\errors.log

# Add guardrail to prevent repeat
notepad .ralph\guardrails.md

Add this format:

### Sign: [What went wrong]
- **Trigger**: [When it happens]
- **Instruction**: [What to do instead]

Commands Cheat Sheet

Command	What
ralph.bat	Run default
ralph.bat openai	Use OpenAI
ralph.bat ollama -Model codellama:34b	Use specific model
ralph.bat watch	Monitor logs
ralph.bat models ollama	List models
ralph.bat init	Reset state

This site is open source. Improve this page.