claude_swarm 1.0.0 → 1.0.2

data/docs/v2/guides/getting-started.md

@@ -0,0 +1,1456 @@
+# Getting Started with SwarmSDK
+
+## What You'll Learn
+
+- What SwarmSDK is and why it's valuable for complex AI tasks
+- How to install and configure SwarmSDK in your Ruby project
+- The fundamental concepts: agents, delegation, tools, and swarms
+- How to create your first agent using YAML, Ruby DSL, and Markdown
+- Essential workflows for building single and multi-agent systems
+- How to interpret results and debug common issues
+
+## Prerequisites
+
+- **Ruby 3.2.0 or higher** installed on your system
+- **Basic Ruby knowledge** - familiarity with hashes, symbols, and blocks
+- **LLM API access** - an API key for OpenAI, Anthropic, or another provider
+- **15-20 minutes** to complete this guide
+
+## What is SwarmSDK?
+
+SwarmSDK is a Ruby framework for orchestrating **teams of AI agents** that collaborate to solve complex problems. Unlike traditional single-agent systems, SwarmSDK enables you to build specialized agents that work together, each focusing on what they do best.
+
+### Why Multi-Agent Systems?
+
+Complex tasks often require different types of expertise. Consider a software development project:
+
+- A **planner** agent breaks down requirements into tasks
+- A **coder** agent implements features
+- A **reviewer** agent checks code quality and suggests improvements
+- A **tester** agent validates functionality
+
+With SwarmSDK, each agent specializes in one area and delegates work to others when needed. This mirrors how human teams collaborate, leading to better results than a single generalist AI trying to do everything.
+
+### Core Benefits
+
+**Agent Specialization**: Each agent has its own system prompt, model, tools, and working directory, optimized for specific tasks.
+
+**Flexible Delegation**: Agents can delegate work to specialists through simple function calls, creating dynamic collaboration patterns.
+
+**Comprehensive Tooling**: Built-in tools for file operations (Read, Write, Edit), code search (Grep, Glob), command execution (Bash), and shared memory (Scratchpad).
+
+**Fine-Grained Control**: Permissions system restricts file access, hooks customize behavior at every step, and parameters tune each agent's LLM settings.
+
+**Multiple Interfaces**: Define swarms using declarative YAML, fluent Ruby DSL, Markdown agent files, or direct API calls—whichever fits your workflow.
+
+## Installation
+
+Add SwarmSDK to your Ruby project's Gemfile:
+
+```ruby
+gem 'swarm_sdk'
+```
+
+Then install:
+
+```bash
+bundle install
+```
+
+Or install directly:
+
+```bash
+gem install swarm_sdk
+```
+
+### Configure Your API Key
+
+SwarmSDK requires an LLM provider to run. Set your API key as an environment variable:
+
+```bash
+# For OpenAI
+export OPENAI_API_KEY="sk-your-key-here"
+
+# For Anthropic Claude
+export ANTHROPIC_API_KEY="sk-ant-your-key-here"
+```
+
+**Tip**: Create a `.env` file in your project root for convenience:
+
+```bash
+# .env
+OPENAI_API_KEY=sk-your-key-here
+```
+
+Load it in your Ruby script:
+
+```ruby
+require 'dotenv/load'
+require 'swarm_sdk'
+```
+
+### Verify Installation
+
+Quick test to ensure everything works:
+
+```bash
+ruby -e "require 'swarm_sdk'; puts 'SwarmSDK loaded successfully!'"
+```
+
+## Core Concepts
+
+Before writing code, let's understand SwarmSDK's architecture.
+
+### Swarms
+
+A **swarm** is a team of agents working together. Every swarm has:
+- A **name** (for logging and debugging)
+- A **lead agent** (the entry point that receives your tasks)
+- One or more **agents** (the team members)
+
+When you call `swarm.execute("task")`, the task goes to the lead agent, which can either handle it directly or delegate to other agents.
+
+### Agents
+
+An **agent** is an AI with specific capabilities and constraints:
+- **Description**: What the agent does (required)
+- **Model**: Which LLM to use (e.g., "gpt-4", "claude-sonnet-4")
+- **System Prompt**: Instructions defining the agent's behavior and expertise
+- **Tools**: Functions the agent can call (Read, Write, Bash, etc.)
+- **Directory**: The agent's working directory for file operations
+- **Delegates To**: Other agents this agent can hand off work to
+
+### Delegation
+
+Delegation is how agents collaborate. When an agent is configured with `delegates_to: [other_agent]`, it gains a special delegation tool.
+
+**Delegation tool naming**: Tools are named `DelegateTaskTo{AgentName}` where AgentName is capitalized. For example:
+- `delegates_to: [reviewer]` → creates `DelegateTaskToReviewer` tool
+- `delegates_to: [backend]` → creates `DelegateTaskToBackend` tool
+- `delegates_to: [qa_tester]` → creates `DelegateTaskToQa_tester` tool
+
+**Example flow**:
+1. You send "Build a login page" to the **lead** agent
+2. Lead delegates to **frontend** agent: "Create the React component"
+3. Frontend delegates to **reviewer** agent: "Check this code"
+4. Reviewer returns feedback to frontend
+5. Frontend returns completed component to lead
+6. Lead returns final result to you
+
+Delegation creates flexible collaboration patterns without rigid workflows.
+
+### Tools
+
+Tools are functions agents call to interact with the world. SwarmSDK provides:
+
+| Category | Tools | Purpose |
+|----------|-------|---------|
+| **File I/O** | Read, Write, Edit, MultiEdit | Read and modify files |
+| **Search** | Grep, Glob | Find content and files |
+| **Execution** | Bash | Run shell commands |
+| **Web** | WebFetch | Fetch and process web content |
+| **Task Management** | TodoWrite | Track progress on multi-step tasks |
+| **Shared Scratchpad** | ScratchpadWrite, ScratchpadRead, ScratchpadList | Share work-in-progress between agents (volatile) |
+| **Per-Agent Memory** | MemoryWrite, MemoryRead, MemoryEdit, MemoryMultiEdit, MemoryGlob, MemoryGrep, MemoryDelete | Persistent learning and knowledge storage (opt-in) |
+| **Reasoning** | Think | Extended reasoning for complex problems |
+
+**Default tools**: Every agent automatically gets Read, Grep, Glob, TodoWrite, Think, WebFetch, and scratchpad tools (ScratchpadWrite, ScratchpadRead, ScratchpadList) unless you explicitly disable them with `disable_default_tools: true`.
+
+**Memory tools** are opt-in for learning agents that need persistent knowledge storage.
+
+## Configuration Formats: YAML vs Ruby DSL vs Markdown
+
+SwarmSDK supports three equally powerful configuration formats.
+
+### YAML Configuration
+
+**Use YAML when**:
+- You want simple, declarative configuration files
+- Your hooks are shell scripts (not Ruby code)
+- You prefer separating configuration from code
+- You're new to SwarmSDK
+
+**Example**:
+```yaml
+version: 2
+swarm:
+  name: "My First Swarm"
+  lead: assistant
+  agents:
+    assistant:
+      description: "A helpful assistant"
+      model: "gpt-4"
+      system_prompt: "You are a helpful assistant."
+      tools:
+        - Write
+```
+
+### Ruby DSL
+
+**Use Ruby DSL when**:
+- You need dynamic configuration (variables, conditionals, loops)
+- You want to write hooks as Ruby blocks (inline logic)
+- You prefer IDE autocomplete and type checking
+- You're building reusable agent libraries
+
+**Example - Inline Definition**:
+```ruby
+swarm = SwarmSDK.build do
+  name "My First Swarm"
+  lead :assistant
+
+  agent :assistant do
+    description "A helpful assistant"
+    model "gpt-4"
+    system_prompt "You are a helpful assistant."
+    tools :Write
+  end
+end
+```
+
+**Example - Load from Markdown File**:
+```ruby
+swarm = SwarmSDK.build do
+  name "My First Swarm"
+  lead :assistant
+
+  # Load agent from markdown file content
+  # Name is always required (overrides any name in frontmatter)
+  agent :assistant, File.read("agents/assistant.md")
+end
+```
+
+**Example - Mix Inline and Markdown**:
+```ruby
+swarm = SwarmSDK.build do
+  name "Dev Team"
+  lead :frontend
+
+  # Inline DSL agent
+  agent :frontend do
+    description "Frontend developer"
+    model "gpt-4"
+    system_prompt "You build UIs"
+  end
+
+  # Load backend from markdown file
+  agent :backend, File.read("agents/backend.md")
+end
+```
+
+**Note**: Use `system_prompt` in the Ruby DSL to set the agent's instructions.
+
+### Markdown Agent Files
+
+**Use Markdown when**:
+- You want to define agents in separate, readable files
+- You prefer writing prompts in a documentation format
+- You want to version control agent definitions independently
+- You're sharing agent configurations with non-developers
+
+**Example**: Create `agents/assistant.md`:
+
+```markdown
+---
+description: "A helpful assistant"
+model: "gpt-4"
+tools:
+  - Write
+  - Edit
+delegates_to:
+  - reviewer
+---
+
+# Assistant Agent
+
+You are a helpful assistant who writes clear, maintainable code.
+
+## Your Responsibilities
+
+- Write clean, well-documented code
+- Follow best practices and style guides
+- Delegate code reviews to the reviewer agent
+
+## Best Practices
+
+- Always add comments for complex logic
+- Use descriptive variable names
+- Test your code before finalizing
+```
+
+Then reference it in your YAML configuration:
+
+```yaml
+# swarm.yml
+version: 2
+swarm:
+  name: "My Swarm"
+  lead: assistant
+  agents:
+    assistant: "agents/assistant.md"
+```
+
+**Note**: When the agent value is a string, it's treated as a file path. You can also use the hash format if you need to override settings from the markdown file:
+
+```yaml
+agents:
+  assistant:
+    agent_file: "agents/assistant.md"
+    timeout: 120  # Override timeout from file
+```
+
+```ruby
+require 'swarm_sdk'
+
+# Load swarm from YAML (which references the Markdown file)
+swarm = SwarmSDK::Swarm.load('swarm.yml')
+result = swarm.execute("Your task here")
+```
+
+**Note**: Markdown agent definitions can be used in multiple ways:
+- **YAML**: Reference file path: `assistant: "agents/assistant.md"`
+- **Ruby DSL**: Pass file content: `agent :assistant, File.read("agents/assistant.md")`
+- **Ruby DSL**: Or inline markdown: `agent :assistant, <<~MD ... MD`
+
+In the Ruby DSL, the agent name is **always required** and will override any `name` field in the markdown frontmatter.
+
+**All three approaches are powerful**—choose what fits your workflow:
+- **YAML**: Best for simple, declarative configs with file references
+- **Ruby DSL**: Best for dynamic configs, inline hooks, and programmatic agent definitions
+- **Markdown**: Best for reusable agent definitions that can be version controlled independently
+
+The rest of this guide shows examples in all three formats.
+
+## Your First Agent
+
+Let's create the simplest possible agent: one that answers questions.
+
+### Understanding What We're Building
+
+We'll create:
+1. A swarm named "My First Swarm"
+2. One agent called "assistant"
+3. The agent will use GPT-4 and have basic file reading capability
+
+### YAML Approach
+
+Create a file called `swarm.yml`:
+
+```yaml
+version: 2
+swarm:
+  name: "My First Swarm"
+  lead: assistant
+
+  agents:
+    assistant:
+      description: "A helpful assistant"
+      model: "gpt-4"
+      system_prompt: |
+        You are a helpful assistant.
+        Answer questions clearly and concisely.
+      tools:
+        - Write
+```
+
+Create a file called `run.rb`:
+
+```ruby
+require 'swarm_sdk'
+
+# Load swarm from YAML
+swarm = SwarmSDK::Swarm.load('swarm.yml')
+
+# Execute a task
+result = swarm.execute("What is 2 + 2?")
+
+puts "Response: #{result.content}"
+puts "Success: #{result.success?}"
+puts "Duration: #{result.duration}s"
+```
+
+### Ruby DSL Approach
+
+Create a file called `run.rb`:
+
+```ruby
+require 'swarm_sdk'
+
+# Build swarm with DSL
+swarm = SwarmSDK.build do
+  name "My First Swarm"
+  lead :assistant
+
+  agent :assistant do
+    description "A helpful assistant"
+    model "gpt-4"
+    system_prompt "You are a helpful assistant. Answer questions clearly and concisely."
+    tools :Write
+  end
+end
+
+# Execute a task
+result = swarm.execute("What is 2 + 2?")
+
+puts "Response: #{result.content}"
+puts "Success: #{result.success?}"
+puts "Duration: #{result.duration}s"
+```
+
+### Markdown Approach
+
+Create `agents/assistant.md`:
+
+```markdown
+---
+description: "A helpful assistant"
+model: "gpt-4"
+tools:
+  - Write
+---
+
+# Helpful Assistant
+
+You are a helpful assistant.
+
+Answer questions clearly and concisely.
+```
+
+Create `swarm.yml`:
+
+```yaml
+version: 2
+swarm:
+  name: "My First Swarm"
+  lead: assistant
+  agents:
+    assistant: "agents/assistant.md"
+```
+
+Create `run.rb`:
+
+```ruby
+require 'swarm_sdk'
+
+# Load swarm from YAML (which references the Markdown file)
+swarm = SwarmSDK::Swarm.load('swarm.yml')
+
+# Execute a task
+result = swarm.execute("What is 2 + 2?")
+
+puts "Response: #{result.content}"
+puts "Success: #{result.success?}"
+puts "Duration: #{result.duration}s"
+```
+
+### Run Your Agent
+
+```bash
+ruby run.rb
+```
+
+**Expected output**:
+```
+Response: The answer is 4.
+Success: true
+Duration: 1.23s
+```
+
+### What's Happening?
+
+Let's break down each part:
+
+1. **`version: 2`** (YAML only): Specifies SwarmSDK configuration version
+2. **`name "My First Swarm"`**: Names your swarm (useful for logging)
+3. **`lead :assistant`**: Designates which agent receives your tasks
+4. **`agent :assistant do`**: Defines an agent with a unique name
+5. **`description`**: Explains the agent's role (required field)
+6. **`model "gpt-4"`**: Specifies which LLM to use
+7. **`system_prompt`**: Instructions that guide the agent's behavior
+8. **`tools :Write`**: Grants the agent Write capability (Read is already included by default)
+9. **`swarm.execute("...")`**: Sends a task to the lead agent and returns the result
+
+## Understanding Results
+
+The `execute` method returns a `Result` object with detailed information:
+
+```ruby
+result = swarm.execute("What is 2 + 2?")
+
+# Access response content
+result.content      # => "The answer is 4."
+
+# Check success status
+result.success?     # => true
+
+# Get performance metrics
+result.duration     # => 1.23 (seconds)
+result.total_cost   # => 0.0015 (USD)
+result.total_tokens # => 450
+
+# Get involved agents (useful in multi-agent swarms)
+result.agents_involved  # => [:assistant]
+
+# Check for errors
+result.error        # => nil (or exception object if failed)
+```
+
+### Result Object Reference
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `content` | String | The final response from the lead agent |
+| `success?` | Boolean | Whether execution completed without errors |
+| `duration` | Float | Total execution time in seconds |
+| `total_cost` | Float | Total cost in USD (if provider reports it) |
+| `total_tokens` | Integer | Total tokens used across all agents |
+| `agents_involved` | Array[Symbol] | List of agents that participated |
+| `error` | Exception \| nil | Error object if execution failed |
+
+### Common Result Patterns
+
+```ruby
+# Check if successful before using result
+if result.success?
+  puts result.content
+else
+  puts "Error: #{result.error.message}"
+end
+
+# Log cost and performance
+puts "Cost: $#{result.total_cost.round(4)}"
+puts "Time: #{result.duration.round(2)}s"
+puts "Tokens: #{result.total_tokens}"
+
+# Check which agents were involved
+puts "Agents used: #{result.agents_involved.join(', ')}"
+
+# Handle errors gracefully
+begin
+  result = swarm.execute("Complex task")
+
+  # Check result success instead of catching exceptions
+  unless result.success?
+    puts "Execution failed: #{result.error.message}"
+  end
+rescue SwarmSDK::ConfigurationError => e
+  puts "Configuration error: #{e.message}"
+end
+```
+
+## Common Workflows
+
+Now that you understand the basics, let's explore common patterns.
+
+### Single Agent with File Operations
+
+**Use case**: An agent that reads a file and summarizes it.
+
+**YAML** (`file-reader.yml`):
+```yaml
+version: 2
+swarm:
+  name: "File Reader"
+  lead: reader
+
+  agents:
+    reader:
+      description: "Reads and summarizes files"
+      model: "gpt-4"
+      system_prompt: |
+        You are a file analysis expert.
+        Read files and provide concise, insightful summaries.
+      # Read is included by default - no tools needed!
+      directory: "."
+```
+
+**Ruby DSL**:
+```ruby
+swarm = SwarmSDK.build do
+  name "File Reader"
+  lead :reader
+
+  agent :reader do
+    description "Reads and summarizes files"
+    model "gpt-4"
+    system_prompt "You are a file analysis expert. Read files and provide concise, insightful summaries."
+    directory "."
+  end
+end
+
+# Use it
+result = swarm.execute("Read README.md and summarize its key points")
+```
+
+**Markdown** (`agents/reader.md`):
+```markdown
+---
+description: "Reads and summarizes files"
+model: "gpt-4"
+directory: "."
+---
+
+# File Analysis Expert
+
+You are a file analysis expert.
+
+Read files and provide concise, insightful summaries.
+```
+
+**Why this works**: The agent has Read tool by default and can access files in the specified directory.
+
+### Multi-Agent Collaboration
+
+**Use case**: A code review system with a coder and reviewer.
+
+**YAML** (`code-review.yml`):
+```yaml
+version: 2
+swarm:
+  name: "Code Review Team"
+  lead: coder
+
+  agents:
+    coder:
+      description: "Writes clean, maintainable code"
+      model: "gpt-4"
+      system_prompt: |
+        You are an expert programmer.
+        Write code and delegate to the reviewer for feedback.
+      tools:
+        - Write
+        - Edit
+      delegates_to:
+        - reviewer
+
+    reviewer:
+      description: "Reviews code for quality and issues"
+      model: "claude-sonnet-4"
+      system_prompt: |
+        You are a code review expert.
+        Analyze code for bugs, style issues, and improvements.
+      # Only needs Read (included by default)
+```
+
+**Ruby DSL**:
+```ruby
+swarm = SwarmSDK.build do
+  name "Code Review Team"
+  lead :coder
+
+  agent :coder do
+    description "Writes clean, maintainable code"
+    model "gpt-4"
+    system_prompt "You are an expert programmer. Write code and delegate to the reviewer for feedback."
+    tools :Write, :Edit
+    delegates_to :reviewer
+  end
+
+  agent :reviewer do
+    description "Reviews code for quality and issues"
+    model "claude-sonnet-4"
+    system_prompt "You are a code review expert. Analyze code for bugs, style issues, and improvements."
+  end
+end
+
+# Use it
+result = swarm.execute("Write a function to validate email addresses and get it reviewed")
+```
+
+**Markdown** (`agents/coder.md` + `agents/reviewer.md` + `swarm.yml`):
+
+`agents/coder.md`:
+```markdown
+---
+description: "Writes clean, maintainable code"
+model: "gpt-4"
+tools:
+  - Write
+  - Edit
+delegates_to:
+  - reviewer
+---
+
+# Expert Programmer
+
+You are an expert programmer.
+
+Write code and delegate to the reviewer for feedback.
+```
+
+`agents/reviewer.md`:
+```markdown
+---
+description: "Reviews code for quality and issues"
+model: "claude-sonnet-4"
+---
+
+# Code Review Expert
+
+You are a code review expert.
+
+Analyze code for bugs, style issues, and improvements.
+```
+
+`swarm.yml`:
+```yaml
+version: 2
+swarm:
+  name: "Code Review Team"
+  lead: coder
+  agents:
+    coder: "agents/coder.md"
+    reviewer: "agents/reviewer.md"
+```
+
+Then load and use:
+```ruby
+swarm = SwarmSDK::Swarm.load('swarm.yml')
+result = swarm.execute("Write a function to validate email addresses and get it reviewed")
+```
+
+**How delegation works**: The coder writes code, then calls the `DelegateTaskToReviewer` tool to get feedback. The reviewer analyzes and returns suggestions. The coder can iterate based on feedback.
+
+### Using Multiple Tools
+
+**Use case**: An agent that searches, reads, and modifies files.
+
+**YAML** (`refactor-assistant.yml`):
+```yaml
+version: 2
+swarm:
+  name: "Refactor Assistant"
+  lead: refactorer
+
+  agents:
+    refactorer:
+      description: "Finds and refactors code patterns"
+      model: "gpt-4"
+      system_prompt: |
+        You are a refactoring expert.
+        Use Grep to find code patterns, Read to understand context,
+        and Edit to make improvements.
+      tools:
+        - Edit
+        - Bash
+      # Grep and Read included by default
+      directory: "./src"
+```
+
+**Ruby DSL**:
+```ruby
+swarm = SwarmSDK.build do
+  name "Refactor Assistant"
+  lead :refactorer
+
+  agent :refactorer do
+    description "Finds and refactors code patterns"
+    model "gpt-4"
+    system_prompt "You are a refactoring expert. Use Grep to find code patterns, Read to understand context, and Edit to make improvements."
+    tools :Edit, :Bash
+    directory "./src"
+  end
+end
+
+# Use it
+result = swarm.execute("Find all functions longer than 50 lines and suggest refactorings")
+```
+
+**Tool workflow**:
+1. Agent uses Grep to find long functions: `Grep(pattern: "def .*", path: ".")`
+2. Agent uses Read to examine each file: `Read(file_path: "file.rb")`
+3. Agent uses Edit to refactor: `Edit(file_path: "file.rb", old_string: "...", new_string: "...")`
+
+### Customizing Agent Behavior
+
+**Use case**: Different agents need different LLM parameters.
+
+**YAML** (`customized-agents.yml`):
+```yaml
+version: 2
+swarm:
+  name: "Customized Team"
+  lead: creative
+
+  agents:
+    creative:
+      description: "Creative brainstormer"
+      model: "gpt-4"
+      system_prompt: "Generate creative ideas."
+      parameters:
+        temperature: 1.5  # More creative/random
+        top_p: 0.95
+
+    analytical:
+      description: "Analytical reviewer"
+      model: "claude-sonnet-4"
+      system_prompt: "Analyze ideas critically."
+      parameters:
+        temperature: 0.3  # More focused/deterministic
+```
+
+**Ruby DSL**:
+```ruby
+swarm = SwarmSDK.build do
+  name "Customized Team"
+  lead :creative
+
+  agent :creative do
+    description "Creative brainstormer"
+    model "gpt-4"
+    system_prompt "Generate creative ideas."
+    parameters temperature: 1.5, top_p: 0.95
+  end
+
+  agent :analytical do
+    description "Analytical reviewer"
+    model "claude-sonnet-4"
+    system_prompt "Analyze ideas critically."
+    parameters temperature: 0.3
+  end
+end
+```
+
+**Why customize parameters**:
+- **Temperature**: Controls randomness (0.0-2.0). Higher = more creative, lower = more focused.
+- **Top P**: Controls diversity (0.0-1.0). Higher = more diverse outputs.
+- **Max tokens**: Limits response length.
+
+### Using Scratchpad for Agent Communication
+
+**Use case**: Agents sharing data without explicit delegation.
+
+**YAML** (`research-team.yml`):
+```yaml
+version: 2
+swarm:
+  name: "Research Team"
+  lead: coordinator
+
+  agents:
+    coordinator:
+      description: "Coordinates research tasks"
+      model: "gpt-4"
+      system_prompt: |
+        Coordinate research. Use ScratchpadWrite to share findings.
+      delegates_to:
+        - researcher
+        - analyst
+
+    researcher:
+      description: "Gathers raw data"
+      model: "gpt-4"
+      system_prompt: |
+        Research topics and write findings to scratchpad using file path "research/data".
+
+    analyst:
+      description: "Analyzes research data"
+      model: "claude-sonnet-4"
+      system_prompt: |
+        Read "research/data" from scratchpad and provide analysis.
+```
+
+**Ruby DSL**:
+```ruby
+swarm = SwarmSDK.build do
+  name "Research Team"
+  lead :coordinator
+
+  agent :coordinator do
+    description "Coordinates research tasks"
+    model "gpt-4"
+    system_prompt "Coordinate research. Use ScratchpadWrite to share findings."
+    delegates_to :researcher, :analyst
+  end
+
+  agent :researcher do
+    description "Gathers raw data"
+    model "gpt-4"
+    system_prompt 'Research topics and write findings to scratchpad using file path "research/data".'
+  end
+
+  agent :analyst do
+    description "Analyzes research data"
+    model "claude-sonnet-4"
+    system_prompt 'Read "research/data" from scratchpad and provide analysis.'
+  end
+end
+
+# Use it
+result = swarm.execute("Research Ruby performance optimization techniques and analyze them")
+```
+
+**How it works**:
+1. Coordinator delegates to researcher: "Gather data on Ruby performance"
+2. Researcher uses `ScratchpadWrite(file_path: "research/data", content: "...", title: "Research Data")`
+3. Coordinator delegates to analyst: "Analyze the research"
+4. Analyst uses `ScratchpadRead(file_path: "research/data")` to get the data
+5. Analyst provides analysis
+
+## Common Pitfalls and Solutions
+
+### Pitfall 1: Missing Lead Agent
+
+**Error**:
+```ruby
+# ❌ No lead agent specified
+swarm = SwarmSDK.build do
+  name "My Swarm"
+  agent :assistant do
+    description "Helper"
+    model "gpt-4"
+  end
+end
+# => ConfigurationError: No lead agent set
+```
+
+**Solution**:
+```ruby
+# ✅ Always specify lead
+swarm = SwarmSDK.build do
+  name "My Swarm"
+  lead :assistant  # Don't forget this!
+  agent :assistant do
+    description "Helper"
+    model "gpt-4"
+  end
+end
+```
+
+### Pitfall 2: Missing Required Description
+
+**Error**:
+```ruby
+# ❌ Description required
+agent :assistant do
+  model "gpt-4"
+  system_prompt "You are helpful."
+end
+# => ConfigurationError: missing required 'description'
+```
+
+**Solution**:
+```ruby
+# ✅ Always include description
+agent :assistant do
+  description "A helpful assistant"  # Required!
+  model "gpt-4"
+  system_prompt "You are helpful."
+end
+```
+
+### Pitfall 3: Invalid Model Names
+
+**Error**:
+```ruby
+# ❌ Model doesn't exist
+agent :assistant do
+  description "Helper"
+  model "gpt-99"  # Typo or non-existent model
+end
+# => May fail at runtime when trying to call LLM
+```
+
+**Solution**:
+```ruby
+# ✅ Use valid model identifiers
+agent :assistant do
+  description "Helper"
+  model "gpt-4"  # Valid: gpt-4, gpt-4-turbo, claude-sonnet-4, etc.
+end
+```
+
+**Valid model examples**:
+- OpenAI: `"gpt-4"`, `"gpt-4-turbo"`, `"gpt-3.5-turbo"`
+- Anthropic: `"claude-sonnet-4"`, `"claude-opus-4"`, `"claude-haiku-4"`
+- Others: Depends on your provider configuration
+
+### Pitfall 4: Delegation Without Configuration
+
+**Error**:
+```ruby
+# ❌ Trying to delegate to unconfigured agent
+swarm = SwarmSDK.build do
+  name "Team"
+  lead :leader
+
+  agent :leader do
+    description "Leader"
+    model "gpt-4"
+    # Missing: delegates_to :helper
+  end
+
+  agent :helper do
+    description "Helper"
+    model "gpt-4"
+  end
+end
+# Leader won't have DelegateTaskToHelper tool!
+```
+
+**Solution**:
+```ruby
+# ✅ Configure delegation explicitly
+agent :leader do
+  description "Leader"
+  model "gpt-4"
+  delegates_to :helper  # Now leader can delegate!
+end
+```
+
+### Pitfall 5: Tool Not Available
+
+**Error**:
+```ruby
+# Agent tries to use Write but doesn't have it
+agent :reader do
+  description "Reader"
+  model "gpt-4"
+  disable_default_tools true  # Disables all default tools
+  tools :Read  # Only Read available
+end
+# If agent tries to write: ToolNotFoundError
+```
+
+**Solution**:
+```ruby
+# ✅ Add tools you need
+agent :reader do
+  description "Reader"
+  model "gpt-4"
+  tools :Read, :Write  # Explicitly add Write
+end
+```
+
+### Pitfall 6: Incorrect Gem Name
+
+**Error**:
+```ruby
+# ❌ Wrong gem name
+gem 'swarm-sdk'  # Wrong!
+gem 'swarmcore'  # Wrong!
+
+# ❌ Wrong require
+require 'swarm-sdk'  # Wrong!
+```
+
+**Solution**:
+```ruby
+# ✅ Correct gem name and require
+# In Gemfile:
+gem 'swarm_sdk'
+
+# In Ruby code:
+require 'swarm_sdk'
+
+# Installation:
+# gem install swarm_sdk
+```
+
+### Pitfall 7: Wrong Ruby Version
+
+**Error**:
+```bash
+# Using Ruby 3.1 or earlier
+ruby -v
+# => ruby 3.1.0
+
+gem install swarm_sdk
+# => ERROR: swarm_sdk requires Ruby >= 3.2.0
+```
+
+**Solution**:
+```bash
+# ✅ Use Ruby 3.2.0 or higher
+# Check version
+ruby -v
+# => ruby 3.2.0 (or higher)
+
+# Install/upgrade Ruby if needed:
+rbenv install 3.2.0
+rbenv global 3.2.0
+# or
+rvm install 3.2.0
+rvm use 3.2.0
+```
+
+## Testing Your Setup
+
+Here's a comprehensive test script to verify SwarmSDK is working:
+
+```ruby
+require 'swarm_sdk'
+
+puts "Testing SwarmSDK installation..."
+puts "Ruby version: #{RUBY_VERSION}"
+puts ""
+
+begin
+  # Test 1: Basic swarm creation
+  puts "Test 1: Creating swarm..."
+  swarm = SwarmSDK.build do
+    name "Setup Test"
+    lead :test_agent
+
+    agent :test_agent do
+      description "Test agent"
+      model "gpt-4"
+      system_prompt "Reply with exactly: 'SwarmSDK is working!'"
+    end
+  end
+  puts "✅ Swarm created successfully"
+  puts ""
+
+  # Test 2: Simple execution
+  puts "Test 2: Executing task..."
+  result = swarm.execute("Test")
+
+  if result.success?
+    puts "✅ Execution successful"
+    puts "   Response: #{result.content}"
+    puts "   Duration: #{result.duration.round(2)}s"
+    puts "   Cost: $#{result.total_cost.round(4)}"
+    puts "   Agents: #{result.agents_involved.inspect}"
+  else
+    puts "❌ Execution failed: #{result.error.message}"
+    puts "   Check your API key and network connection"
+  end
+  puts ""
+
+  # Test 3: Multi-agent delegation
+  puts "Test 3: Testing delegation..."
+  swarm2 = SwarmSDK.build do
+    name "Delegation Test"
+    lead :primary
+
+    agent :primary do
+      description "Primary agent"
+      model "gpt-4"
+      system_prompt "Delegate to helper with message: 'Hello!'"
+      delegates_to :helper
+    end
+
+    agent :helper do
+      description "Helper agent"
+      model "gpt-4"
+      system_prompt "Reply with: 'Delegation works!'"
+    end
+  end
+
+  result2 = swarm2.execute("Test delegation")
+  if result2.success? && result2.agents_involved.length > 1
+    puts "✅ Delegation works"
+    puts "   Agents involved: #{result2.agents_involved.inspect}"
+  else
+    puts "❌ Delegation failed"
+  end
+
+rescue StandardError => e
+  puts "❌ Error: #{e.class.name}"
+  puts "   Message: #{e.message}"
+  puts "   Backtrace: #{e.backtrace.first(3).join("\n   ")}"
+  puts ""
+  puts "Common fixes:"
+  puts "  - Verify API key is set: echo $OPENAI_API_KEY"
+  puts "  - Check network connection"
+  puts "  - Ensure Ruby >= 3.2.0"
+end
+```
+
+**Expected output**:
+```
+Testing SwarmSDK installation...
+Ruby version: 3.2.0
+
+Test 1: Creating swarm...
+✅ Swarm created successfully
+
+Test 2: Executing task...
+✅ Execution successful
+   Response: SwarmSDK is working!
+   Duration: 1.23s
+   Cost: $0.0015
+   Agents: [:test_agent]
+
+Test 3: Testing delegation...
+✅ Delegation works
+   Agents involved: [:primary, :helper]
+```
+
+## Next Steps
+
+Congratulations! You've learned the fundamentals of SwarmSDK.
+
+### Continue Learning
+
+**Core Tutorials** (in recommended order):
+1. **[Complete Tutorial](complete-tutorial.md)** - Comprehensive guide building a real application
+2. **[Quick Start CLI](quick-start-cli.md)** - Using SwarmSDK from the command line
+3. **[Rails Integration](rails-integration.md)** - Integrating SwarmSDK with Rails applications
+
+**Advanced Topics**:
+- **[Permissions Guide](permissions.md)** - Control file and command access securely
+- **[Hooks Complete Guide](hooks-complete-guide.md)** - Customize behavior at every step
+- **[Scratchpad Guide](how-to-use-scratchpad.md)** - Share data between agents
+- **[Performance Tuning](performance-tuning.md)** - Optimize for speed and cost
+
+### Key Concepts to Explore
+
+**Agent Delegation Patterns**: Learn advanced collaboration strategies like hierarchical delegation, peer collaboration, and specialist chains.
+
+**Permissions System**: Discover how to restrict agents to specific directories, prevent dangerous commands, and create secure multi-agent systems.
+
+**Hooks and Customization**: Master pre/post hooks for tool calls, prompt injection, and workflow automation.
+
+**MCP Server Integration**: Connect to external tools and services using the Model Context Protocol.
+
+**Node-Based Workflows**: Build multi-stage pipelines where different agent teams handle each stage.
+
+### Real-World Examples
+
+Check out complete working examples in the `examples/` directory:
+
+- **Code Review Bot** - Multi-agent code review system
+- **Documentation Generator** - Automated documentation from code
+- **Test Suite Builder** - Generate comprehensive test suites
+- **Refactoring Assistant** - Find and fix code smells
+
+## Where to Get Help
+
+- **Documentation**: [SwarmSDK Guides](../README.md)
+- **API Reference**: [API Documentation](../../api/)
+- **Examples**: [Example Swarms](../../../examples/v2/)
+- **Issues**: [GitHub Issues](https://github.com/parruda/claude-swarm/issues)
+
+## Summary
+
+You've learned:
+
+✅ **What SwarmSDK is** - A framework for building collaborative AI agent teams
+
+✅ **Installation** - How to add SwarmSDK to your project and configure API keys
+
+✅ **Core concepts** - Swarms, agents, delegation, and tools
+
+✅ **Configuration formats** - YAML, Ruby DSL, and Markdown approaches
+
+✅ **Creating agents** - How to define agents with models, prompts, and tools
+
+✅ **Understanding results** - How to interpret execution results and metrics
+
+✅ **Common workflows** - Single agent, multi-agent, and tool usage patterns
+
+✅ **Avoiding pitfalls** - Common mistakes and how to fix them
+
+✅ **Testing setup** - How to verify everything works correctly
+
+**Next**: [Complete Tutorial →](complete-tutorial.md)
+
+---
+
+## Quick Reference Card
+
+### Minimal Working Swarm (Ruby DSL)
+
+```ruby
+require 'swarm_sdk'
+
+swarm = SwarmSDK.build do
+  name "My Swarm"
+  lead :agent_name
+
+  agent :agent_name do
+    description "What it does"
+    model "gpt-4"
+    system_prompt "Instructions"
+    tools :Write, :Edit
+  end
+end
+
+result = swarm.execute("Your task here")
+puts result.content if result.success?
+```
+
+### Minimal Working Swarm (YAML)
+
+```yaml
+# swarm.yml
+version: 2
+swarm:
+  name: "My Swarm"
+  lead: agent_name
+  agents:
+    agent_name:
+      description: "What it does"
+      model: "gpt-4"
+      system_prompt: "Instructions"
+      tools:
+        - Write
+        - Edit
+```
+
+```ruby
+# run.rb
+require 'swarm_sdk'
+swarm = SwarmSDK::Swarm.load('swarm.yml')
+result = swarm.execute("Your task here")
+puts result.content if result.success?
+```
+
+### Minimal Working Agent (Markdown)
+
+```markdown
+<!-- agents/agent_name.md -->
+---
+description: "What it does"
+model: "gpt-4"
+tools:
+  - Write
+  - Edit
+---
+
+# Agent Name
+
+Instructions for the agent.
+```
+
+```yaml
+# swarm.yml
+version: 2
+swarm:
+  name: "My Swarm"
+  lead: agent_name
+  agents:
+    agent_name: "agents/agent_name.md"
+```
+
+```ruby
+# run.rb
+require 'swarm_sdk'
+swarm = SwarmSDK::Swarm.load('swarm.yml')
+result = swarm.execute("Your task here")
+```
+
+### Essential Fields
+
+**Required**:
+- `name` - Swarm name (for logging)
+- `lead` - Entry point agent (symbol)
+- `description` - Agent's role (required for each agent)
+
+**Common**:
+- `model` - LLM to use (e.g., "gpt-4", "claude-sonnet-4")
+- `system_prompt` - Agent instructions
+- `tools` - Additional tools (Write, Edit, Bash, etc.)
+- `delegates_to` - Agents this agent can delegate to
+- `directory` - Working directory for file operations
+
+### Default Tools (Always Available)
+
+Unless `disable_default_tools: true`:
+- **Read** - Read files
+- **Grep** - Search file contents
+- **Glob** - Find files by pattern
+- **TodoWrite** - Track tasks
+- **Think** - Extended reasoning
+- **WebFetch** - Fetch and process web content
+- **ScratchpadWrite, ScratchpadRead, ScratchpadList** - Shared scratchpad (volatile)
+
+### Common Tools (Add Explicitly)
+
+```ruby
+tools :Write        # Create/overwrite files
+tools :Edit         # Modify existing files
+tools :MultiEdit    # Batch file edits
+tools :Bash         # Run shell commands
+```
+
+### Delegation Pattern
+
+```ruby
+agent :leader do
+  description "Coordinates work"
+  model "gpt-4"
+  delegates_to :worker  # Creates DelegateTaskToWorker tool
+end
+
+agent :worker do
+  description "Does the work"
+  model "gpt-4"
+end
+```
+
+### Result Object Methods
+
+```ruby
+result = swarm.execute("task")
+
+result.content         # Response text
+result.success?        # Boolean
+result.duration        # Float (seconds)
+result.total_cost      # Float (USD)
+result.total_tokens    # Integer
+result.agents_involved # Array[Symbol]
+result.error           # Exception | nil
+```
+
+### Model Examples
+
+**OpenAI**:
+- `"gpt-4"` - Most capable
+- `"gpt-4-turbo"` - Faster, cheaper
+- `"gpt-3.5-turbo"` - Fast, economical
+
+**Anthropic**:
+- `"claude-opus-4"` - Most capable
+- `"claude-sonnet-4"` - Balanced
+- `"claude-haiku-4"` - Fast, economical
+
+### Parameters for Tuning
+
+```ruby
+parameters temperature: 0.7    # 0.0 (focused) to 2.0 (creative)
+parameters top_p: 0.9          # 0.0 to 1.0 (diversity)
+parameters max_tokens: 2000    # Response length limit
+```
+
+### Common Patterns
+
+**Single agent**:
+```ruby
+swarm = SwarmSDK.build do
+  name "Solo"
+  lead :worker
+  agent :worker do
+    description "Does everything"
+    model "gpt-4"
+  end
+end
+```
+
+**Delegation chain**:
+```ruby
+swarm = SwarmSDK.build do
+  name "Chain"
+  lead :first
+  agent :first { description "First"; model "gpt-4"; delegates_to :second }
+  agent :second { description "Second"; model "gpt-4"; delegates_to :third }
+  agent :third { description "Third"; model "gpt-4" }
+end
+```
+
+**Parallel team**:
+```ruby
+swarm = SwarmSDK.build do
+  name "Team"
+  lead :coordinator
+  agent :coordinator do
+    description "Coordinates"
+    model "gpt-4"
+    delegates_to :frontend, :backend, :testing
+  end
+  agent :frontend { description "Frontend"; model "gpt-4" }
+  agent :backend { description "Backend"; model "gpt-4" }
+  agent :testing { description "Testing"; model "gpt-4" }
+end
+```