Quick Start Guide

This guide will get you up and running with Pulsar in under 10 minutes.

Installation

Install Pulsar using pip:

pip install pulsar-compose

Or using Poetry:

poetry add pulsar-compose

For local AI models, install Ollama:

# Install Ollama
curl -fsSL https://ollama.ai/install.sh | sh

# Pull a model
ollama pull llama2

Your First Workflow

Create a file called hello_world.yml:

version: "0.1"
name: "Hello World Workflow"
description: "A simple workflow that greets the user"

agents:
  greeter:
    model: "gpt-4"
    provider: "openai"
    prompt: "You are a friendly assistant"

workflow:
  - type: "agent"
    step: "greet"
    agent: "greeter"
    prompt: "Say hello to {{input}} in a creative way"
    save_to: "greeting"

Run the workflow:

# Set your OpenAI API key
export OPENAI_API_KEY="your-api-key-here"

# Run the workflow
pulsar run hello_world.yml --input "Alice"

You should see output like:

🚀 Starting workflow: Hello World Workflow
📝 Step 1/1: greet
✅ Workflow completed successfully

Results:
greeting: "Hello Alice! 🌟 What a wonderful day to meet someone as amazing as you! How can I brighten your day today?"

Multi-Agent Workflow

Create a more complex workflow with multiple agents:

version: "0.1"
name: "Content Creation Pipeline"
description: "Generate and improve content using multiple AI agents"

agents:
  researcher:
    model: "gpt-4"
    provider: "openai"
    prompt: "You are a research specialist"

  writer:
    model: "claude-3-sonnet-20240229"
    provider: "anthropic"
    prompt: "You are a creative writer"

  editor:
    model: "gpt-4"
    provider: "openai"
    prompt: "You are an expert editor"

workflow:
  - type: "agent"
    step: "research"
    agent: "researcher"
    prompt: "Research key facts about: {{input}}"
    save_to: "research"

  - type: "agent"
    step: "write"
    agent: "writer"
    prompt: "Write an engaging article using this research: {{research}}"
    save_to: "draft"

  - type: "agent"
    step: "edit"
    agent: "editor"
    prompt: "Edit and improve this article: {{draft}}"
    save_to: "final"

Run it:

export OPENAI_API_KEY="your-openai-key"
export ANTHROPIC_API_KEY="your-anthropic-key"

pulsar run content_creation.yml --input "renewable energy"

Conditional Workflows

Add decision-making to your workflows:

version: "0.1"
name: "Smart Content Processor"
description: "Process content with conditional logic"

agents:
  analyzer:
    model: "gpt-4"
    provider: "openai"
    prompt: "You analyze content quality"

  improver:
    model: "claude-3-haiku-20240307"
    provider: "anthropic"
    prompt: "You improve content"

workflow:
  - type: "agent"
    step: "analyze"
    agent: "analyzer"
    prompt: "Rate this content quality on a scale of 1-10: {{input}}"
    save_to: "quality_score"

  - type: "conditional"
    step: "check_quality"
    if: "{{int(quality_score) >= 7}}"
    then:
      - type: "agent"
        step: "publish"
        agent: "analyzer"
        prompt: "Content is good enough: {{input}}"
        save_to: "result"
    else:
      - type: "agent"
        step: "improve"
        agent: "improver"
        prompt: "Improve this content: {{input}}"
        save_to: "improved_content"
      - type: "agent"
        step: "final_review"
        agent: "analyzer"
        prompt: "Review the improved content: {{improved_content}}"
        save_to: "result"

Using Templates

Pulsar supports Jinja2 templating for dynamic content:

version: "0.1"
name: "Template Example"
description: "Using templates for dynamic content"

variables:
  style: "professional"
  tone: "friendly"
  max_length: 500

agents:
  writer:
    model: "gpt-4"
    provider: "openai"
    prompt: "You are a {{style}} writer with a {{tone}} tone"

workflow:
  - type: "agent"
    step: "write"
    agent: "writer"
    prompt: |
      Write a {{style}} article about {{input}}.
      Keep it under {{max_length}} characters.
      Use a {{tone}} tone throughout.
    save_to: "article"

Run with custom variables:

pulsar run template_example.yml --input "artificial intelligence" --var style=creative --var tone=enthusiastic

Local AI Models

Use local models with Ollama for privacy and cost savings:

version: "0.1"
name: "Local AI Workflow"
description: "Using local Ollama models"

agents:
  local_assistant:
    model: "llama2"
    provider: "local"
    prompt: "You are a helpful local AI assistant"

workflow:
  - type: "agent"
    step: "assist"
    agent: "local_assistant"
    prompt: "Help with: {{input}}"
    save_to: "response"

Make sure Ollama is running:

# Start Ollama service
ollama serve

# Run the workflow
pulsar run local_workflow.yml --input "explain quantum computing"

Next Steps

Now that you have the basics, explore:

Workflow Fundamentals - Advanced workflow patterns
../user_guide/agents - AI provider configuration
../user_guide/templates - Template system details
../user_guide/examples - More example workflows

For API usage, see the ../api/index.