Skip to content

workshop.md

Latest commit

 

History

History
217 lines (158 loc) · 7.09 KB

workshop.md

File metadata and controls

217 lines (158 loc) · 7.09 KB

Building with Prism: A Hands-On Workshop

Workshop Overview

This 3-hour workshop will guide participants through Prism's powerful capabilities for integrating AI into Laravel applications. Participants will build 2-3 practical AI workflows of increasing complexity, gaining hands-on experience with Prism's unified provider interface, multimodal inputs, structured output, and tools system.

Prerequisites

  • Basic knowledge of Laravel and PHP
  • PHP 8.2+ and Laravel 11.0+
  • API keys for at least one of these providers:
    • OpenAI (recommended)
    • Anthropic (recommended for advanced features)
    • Mistral, Ollama, or other supported providers (optional)

Workshop Structure

  • Setup & Introduction (30 minutes)
  • Project 1: Content Generation with Prism (45 minutes)
  • Project 2: Multimodal Analysis with Prism (60 minutes)
  • Project 3: Building an AI Assistant with Prism Tools (45 minutes)

Setup & Introduction (30 minutes)

  1. Prism Overview (10 minutes)

    • Introduction to Prism's architecture and philosophy
    • Provider abstraction and switching between providers
    • Key features: tools, structured output, multimodal support

  2. Prism Setup and Configuration (20 minutes)

    • Installing Prism in a Laravel project
    • Configuring providers and API keys
    • Understanding the Prism configuration file
    • Testing the installation with a simple text generation example

Project 1: Content Generation with Prism (45 minutes)

Difficulty: Beginner

This project will demonstrate Prism's core text generation capabilities and provider abstraction, allowing participants to understand how to build flexible AI-powered content generation.

Implementation Steps

  1. Setting up Basic Text Generation

    • Using the Prism::text() API
    • Adding system prompts using withSystemPrompt()
    • Creating prompt templates as Laravel views
    • Switching between different providers with the same code

  2. Implementing Content Generation

    • Creating a reusable content generation service
    • Handling different content types with prompt engineering
    • Implementing different generation parameters (temperature, max tokens)
    • Error handling for generation failures

  3. Building a Content Generation API

    • Creating an API endpoint for content generation
    • Implementing provider switching based on request parameters
    • Returning generated content to clients

Learning Outcomes

  • Core Prism text generation workflows
  • Provider abstraction and switching between providers
  • Using system prompts and templates effectively
  • Managing generation parameters
  • Basic error handling

Project 2: Multimodal Analysis with Prism (60 minutes)

Difficulty: Intermediate

This project will explore Prism's support for multimodal inputs and structured output, focusing on document and image analysis capabilities.

Implementation Steps

  1. Working with Images and Documents

    • Using the Image and Document value objects
    • Handling different input methods (path, base64, URL)
    • Understanding provider capabilities and limitations
    • Creating multimodal message sequences

  2. Implementing Structured Output

    • Defining schemas using Prism's schema system
    • Using ObjectSchema, StringSchema, and other schema types
    • Working with Prism::structured() API
    • Handling structured parsing failures

  3. Building a Document Analysis Service

    • Creating a document processing pipeline
    • Extracting structured information from documents
    • Handling different document types (PDF, images, text)
    • Implementing fallbacks when structured output fails

Learning Outcomes

  • Working with multimodal inputs in Prism
  • Implementing structured output with schemas
  • Handling provider-specific capabilities and limitations
  • Building more complex AI workflows
  • Advanced error handling strategies

Project 3: Building an AI Assistant with Prism Tools (45 minutes)

Difficulty: Advanced

This project will demonstrate Prism's powerful tools system, allowing participants to build interactive AI assistants that can take actions within their application.

Implementation Steps

  1. Understanding Prism's Tools System

    • Tool concepts and architecture
    • Creating basic tools using Tool::as()
    • Defining tool parameters and schemas
    • Handling tool execution

  2. Implementing Custom Tools

    • Creating a database search tool
    • Building a calculation/processing tool
    • Implementing an API integration tool
    • Managing tool results and error handling

  3. Building a Conversational Assistant

    • Managing conversation state with message chains
    • Controlling tool choice with withToolChoice()
    • Implementing multi-step interactions
    • Streaming responses with Prism

Learning Outcomes

  • Creating and using custom tools with Prism
  • Building interactive AI assistants
  • Managing multi-step conversations
  • Implementing streaming responses
  • Advanced Prism workflows

Conclusion and Extensions

  • Recap of Prism's key capabilities
  • Comparing provider-specific features
  • Suggestions for extending the projects
  • Resources for further learning

Workshop Materials

  • Workshop repository with starter code
  • Solution code for reference
  • Sample prompts, schemas, and tools
  • Example documents and images for testing

Setup Instructions

1. Create a new Laravel Project and Install Prism

composer create-project laravel/laravel prism-workshop
cd prism-workshop
composer require prism-php/prism
php artisan vendor:publish --tag=prism-config

2. Configure Prism Providers

Edit your .env file to add API keys:

OPENAI_API_KEY=your_openai_key_here
ANTHROPIC_API_KEY=your_anthropic_key_here
# Add other provider keys as needed

3. Create Project Directories

php artisan make:controller PrismController
mkdir -p resources/views/prism
mkdir -p resources/views/prompts

4. Setup Routes

Edit routes/web.php to add routes for each project:

Route::get('/', function () {
    return view('welcome');
});

Route::prefix('prism')->group(function () {
    Route::get('/project1', [App\Http\Controllers\PrismController::class, 'project1']);
    Route::get('/project2', [App\Http\Controllers\PrismController::class, 'project2']);
    Route::get('/project3', [App\Http\Controllers\PrismController::class, 'project3']);
    
    // Add API endpoints as needed
    Route::post('/generate', [App\Http\Controllers\PrismController::class, 'generate']);
    Route::post('/analyze', [App\Http\Controllers\PrismController::class, 'analyze']);
    Route::post('/assistant', [App\Http\Controllers\PrismController::class, 'assistant']);
});

5. Start the Development Server

php artisan serve

Additional Workshop Tips

  • Have participants experiment with different providers to see how Prism's abstraction works
  • Prepare examples of structured output schemas of varying complexity
  • Create sample tools that demonstrate different parameter types
  • Provide example documents and images for multimodal testing
  • Consider comparing the same workflow across different providers to highlight Prism's flexibility