Appearance

Prompt Library ​

The Prompt Library is a centralized repository for saving, organizing, and reusing prompts. Tag prompts, search by content, and load them directly into the Prompt Lab or Playground.

Access ​

Navigate to /ai-orbit/prompts.

Prompt Cards ​

Each saved prompt displays:

FieldDescription
NameDescriptive title
ContentThe system prompt or full prompt text
InstructionOptional user instruction template
TagsCategory labels for organization
MetadataTemperature, max tokens, top P, context
UpdatedLast modified timestamp

Creating a Prompt ​

Click "New Prompt" and fill in the form:

Name:        Customer Support Greeting
Content:     You are a friendly customer support agent for Acme Corp.
             Be concise, professional, and helpful.
Instruction: Greet the customer and ask how you can help.
Tags:        support, greeting, customer-service
Temperature: 0.7
Max Tokens:  200
Top P:       1.0
Context:     {"previous_interaction": "none"}

Tags ​

Tags help you organize prompts. Add multiple tags separated by commas:

  • support, sales, onboarding
  • gpt-4, claude, comparison
  • high-temperature, low-temperature

Click a tag to filter the library to prompts with that tag.

Searching ​

The Prompt Library supports full-text search across:

  • Prompt names
  • Prompt content
  • Instructions

Type in the search box and results update in real time.

Loading Prompts ​

Into Prompt Lab ​

  1. Open the Prompt Lab
  2. Browse recent prompts in the sidebar
  3. Click any prompt to populate the form
  4. Adjust model slots and run

Into Playground ​

  1. Open an Agent Sandbox
  2. Use the "Load from Library" feature
  3. Select a prompt to set the system prompt

Editing Prompts ​

Click the edit icon on any prompt card to modify:

  • Name and content
  • Instruction
  • Tags
  • Metadata (temperature, max tokens, top P, context)

Changes are saved immediately.

Deleting Prompts ​

Click the delete icon to remove a prompt. This action is permanent.

Programmatic Access ​

php
use Ashrafic\AiOrbit\Models\SavedPrompt;

// Create a prompt
$prompt = SavedPrompt::create([
    'name' => 'Support Greeting',
    'content' => 'You are a friendly support agent...',
    'instruction' => 'Greet the customer...',
    'meta' => [
        'temperature' => 0.7,
        'max_tokens' => 200,
        'top_p' => 1.0,
        'context' => ['previous' => 'none'],
    ],
    'tags' => ['support', 'greeting'],
]);

// Search prompts
$results = SavedPrompt::where('name', 'like', '%support%')
    ->orWhere('content', 'like', '%support%')
    ->get();

// Get prompts by tag
$supportPrompts = SavedPrompt::whereJsonContains('tags', 'support')->get();

Use Cases ​

Prompt Versioning ​

Save iterations of prompts as you refine them. Compare versions by loading them into the Prompt Lab.

Team Sharing ​

Share prompt templates with your team. Everyone loads the same prompt from the library, ensuring consistency.

A/B Testing ​

Save two variants of a prompt with different temperatures or instructions. Load both into the Prompt Lab to compare.

Documentation ​

Document your best-performing prompts with tags and metadata. New team members can browse the library to understand your prompt patterns.

Quick Start ​

Build a library of starter prompts for common tasks. Load them into the Playground to bootstrap new agent development.

Customization ​

Override the Prompt Library view:

bash
php artisan vendor:publish --tag=ai-orbit-views

Then edit resources/views/vendor/ai-orbit/livewire/prompt-library.blade.php.

Best Practices ​

  1. Use descriptive names — "Support Greeting" is better than "Prompt 1"
  2. Tag consistently — Establish a tagging convention for your team
  3. Include instructions — Separate system prompt from user instruction
  4. Save metadata — Record temperature and max tokens for reproducibility
  5. Review regularly — Delete outdated prompts to keep the library clean