​
Presets: Configuration That Actually Makes Sense

Stop hardcoding AI settings like it’s 2019 Presets transform your LLM workflows from messy, scattered configurations into clean, reusable templates. Think of them as your AI’s favorite outfits—perfectly tailored for specific occasions and ready to wear at a moment’s notice.

​
What Are Presets?

Your AI’s wardrobe for every occasion Presets are named configuration bundles that package everything your model needs to shine in specific scenarios. Instead of repeating the same complex settings across your codebase, create once and reference everywhere. Real-world preset examples:
  • 🎯 email-copywriter → Persuasive marketing content with high creativity
  • 📞 inbound-classifier → Lightning-fast customer inquiry sorting
  • 👨‍💻 code-reviewer → Thorough, constructive PR analysis
  • 💬 customer-support → Empathetic, solution-focused responses
Each preset controls:
  • 🎭 Model selection and fallback chains
  • 🚀 Provider routing (optimize for speed, cost, or quality)
  • 🧠 System prompts that set the perfect tone
  • ⚙️ Generation parameters (temperature, top_p, max tokens)
  • 🎛️ Provider inclusion/exclusion rules

​
⚡ Quick Start

From zero to AI hero in 30 seconds Step 1: Create your preset in the AnyAPI dashboard Step 2: Reference it in your code 
{
  "model": "@preset/email-copywriter",
  "messages": [
    {
      "role": "user",
      "content": "Write a launch announcement for our new AI feature"
    }
  ]
}
That’s it. Your AI now knows exactly how to behave, what tone to use, and which models to prefer—all without cluttering your application code.

​
🎯 Why Presets Are a Game-Changer

​
Clean Code, Clear Intent

// Before: What does this even do?
const result = await client.chat.completions.create({
  model: "anthropic/claude-3.5-sonnet",
  temperature: 0.7,
  max_tokens: 500,
  system: "You are a helpful marketing copywriter who writes engaging email content...",
  // ... 15 more configuration lines
});

// After: Crystal clear purpose
const result = await client.chat.completions.create({
  model: "@preset/email-copywriter",
  messages: [{ role: "user", content: userPrompt }]
});

​
Deploy-Free Optimization

Update your AI behavior without touching code:
  • 🔄 Switch to newer, better models instantly
  • 📝 Refine system prompts based on performance data
  • ⚡ Adjust parameters for different quality/speed trade-offs
  • 💰 Optimize provider routing to reduce costs

​
Team Collaboration Made Simple

  • Marketing tweaks copy generation presets
  • DevOps optimizes routing preferences
  • Product managers adjust model behavior
  • Developers focus on application logic

​
🛠️ Three Ways to Use Presets

Clean and semantic—your code tells a story: 
{
  "model": "@preset/customer-support",
  "messages": [
    {
      "role": "user",
      "content": "My order hasn't arrived and I'm frustrated"
    }
  ]
}

​
Explicit Preset Field

When you need to be extra clear: 
{
  "preset": "customer-support",
  "messages": [
    {
      "role": "user", 
      "content": "My order hasn't arrived and I'm frustrated"
    }
  ]
}

​
Smart Overrides

Use preset defaults but customize on the fly: 
{
  "model": "@preset/creative-writer",
  "temperature": 0.9,  // Boost creativity for this specific request
  "messages": [
    {
      "role": "user",
      "content": "Write a wild sci-fi story about time-traveling cats"
    }
  ]
}

​
🎛️ Preset Management

Your mission control for AI configurations Head to the AnyAPI dashboard to craft your perfect presets. Each one includes: 🏷️ Name & Identity
  • Unique identifier that makes sense to your team
  • Description explaining the preset’s purpose
🤖 Model Strategy
  • Primary model selection
  • Backup models for bulletproof reliability
  • Provider routing preferences (speed vs. cost vs. quality)
💬 Behavioral Settings
  • System prompts that set the perfect personality
  • Temperature, top_p, and other generation parameters
  • Token limits and response formatting rules
🚀 Advanced Options
  • Provider inclusion/exclusion lists
  • Custom headers and metadata
  • Usage tracking and analytics tags

​
💡 Pro Tips

🎯 Be Specific: Name your presets based on their job, not their settings. blog-writer-casual beats gpt4-temp07-1000tokens. 🔄 Version Control: Use descriptive names like email-v2-summer-campaign when testing major changes. ⚖️ Balance Act: Mix high-quality primary models with cost-effective fallbacks for optimal spend. 📊 Monitor Performance: Track which presets perform best and iterate based on real usage data.
Ready to bring order to your AI chaos? Start building presets that actually make sense.