Presets

Reusable request configurations for model routing and behavior

Presets let you move request configuration out of application code. Each preset can define model defaults, provider preferences, system prompts, and generation parameters. Your app only references a stable preset ID.

Define a preset

preset-support-assistant.json

{
  "id": "support-assistant-v1",
  "model": "anthropic/claude-sonnet-4.5",
  "messages": [
    {
      "role": "system",
      "content": "You are a concise customer support assistant."
    }
  ],
  "provider": {
    "order": ["bedrock-us-east-2", "openai-api"],
    "allow_fallbacks": true,
    "sort": "latency"
  },
  "temperature": 0.2,
  "max_tokens": 700
}

Apply a preset in requests

Model Alias

{
  "model": "@preset/support-assistant-v1",
  "messages": [{ "role": "user", "content": "Summarize this ticket thread." }]
}

Request values override preset values

Request-level keys are shallow-merged on top of preset defaults. Use this to keep a stable baseline while adjusting values like `temperature` per call.