Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.ai-stats.phaseo.app/llms.txt

Use this file to discover all available pages before exploring further.

Presets are reusable configurations that help teams standardize prompts, model preferences, and routing defaults. They are managed in the AI Stats dashboard and can be shared across a team.

​
What a preset can include

  • A system prompt to prepend to every request.
  • Allowed models or model families.
  • Provider allow/ignore lists for routing preferences.
  • Default parameters (temperature, top_p, max_tokens, and similar settings).
  • Optional reasoning defaults for supported models.
Preset names are prefixed with @ to keep them easy to recognize.

​
How preset merging works

When a request resolves with a preset in gateway context, the preset is applied before provider routing:
  • Default parameters only fill fields that are missing in the request body. They do not overwrite values already provided by the caller.
  • If the request already has a system message, the preset prompt is prepended. If the request uses an Anthropic-style system field, the preset prompt is prepended there instead.
  • Provider allow/ignore lists are applied before provider selection, so they narrow the fallback pool instead of acting as a cosmetic label.
  • Requests outside the preset’s allowed model list are rejected early instead of being silently rerouted.
This makes presets the main public surface for reusable request defaults and lightweight compatibility transforms without forcing every caller to duplicate the same prompt or parameter logic.

​
Current public preset surface

The dashboard preset flow is intentionally scoped to a stable, explicit subset of request shaping:
  • system prompt injection
  • model allowlists
  • provider allow/ignore routing constraints
  • decoding and generation defaults
  • reasoning defaults
If you need more complex caller-specific transforms, keep them in one application boundary layer and let presets carry the reusable team-wide defaults.

​
Where to manage presets

Create and manage presets in Dashboard -> Settings -> Presets. Use them consistently across experiments and production workloads so the team is aligned on prompt and routing behavior.

​
When to use presets

  • Standardize system prompts across multiple services.
  • Constrain routing to approved providers for compliance.
  • Keep default parameters consistent across environments.
  • Give migration projects one durable place to carry prompt, routing, and parameter defaults while the application code stays mostly unchanged.
Last modified on May 6, 2026