Overview
Parameter Control
Sliders, toggles, and selects for fine-tuning AI parameters
settingscontrolsparameters
Props
| Prop | Type | Default | Description |
|---|---|---|---|
parameters | ParameterControlItem[] | required | Array of parameter definitions with id, label, type (slider | toggle | select | matrix), value, min, max, step, options, description |
onChange | (id: string, value: unknown) => void | required | Callback fired when a parameter value changes, with the parameter id and new value |
title | string | undefined | Optional heading above the parameter controls |
layout | "vertical" | "horizontal" | "vertical" | Layout direction for the controls |
Docs
Parameters sit between user intent and model behavior, acting as the dials that control how the AI interprets input and generates output. They let users specify how exploratory or constrained, how creative or deterministic, and how concise or verbose the generation should be. The pattern has evolved from typed inline flags — popularized by early image generators — into more accessible controls like sliders, toggles, dropdowns, and preset bundles that welcome users at every skill level.
Variants
- Inline flags — Typed tokens within the prompt itself that modify generation behavior. High precision for experts but a steep learning curve for newcomers.
- Toggles and switches — Binary controls for two-state parameters such as formal vs. casual tone, or speed vs. quality trade-offs.
- Sliders — Continuous or stepped controls for adjusting numerical weights like temperature, length, or creativity levels.
- Presets — Bundled configurations that wrap multiple parameters into clear choices like 'Draft' vs. 'Publish-ready', simplifying decision-making.
Use Cases
- Fine-tuning generation temperature and creativity
- Controlling output length, format, and aspect ratio
- Selecting tone, voice, or writing style before generation
- Adjusting speed vs. quality trade-offs for different workflows
- Bundling advanced settings into beginner-friendly presets
Best Practices
- Ship with sensible defaults — most users won't adjust parameters, so the out-of-the-box experience should produce good results.
- Use progressive disclosure: surface one or two key controls by default and tuck advanced options behind expandable panels.
- Bundle complex parameter combinations into named presets that clearly communicate their effect (e.g. 'Creative Draft' vs. 'Precise Report').
- Label any option that affects cost, speed, or resource usage so users can make informed choices before generation.
- Guard against extreme settings that could produce failures — warn when values are likely to cause truncation, nonsensical output, or excessive cost.