Config

Config

Configuration types for controlling model behavior.

Config

The Config function creates a Zod schema for parsing model configuration. All fields are optional and provider-specific.

import { Config } from "@adaline/types";

const config = Config().parse({
  temperature: 0.7,
  maxTokens: 500,
  topP: 0.9,
  frequencyPenalty: 0,
  presencePenalty: 0,
  stop: ["\n"],
});

Common Fields

temperature · number

Controls randomness. 0 is deterministic, higher values are more creative. Range: 0 to 2.

maxTokens · number

Maximum number of tokens to generate.

topP · number

Nucleus sampling threshold. Range: 0 to 1.

frequencyPenalty · number

Penalize tokens based on frequency. Range: -2 to 2.

presencePenalty · number

Penalize tokens based on presence. Range: -2 to 2.

stop · string[]

Stop sequences that halt generation.

seed · number

Random seed for deterministic outputs (provider support varies).

Embedding Config

For embedding models, the config supports:

const embeddingConfig = Config().parse({
  encodingFormat: "float",
  dimensions: 256,
});

encodingFormat · string

The encoding format for embeddings. Common values: "float", "base64".

dimensions · number

The number of dimensions for the output embeddings.