Interface ChatMistralAIInput

Input to chat model class.

interface ChatMistralAIInput {
    apiKey?: string;
    endpoint?: string;
    maxTokens?: number;
    model?: string;
    modelName?: string;
    randomSeed?: number;
    safeMode?: boolean;
    safePrompt?: boolean;
    seed?: number;
    streamUsage?: boolean;
    streaming?: boolean;
    temperature?: number;
    topP?: number;
}

Hierarchy (view full)

Implemented by

Properties

apiKey? endpoint? maxTokens? model? modelName? randomSeed? safeMode? safePrompt? seed? streamUsage? streaming? temperature? topP?

Properties

OptionalapiKey

apiKey?: string

The API key to use.

Default

{process.env.MISTRAL_API_KEY}

Optionalendpoint

endpoint?: string

Override the default endpoint.

OptionalmaxTokens

maxTokens?: number

The maximum number of tokens to generate in the completion. The token count of your prompt plus max_tokens cannot exceed the model's context length.

Optionalmodel

model?: string

The name of the model to use.

Default

{"mistral-small-latest"}

OptionalmodelName

modelName?: string

The name of the model to use. Alias for model

Default

{"mistral-small-latest"}

OptionalrandomSeed

randomSeed?: number

The seed to use for random sampling. If set, different calls will generate deterministic results. Alias for seed

OptionalsafeMode

safeMode?: boolean

Whether to inject a safety prompt before all conversations.

Default

{false}

Deprecated

use safePrompt instead

OptionalsafePrompt

safePrompt?: boolean

Whether to inject a safety prompt before all conversations.

Default

{false}

Optionalseed

seed?: number

The seed to use for random sampling. If set, different calls will generate deterministic results.

OptionalstreamUsage

streamUsage?: boolean

Whether or not to include token usage in the stream.

Default

{true}

Optionalstreaming

streaming?: boolean

Whether or not to stream the response.

Default

{false}

Optionaltemperature

temperature?: number

What sampling temperature to use, between 0.0 and 2.0. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic.

Default

{0.7}

OptionaltopP

topP?: number

Nucleus sampling, where the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. Should be between 0 and 1.

Default

{1}

Settings

Member Visibility

On This Page

Properties