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?: string

The API key to use.

{process.env.MISTRAL_API_KEY}
endpoint?: string

Override the default endpoint.

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.

model?: string

The name of the model to use.

{"mistral-small-latest"}
modelName?: string

The name of the model to use. Alias for model

{"mistral-small-latest"}
randomSeed?: number

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

safeMode?: boolean

Whether to inject a safety prompt before all conversations.

{false}

use safePrompt instead

safePrompt?: boolean

Whether to inject a safety prompt before all conversations.

{false}
seed?: number

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

streamUsage?: boolean

Whether or not to include token usage in the stream.

{true}
streaming?: boolean

Whether or not to stream the response.

{false}
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.

{0.7}
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.

{1}
""