new·The score now tells you which way it movedA brain's exam only ever grows: its own material writes questions, and so does every question a real caller asked and did not get answered. The score is a percentage over that growing set, so a brain that learned more could post a smaller number — and this week three did. One of them answered two MORE questions than the week before and showed eighteen points less. Printed as a single percentage, that reads as decline to a reader and as punishment to anyone who contributes material.all news →
mozg.beta
Sign in

AI SDK · Providers · all subjects

mistral/options

4 notes, read out of this brain and free to use. Each one was extracted from a source and is re-checked against its exam.

Mistral provider customization options

When creating a customized Mistral provider instance with `createMistral()`, the following optional settings are available: baseURL (string, defaults to https://api.mistral.ai/v1), apiKey (string, defaults to MISTRAL_API_KEY environment variable), headers (Record<string,string> for custom headers), and fetch (custom fetch implementation).

Mistral language model provider options

Mistral language models support additional provider options via `providerOptions.mistral` using `MistralLanguageModelChatOptions`: safePrompt (boolean, defaults to false, injects safety prompt), documentImageLimit (number, maximum images in document), documentPageLimit (number, maximum pages in document), strictJsonSchema (boolean, defaults to false, enables strict JSON schema validation), structuredOutputs (boolean, defaults to true, enables structured outputs), and parallelToolCalls (boolean, defaults to true, enables parallel function calling).

Mistral transcription model provider options

Mistral transcription models support additional provider options via `providerOptions.mistral` using `MistralTranscriptionModelOptions`: language (string, language code like 'en' to improve accuracy), temperature (number, sampling temperature), timestampGranularities (Array<'segment' | 'word'>, granularities for timestamps - cannot be combined with language), diarize (boolean, identifies speakers), and contextBias (string[], up to 100 words/phrases for spelling guidance, items cannot contain commas or whitespace and use underscores for multi-word phrases).

Mistral embedding model provider options

Mistral embedding models support additional provider options via `providerOptions.mistral` using `MistralEmbeddingModelOptions`: metadata (Record<string, unknown>, additional metadata to attach), outputDimension (number, output embedding dimension when supported), and outputDtype (string, output data type when supported - accepts 'float', 'int8', 'uint8', 'binary', or 'ubinary').

Give your agent this brain