UNPKG

@ai-sdk/mistral

Version:

The **[Mistral provider](https://ai-sdk.dev/providers/ai-sdk-providers/mistral)** for the [AI SDK](https://ai-sdk.dev/docs) contains language model, embedding model, and speech model support for Mistral APIs.

72 lines (63 loc) 1.78 kB
import { z } from 'zod/v4'; // https://docs.mistral.ai/getting-started/models/models_overview/ export type MistralChatModelId = | 'ministral-3b-latest' | 'ministral-8b-latest' | 'ministral-14b-latest' | 'mistral-large-latest' | 'mistral-medium-latest' | 'mistral-medium-3' | 'mistral-large-2512' | 'mistral-medium-2508' | 'mistral-medium-2505' | 'mistral-small-2506' | 'pixtral-large-latest' // reasoning config support models | 'mistral-medium-3.5' | 'mistral-small-latest' | 'mistral-small-2603' // reasoning models | 'magistral-medium-latest' | 'magistral-small-latest' | 'magistral-medium-2509' | 'magistral-small-2509' | (string & {}); export const mistralLanguageModelChatOptions = z.object({ /** * Whether to inject a safety prompt before all conversations. * * Defaults to `false`. */ safePrompt: z.boolean().optional(), documentImageLimit: z.number().optional(), documentPageLimit: z.number().optional(), /** * Whether to use structured outputs. * * @default true */ structuredOutputs: z.boolean().optional(), /** * Whether to use strict JSON schema validation. * * @default false */ strictJsonSchema: z.boolean().optional(), /** * Whether to enable parallel function calling during tool use. * When set to false, the model will use at most one tool per response. * * @default true */ parallelToolCalls: z.boolean().optional(), /** * Controls the reasoning effort for models that support adjustable reasoning. * * - `'high'`: Enable reasoning * - `'none'`: Disable reasoning */ reasoningEffort: z.enum(['high', 'none']).optional(), }); export type MistralLanguageModelChatOptions = z.infer< typeof mistralLanguageModelChatOptions >;