@mistralai/mistralai 1.10.0 → 1.11.0

package/src/models/components/agent.ts

@@ -90,11 +90,15 @@ export type Agent = {
   name: string;
   description?: string | null | undefined;
   handoffs?: Array<string> | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
   object?: AgentObject | undefined;
   id: string;
   version: number;
+  versions: Array<number>;
   createdAt: Date;
   updatedAt: Date;
+  deploymentChat: boolean;
+  source: string;
 };
 
 /** @internal */
@@ -271,20 +275,25 @@ export const Agent$inboundSchema: z.ZodType<Agent, z.ZodTypeDef, unknown> = z
     name: z.string(),
     description: z.nullable(z.string()).optional(),
     handoffs: z.nullable(z.array(z.string())).optional(),
+    metadata: z.nullable(z.record(z.any())).optional(),
     object: AgentObject$inboundSchema.default("agent"),
     id: z.string(),
     version: z.number().int(),
+    versions: z.array(z.number().int()),
     created_at: z.string().datetime({ offset: true }).transform(v =>
       new Date(v)
     ),
     updated_at: z.string().datetime({ offset: true }).transform(v =>
       new Date(v)
     ),
+    deployment_chat: z.boolean(),
+    source: z.string(),
   }).transform((v) => {
     return remap$(v, {
       "completion_args": "completionArgs",
       "created_at": "createdAt",
       "updated_at": "updatedAt",
+      "deployment_chat": "deploymentChat",
     });
   });
 
@@ -306,11 +315,15 @@ export type Agent$Outbound = {
   name: string;
   description?: string | null | undefined;
   handoffs?: Array<string> | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
   object: string;
   id: string;
   version: number;
+  versions: Array<number>;
   created_at: string;
   updated_at: string;
+  deployment_chat: boolean;
+  source: string;
 };
 
 /** @internal */
@@ -359,16 +372,21 @@ export const Agent$outboundSchema: z.ZodType<
   name: z.string(),
   description: z.nullable(z.string()).optional(),
   handoffs: z.nullable(z.array(z.string())).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
   object: AgentObject$outboundSchema.default("agent"),
   id: z.string(),
   version: z.number().int(),
+  versions: z.array(z.number().int()),
   createdAt: z.date().transform(v => v.toISOString()),
   updatedAt: z.date().transform(v => v.toISOString()),
+  deploymentChat: z.boolean(),
+  source: z.string(),
 }).transform((v) => {
   return remap$(v, {
     completionArgs: "completion_args",
     createdAt: "created_at",
     updatedAt: "updated_at",
+    deploymentChat: "deployment_chat",
   });
 });
 

package/src/models/components/agentconversation.ts

@@ -25,11 +25,16 @@ export type AgentConversation = {
    * Description of the what the conversation is about.
    */
   description?: string | null | undefined;
+  /**
+   * Custom metadata for the conversation.
+   */
+  metadata?: { [k: string]: any } | null | undefined;
   object?: AgentConversationObject | undefined;
   id: string;
   createdAt: Date;
   updatedAt: Date;
   agentId: string;
+  agentVersion?: number | null | undefined;
 };
 
 /** @internal */
@@ -61,16 +66,19 @@ export const AgentConversation$inboundSchema: z.ZodType<
 > = z.object({
   name: z.nullable(z.string()).optional(),
   description: z.nullable(z.string()).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
   object: AgentConversationObject$inboundSchema.default("conversation"),
   id: z.string(),
   created_at: z.string().datetime({ offset: true }).transform(v => new Date(v)),
   updated_at: z.string().datetime({ offset: true }).transform(v => new Date(v)),
   agent_id: z.string(),
+  agent_version: z.nullable(z.number().int()).optional(),
 }).transform((v) => {
   return remap$(v, {
     "created_at": "createdAt",
     "updated_at": "updatedAt",
     "agent_id": "agentId",
+    "agent_version": "agentVersion",
   });
 });
 
@@ -78,11 +86,13 @@ export const AgentConversation$inboundSchema: z.ZodType<
 export type AgentConversation$Outbound = {
   name?: string | null | undefined;
   description?: string | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
   object: string;
   id: string;
   created_at: string;
   updated_at: string;
   agent_id: string;
+  agent_version?: number | null | undefined;
 };
 
 /** @internal */
@@ -93,16 +103,19 @@ export const AgentConversation$outboundSchema: z.ZodType<
 > = z.object({
   name: z.nullable(z.string()).optional(),
   description: z.nullable(z.string()).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
   object: AgentConversationObject$outboundSchema.default("conversation"),
   id: z.string(),
   createdAt: z.date().transform(v => v.toISOString()),
   updatedAt: z.date().transform(v => v.toISOString()),
   agentId: z.string(),
+  agentVersion: z.nullable(z.number().int()).optional(),
 }).transform((v) => {
   return remap$(v, {
     createdAt: "created_at",
     updatedAt: "updated_at",
     agentId: "agent_id",
+    agentVersion: "agent_version",
   });
 });
 

package/src/models/components/agentcreationrequest.ts

@@ -84,6 +84,7 @@ export type AgentCreationRequest = {
   name: string;
   description?: string | null | undefined;
   handoffs?: Array<string> | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
 };
 
 /** @internal */
@@ -248,6 +249,7 @@ export const AgentCreationRequest$inboundSchema: z.ZodType<
   name: z.string(),
   description: z.nullable(z.string()).optional(),
   handoffs: z.nullable(z.array(z.string())).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
 }).transform((v) => {
   return remap$(v, {
     "completion_args": "completionArgs",
@@ -272,6 +274,7 @@ export type AgentCreationRequest$Outbound = {
   name: string;
   description?: string | null | undefined;
   handoffs?: Array<string> | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
 };
 
 /** @internal */
@@ -320,6 +323,7 @@ export const AgentCreationRequest$outboundSchema: z.ZodType<
   name: z.string(),
   description: z.nullable(z.string()).optional(),
   handoffs: z.nullable(z.array(z.string())).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
 }).transform((v) => {
   return remap$(v, {
     completionArgs: "completion_args",

package/src/models/components/agentscompletionrequest.ts

@@ -96,6 +96,7 @@ export type AgentsCompletionRequest = {
    * The seed to use for random sampling. If set, different calls will generate deterministic results.
    */
   randomSeed?: number | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
   /**
    * The prompt(s) to generate completions for, encoded as a list of dict with role and content.
    */
@@ -105,21 +106,27 @@ export type AgentsCompletionRequest = {
     | (UserMessage & { role: "user" })
     | (AssistantMessage & { role: "assistant" })
   >;
+  /**
+   * Specify the format that the model must output. By default it will use `{ "type": "text" }`. Setting to `{ "type": "json_object" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ "type": "json_schema" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide.
+   */
   responseFormat?: ResponseFormat | undefined;
   tools?: Array<Tool> | null | undefined;
   toolChoice?: ToolChoice | ToolChoiceEnum | undefined;
   /**
-   * presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
+   * The `presence_penalty` determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
    */
   presencePenalty?: number | undefined;
   /**
-   * frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
+   * The `frequency_penalty` penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
    */
   frequencyPenalty?: number | undefined;
   /**
    * Number of completions to return for each request, input tokens are only billed once.
    */
   n?: number | null | undefined;
+  /**
+   * Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content.
+   */
   prediction?: Prediction | undefined;
   parallelToolCalls?: boolean | undefined;
   /**
@@ -333,6 +340,7 @@ export const AgentsCompletionRequest$inboundSchema: z.ZodType<
   stream: z.boolean().default(false),
   stop: z.union([z.string(), z.array(z.string())]).optional(),
   random_seed: z.nullable(z.number().int()).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
   messages: z.array(
     z.union([
       SystemMessage$inboundSchema.and(
@@ -388,6 +396,7 @@ export type AgentsCompletionRequest$Outbound = {
   stream: boolean;
   stop?: string | Array<string> | undefined;
   random_seed?: number | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
   messages: Array<
     | (SystemMessage$Outbound & { role: "system" })
     | (ToolMessage$Outbound & { role: "tool" })
@@ -416,6 +425,7 @@ export const AgentsCompletionRequest$outboundSchema: z.ZodType<
   stream: z.boolean().default(false),
   stop: z.union([z.string(), z.array(z.string())]).optional(),
   randomSeed: z.nullable(z.number().int()).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
   messages: z.array(
     z.union([
       SystemMessage$outboundSchema.and(

package/src/models/components/agentscompletionstreamrequest.ts

@@ -95,6 +95,7 @@ export type AgentsCompletionStreamRequest = {
    * The seed to use for random sampling. If set, different calls will generate deterministic results.
    */
   randomSeed?: number | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
   /**
    * The prompt(s) to generate completions for, encoded as a list of dict with role and content.
    */
@@ -104,21 +105,27 @@ export type AgentsCompletionStreamRequest = {
     | (UserMessage & { role: "user" })
     | (AssistantMessage & { role: "assistant" })
   >;
+  /**
+   * Specify the format that the model must output. By default it will use `{ "type": "text" }`. Setting to `{ "type": "json_object" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ "type": "json_schema" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide.
+   */
   responseFormat?: ResponseFormat | undefined;
   tools?: Array<Tool> | null | undefined;
   toolChoice?: ToolChoice | ToolChoiceEnum | undefined;
   /**
-   * presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
+   * The `presence_penalty` determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
    */
   presencePenalty?: number | undefined;
   /**
-   * frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
+   * The `frequency_penalty` penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
    */
   frequencyPenalty?: number | undefined;
   /**
    * Number of completions to return for each request, input tokens are only billed once.
    */
   n?: number | null | undefined;
+  /**
+   * Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content.
+   */
   prediction?: Prediction | undefined;
   parallelToolCalls?: boolean | undefined;
   /**
@@ -344,6 +351,7 @@ export const AgentsCompletionStreamRequest$inboundSchema: z.ZodType<
   stream: z.boolean().default(true),
   stop: z.union([z.string(), z.array(z.string())]).optional(),
   random_seed: z.nullable(z.number().int()).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
   messages: z.array(
     z.union([
       SystemMessage$inboundSchema.and(
@@ -399,6 +407,7 @@ export type AgentsCompletionStreamRequest$Outbound = {
   stream: boolean;
   stop?: string | Array<string> | undefined;
   random_seed?: number | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
   messages: Array<
     | (SystemMessage$Outbound & { role: "system" })
     | (ToolMessage$Outbound & { role: "tool" })
@@ -427,6 +436,7 @@ export const AgentsCompletionStreamRequest$outboundSchema: z.ZodType<
   stream: z.boolean().default(true),
   stop: z.union([z.string(), z.array(z.string())]).optional(),
   randomSeed: z.nullable(z.number().int()).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
   messages: z.array(
     z.union([
       SystemMessage$outboundSchema.and(

package/src/models/components/agentupdaterequest.ts

@@ -84,6 +84,8 @@ export type AgentUpdateRequest = {
   name?: string | null | undefined;
   description?: string | null | undefined;
   handoffs?: Array<string> | null | undefined;
+  deploymentChat?: boolean | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
 };
 
 /** @internal */
@@ -248,9 +250,12 @@ export const AgentUpdateRequest$inboundSchema: z.ZodType<
   name: z.nullable(z.string()).optional(),
   description: z.nullable(z.string()).optional(),
   handoffs: z.nullable(z.array(z.string())).optional(),
+  deployment_chat: z.nullable(z.boolean()).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
 }).transform((v) => {
   return remap$(v, {
     "completion_args": "completionArgs",
+    "deployment_chat": "deploymentChat",
   });
 });
 
@@ -272,6 +277,8 @@ export type AgentUpdateRequest$Outbound = {
   name?: string | null | undefined;
   description?: string | null | undefined;
   handoffs?: Array<string> | null | undefined;
+  deployment_chat?: boolean | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
 };
 
 /** @internal */
@@ -320,9 +327,12 @@ export const AgentUpdateRequest$outboundSchema: z.ZodType<
   name: z.nullable(z.string()).optional(),
   description: z.nullable(z.string()).optional(),
   handoffs: z.nullable(z.array(z.string())).optional(),
+  deploymentChat: z.nullable(z.boolean()).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
 }).transform((v) => {
   return remap$(v, {
     completionArgs: "completion_args",
+    deploymentChat: "deployment_chat",
   });
 });
 

package/src/models/components/audiotranscriptionrequest.ts

@@ -21,6 +21,9 @@ import {
 } from "./timestampgranularity.js";
 
 export type AudioTranscriptionRequest = {
+  /**
+   * ID of the model to be used.
+   */
   model: string;
   file?: FileT | Blob | undefined;
   /**

package/src/models/components/batchjobin.ts

@@ -14,11 +14,26 @@ import {
 } from "./apiendpoint.js";
 
 export type BatchJobIn = {
+  /**
+   * The list of input files to be used for batch inference, these files should be `jsonl` files, containing the input data corresponding to the bory request for the batch inference in a "body" field. An example of such file is the following: ```json {"custom_id": "0", "body": {"max_tokens": 100, "messages": [{"role": "user", "content": "What is the best French cheese?"}]}} {"custom_id": "1", "body": {"max_tokens": 100, "messages": [{"role": "user", "content": "What is the best French wine?"}]}} ```
+   */
   inputFiles: Array<string>;
   endpoint: ApiEndpoint;
+  /**
+   * The model to be used for batch inference.
+   */
   model?: string | null | undefined;
+  /**
+   * In case you want to use a specific agent from the **deprecated** agents api for batch inference, you can specify the agent ID here.
+   */
   agentId?: string | null | undefined;
+  /**
+   * The metadata of your choice to be associated with the batch inference job.
+   */
   metadata?: { [k: string]: string } | null | undefined;
+  /**
+   * The timeout in hours for the batch inference job.
+   */
   timeoutHours?: number | undefined;
 };
 

package/src/models/components/chatcompletionrequest.ts

@@ -77,6 +77,9 @@ export type Messages =
   | (UserMessage & { role: "user" })
   | (AssistantMessage & { role: "assistant" });
 
+/**
+ * Controls which (if any) tool is called by the model. `none` means the model will not call any tool and instead generates a message. `auto` means the model can pick between generating a message or calling one or more tools. `any` or `required` means the model must call one or more tools. Specifying a particular tool via `{"type": "function", "function": {"name": "my_function"}}` forces the model to call that tool.
+ */
 export type ChatCompletionRequestToolChoice = ToolChoice | ToolChoiceEnum;
 
 export type ChatCompletionRequest = {
@@ -108,6 +111,7 @@ export type ChatCompletionRequest = {
    * The seed to use for random sampling. If set, different calls will generate deterministic results.
    */
   randomSeed?: number | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
   /**
    * The prompt(s) to generate completions for, encoded as a list of dict with role and content.
    */
@@ -117,22 +121,37 @@ export type ChatCompletionRequest = {
     | (UserMessage & { role: "user" })
     | (AssistantMessage & { role: "assistant" })
   >;
+  /**
+   * Specify the format that the model must output. By default it will use `{ "type": "text" }`. Setting to `{ "type": "json_object" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ "type": "json_schema" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide.
+   */
   responseFormat?: ResponseFormat | undefined;
+  /**
+   * A list of tools the model may call. Use this to provide a list of functions the model may generate JSON inputs for.
+   */
   tools?: Array<Tool> | null | undefined;
+  /**
+   * Controls which (if any) tool is called by the model. `none` means the model will not call any tool and instead generates a message. `auto` means the model can pick between generating a message or calling one or more tools. `any` or `required` means the model must call one or more tools. Specifying a particular tool via `{"type": "function", "function": {"name": "my_function"}}` forces the model to call that tool.
+   */
   toolChoice?: ToolChoice | ToolChoiceEnum | undefined;
   /**
-   * presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
+   * The `presence_penalty` determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
    */
   presencePenalty?: number | undefined;
   /**
-   * frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
+   * The `frequency_penalty` penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
    */
   frequencyPenalty?: number | undefined;
   /**
    * Number of completions to return for each request, input tokens are only billed once.
    */
   n?: number | null | undefined;
+  /**
+   * Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content.
+   */
   prediction?: Prediction | undefined;
+  /**
+   * Whether to enable parallel function calling during tool use, when enabled the model can call multiple tools in parallel.
+   */
   parallelToolCalls?: boolean | undefined;
   /**
    * Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used.
@@ -329,6 +348,7 @@ export const ChatCompletionRequest$inboundSchema: z.ZodType<
   stream: z.boolean().default(false),
   stop: z.union([z.string(), z.array(z.string())]).optional(),
   random_seed: z.nullable(z.number().int()).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
   messages: z.array(
     z.union([
       SystemMessage$inboundSchema.and(
@@ -388,6 +408,7 @@ export type ChatCompletionRequest$Outbound = {
   stream: boolean;
   stop?: string | Array<string> | undefined;
   random_seed?: number | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
   messages: Array<
     | (SystemMessage$Outbound & { role: "system" })
     | (ToolMessage$Outbound & { role: "tool" })
@@ -419,6 +440,7 @@ export const ChatCompletionRequest$outboundSchema: z.ZodType<
   stream: z.boolean().default(false),
   stop: z.union([z.string(), z.array(z.string())]).optional(),
   randomSeed: z.nullable(z.number().int()).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
   messages: z.array(
     z.union([
       SystemMessage$outboundSchema.and(

package/src/models/components/chatcompletionstreamrequest.ts

@@ -77,6 +77,9 @@ export type ChatCompletionStreamRequestMessages =
   | (UserMessage & { role: "user" })
   | (AssistantMessage & { role: "assistant" });
 
+/**
+ * Controls which (if any) tool is called by the model. `none` means the model will not call any tool and instead generates a message. `auto` means the model can pick between generating a message or calling one or more tools. `any` or `required` means the model must call one or more tools. Specifying a particular tool via `{"type": "function", "function": {"name": "my_function"}}` forces the model to call that tool.
+ */
 export type ChatCompletionStreamRequestToolChoice = ToolChoice | ToolChoiceEnum;
 
 export type ChatCompletionStreamRequest = {
@@ -105,6 +108,7 @@ export type ChatCompletionStreamRequest = {
    * The seed to use for random sampling. If set, different calls will generate deterministic results.
    */
   randomSeed?: number | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
   /**
    * The prompt(s) to generate completions for, encoded as a list of dict with role and content.
    */
@@ -114,22 +118,37 @@ export type ChatCompletionStreamRequest = {
     | (UserMessage & { role: "user" })
     | (AssistantMessage & { role: "assistant" })
   >;
+  /**
+   * Specify the format that the model must output. By default it will use `{ "type": "text" }`. Setting to `{ "type": "json_object" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ "type": "json_schema" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide.
+   */
   responseFormat?: ResponseFormat | undefined;
+  /**
+   * A list of tools the model may call. Use this to provide a list of functions the model may generate JSON inputs for.
+   */
   tools?: Array<Tool> | null | undefined;
+  /**
+   * Controls which (if any) tool is called by the model. `none` means the model will not call any tool and instead generates a message. `auto` means the model can pick between generating a message or calling one or more tools. `any` or `required` means the model must call one or more tools. Specifying a particular tool via `{"type": "function", "function": {"name": "my_function"}}` forces the model to call that tool.
+   */
   toolChoice?: ToolChoice | ToolChoiceEnum | undefined;
   /**
-   * presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
+   * The `presence_penalty` determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
    */
   presencePenalty?: number | undefined;
   /**
-   * frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
+   * The `frequency_penalty` penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
    */
   frequencyPenalty?: number | undefined;
   /**
    * Number of completions to return for each request, input tokens are only billed once.
    */
   n?: number | null | undefined;
+  /**
+   * Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content.
+   */
   prediction?: Prediction | undefined;
+  /**
+   * Whether to enable parallel function calling during tool use, when enabled the model can call multiple tools in parallel.
+   */
   parallelToolCalls?: boolean | undefined;
   /**
    * Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used.
@@ -350,6 +369,7 @@ export const ChatCompletionStreamRequest$inboundSchema: z.ZodType<
   stream: z.boolean().default(true),
   stop: z.union([z.string(), z.array(z.string())]).optional(),
   random_seed: z.nullable(z.number().int()).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
   messages: z.array(
     z.union([
       SystemMessage$inboundSchema.and(
@@ -409,6 +429,7 @@ export type ChatCompletionStreamRequest$Outbound = {
   stream: boolean;
   stop?: string | Array<string> | undefined;
   random_seed?: number | null | undefined;
+  metadata?: { [k: string]: any } | null | undefined;
   messages: Array<
     | (SystemMessage$Outbound & { role: "system" })
     | (ToolMessage$Outbound & { role: "tool" })
@@ -440,6 +461,7 @@ export const ChatCompletionStreamRequest$outboundSchema: z.ZodType<
   stream: z.boolean().default(true),
   stop: z.union([z.string(), z.array(z.string())]).optional(),
   randomSeed: z.nullable(z.number().int()).optional(),
+  metadata: z.nullable(z.record(z.any())).optional(),
   messages: z.array(
     z.union([
       SystemMessage$outboundSchema.and(