@meetkai/mka1 0.51.11 → 0.51.12

package/src/models/components/deployment.ts

@@ -9,11 +9,11 @@ import { ClosedEnum } from "../../types/enums.js";
 import { Result as SafeParseResult } from "../../types/fp.js";
 import { SDKValidationError } from "../errors/sdkvalidationerror.js";
 import {
-  AcceleratorSpec,
-  AcceleratorSpec$inboundSchema,
-  AcceleratorSpec$Outbound,
-  AcceleratorSpec$outboundSchema,
-} from "./acceleratorspec.js";
+  DeploymentAcceleratorSpec,
+  DeploymentAcceleratorSpec$inboundSchema,
+  DeploymentAcceleratorSpec$Outbound,
+  DeploymentAcceleratorSpec$outboundSchema,
+} from "./deploymentacceleratorspec.js";
 import {
   DeploymentEndpoint,
   DeploymentEndpoint$inboundSchema,
@@ -50,13 +50,13 @@ export type DeploymentStatus = ClosedEnum<typeof DeploymentStatus>;
  */
 export type Deployment = {
   /**
-   * A hardware accelerator request — GPU, NPU, TPU, or similar.
+   * A hardware accelerator request for a deployment — GPU, NPU, TPU, or cpu-class serving.
    *
    * @remarks
    *
-   * Request an accelerator by type and count, with optional fallback types.
+   * Request an accelerator by type and count, with optional fallback types. Set class to cpu to deploy a cpu-capable managed recipe on CPU instances (type is then not required; the CPU resource floor comes from the recipe).
    */
-  accelerator: AcceleratorSpec;
+  accelerator: DeploymentAcceleratorSpec;
   createdAt: Date;
   /**
    * Where and how to call the model; set while registered with the gateway
@@ -101,7 +101,7 @@ export const Deployment$inboundSchema: z.ZodType<
   z.ZodTypeDef,
   unknown
 > = z.object({
-  accelerator: AcceleratorSpec$inboundSchema,
+  accelerator: DeploymentAcceleratorSpec$inboundSchema,
   created_at: z.string().datetime({ offset: true }).transform(v => new Date(v)),
   endpoint: z.nullable(DeploymentEndpoint$inboundSchema).optional(),
   engine: Engine$inboundSchema,
@@ -123,7 +123,7 @@ export const Deployment$inboundSchema: z.ZodType<
 });
 /** @internal */
 export type Deployment$Outbound = {
-  accelerator: AcceleratorSpec$Outbound;
+  accelerator: DeploymentAcceleratorSpec$Outbound;
   created_at: string;
   endpoint?: DeploymentEndpoint$Outbound | null | undefined;
   engine: string;
@@ -144,7 +144,7 @@ export const Deployment$outboundSchema: z.ZodType<
   z.ZodTypeDef,
   Deployment
 > = z.object({
-  accelerator: AcceleratorSpec$outboundSchema,
+  accelerator: DeploymentAcceleratorSpec$outboundSchema,
   createdAt: z.date().transform(v => v.toISOString()),
   endpoint: z.nullable(DeploymentEndpoint$outboundSchema).optional(),
   engine: Engine$outboundSchema,

package/src/models/components/deploymentacceleratorspec.ts

@@ -0,0 +1,100 @@
+/*
+ * Code generated by Speakeasy (https://speakeasy.com). DO NOT EDIT.
+ */
+
+import * as z from "zod/v3";
+import { safeParse } from "../../lib/schemas.js";
+import { ClosedEnum } from "../../types/enums.js";
+import { Result as SafeParseResult } from "../../types/fp.js";
+import { SDKValidationError } from "../errors/sdkvalidationerror.js";
+
+/**
+ * Accelerator class: gpu (default) or cpu. For cpu-only recipes this is inferred; type is not required for cpu.
+ */
+export const Class = {
+  Gpu: "gpu",
+  Cpu: "cpu",
+} as const;
+/**
+ * Accelerator class: gpu (default) or cpu. For cpu-only recipes this is inferred; type is not required for cpu.
+ */
+export type Class = ClosedEnum<typeof Class>;
+
+/**
+ * A hardware accelerator request for a deployment — GPU, NPU, TPU, or cpu-class serving.
+ *
+ * @remarks
+ *
+ * Request an accelerator by type and count, with optional fallback types. Set class to cpu to deploy a cpu-capable managed recipe on CPU instances (type is then not required; the CPU resource floor comes from the recipe).
+ */
+export type DeploymentAcceleratorSpec = {
+  /**
+   * Accelerator class: gpu (default) or cpu. For cpu-only recipes this is inferred; type is not required for cpu.
+   */
+  class?: Class | undefined;
+  /**
+   * Accelerators per replica
+   */
+  count?: number | undefined;
+  /**
+   * Ordered fallback types if the primary is unavailable
+   */
+  fallback?: Array<string> | undefined;
+  type?: string | undefined;
+};
+
+/** @internal */
+export const Class$inboundSchema: z.ZodNativeEnum<typeof Class> = z.nativeEnum(
+  Class,
+);
+/** @internal */
+export const Class$outboundSchema: z.ZodNativeEnum<typeof Class> =
+  Class$inboundSchema;
+
+/** @internal */
+export const DeploymentAcceleratorSpec$inboundSchema: z.ZodType<
+  DeploymentAcceleratorSpec,
+  z.ZodTypeDef,
+  unknown
+> = z.object({
+  class: Class$inboundSchema.default("gpu"),
+  count: z.number().int().default(1),
+  fallback: z.array(z.string()).optional(),
+  type: z.string().optional(),
+});
+/** @internal */
+export type DeploymentAcceleratorSpec$Outbound = {
+  class: string;
+  count: number;
+  fallback?: Array<string> | undefined;
+  type?: string | undefined;
+};
+
+/** @internal */
+export const DeploymentAcceleratorSpec$outboundSchema: z.ZodType<
+  DeploymentAcceleratorSpec$Outbound,
+  z.ZodTypeDef,
+  DeploymentAcceleratorSpec
+> = z.object({
+  class: Class$outboundSchema.default("gpu"),
+  count: z.number().int().default(1),
+  fallback: z.array(z.string()).optional(),
+  type: z.string().optional(),
+});
+
+export function deploymentAcceleratorSpecToJSON(
+  deploymentAcceleratorSpec: DeploymentAcceleratorSpec,
+): string {
+  return JSON.stringify(
+    DeploymentAcceleratorSpec$outboundSchema.parse(deploymentAcceleratorSpec),
+  );
+}
+export function deploymentAcceleratorSpecFromJSON(
+  jsonString: string,
+): SafeParseResult<DeploymentAcceleratorSpec, SDKValidationError> {
+  return safeParse(
+    jsonString,
+    (x) => DeploymentAcceleratorSpec$inboundSchema.parse(JSON.parse(x)),
+    `Failed to parse 'DeploymentAcceleratorSpec' from JSON`,
+  );
+}

package/src/models/components/deploymentcreate.ts

@@ -8,11 +8,11 @@ import { safeParse } from "../../lib/schemas.js";
 import { Result as SafeParseResult } from "../../types/fp.js";
 import { SDKValidationError } from "../errors/sdkvalidationerror.js";
 import {
-  AcceleratorSpec,
-  AcceleratorSpec$inboundSchema,
-  AcceleratorSpec$Outbound,
-  AcceleratorSpec$outboundSchema,
-} from "./acceleratorspec.js";
+  DeploymentAcceleratorSpec,
+  DeploymentAcceleratorSpec$inboundSchema,
+  DeploymentAcceleratorSpec$Outbound,
+  DeploymentAcceleratorSpec$outboundSchema,
+} from "./deploymentacceleratorspec.js";
 import {
   DeploymentAPIFormat,
   DeploymentAPIFormat$inboundSchema,
@@ -40,13 +40,13 @@ import {
  */
 export type DeploymentCreate = {
   /**
-   * A hardware accelerator request — GPU, NPU, TPU, or similar.
+   * A hardware accelerator request for a deployment — GPU, NPU, TPU, or cpu-class serving.
    *
    * @remarks
    *
-   * Request an accelerator by type and count, with optional fallback types.
+   * Request an accelerator by type and count, with optional fallback types. Set class to cpu to deploy a cpu-capable managed recipe on CPU instances (type is then not required; the CPU resource floor comes from the recipe).
    */
-  accelerator: AcceleratorSpec;
+  accelerator: DeploymentAcceleratorSpec;
   /**
    * Gateway inference surface exposed by the deployment runtime.
    */
@@ -112,7 +112,7 @@ export const DeploymentCreate$inboundSchema: z.ZodType<
   z.ZodTypeDef,
   unknown
 > = z.object({
-  accelerator: AcceleratorSpec$inboundSchema,
+  accelerator: DeploymentAcceleratorSpec$inboundSchema,
   api_format: DeploymentAPIFormat$inboundSchema.optional(),
   endpoint_auth: z.boolean().default(true),
   engine: Engine$inboundSchema.optional(),
@@ -139,7 +139,7 @@ export const DeploymentCreate$inboundSchema: z.ZodType<
 });
 /** @internal */
 export type DeploymentCreate$Outbound = {
-  accelerator: AcceleratorSpec$Outbound;
+  accelerator: DeploymentAcceleratorSpec$Outbound;
   api_format?: string | undefined;
   endpoint_auth: boolean;
   engine?: string | undefined;
@@ -163,7 +163,7 @@ export const DeploymentCreate$outboundSchema: z.ZodType<
   z.ZodTypeDef,
   DeploymentCreate
 > = z.object({
-  accelerator: AcceleratorSpec$outboundSchema,
+  accelerator: DeploymentAcceleratorSpec$outboundSchema,
   apiFormat: DeploymentAPIFormat$outboundSchema.optional(),
   endpointAuth: z.boolean().default(true),
   engine: Engine$outboundSchema.optional(),

package/src/models/components/index.ts

@@ -135,6 +135,7 @@ export * from "./deleteorgautomodelresponse.js";
 export * from "./deletepromptresponse.js";
 export * from "./deleteresponseobject.js";
 export * from "./deployment.js";
+export * from "./deploymentacceleratorspec.js";
 export * from "./deploymentapiformat.js";
 export * from "./deploymentcreate.js";
 export * from "./deploymentendpoint.js";

package/src/models/components/replica.ts

@@ -8,24 +8,24 @@ import { safeParse } from "../../lib/schemas.js";
 import { Result as SafeParseResult } from "../../types/fp.js";
 import { SDKValidationError } from "../errors/sdkvalidationerror.js";
 import {
-  AcceleratorSpec,
-  AcceleratorSpec$inboundSchema,
-  AcceleratorSpec$Outbound,
-  AcceleratorSpec$outboundSchema,
-} from "./acceleratorspec.js";
+  DeploymentAcceleratorSpec,
+  DeploymentAcceleratorSpec$inboundSchema,
+  DeploymentAcceleratorSpec$Outbound,
+  DeploymentAcceleratorSpec$outboundSchema,
+} from "./deploymentacceleratorspec.js";
 
 /**
  * A single running replica (pod/container) behind a deployment.
  */
 export type Replica = {
   /**
-   * A hardware accelerator request — GPU, NPU, TPU, or similar.
+   * A hardware accelerator request for a deployment — GPU, NPU, TPU, or cpu-class serving.
    *
    * @remarks
    *
-   * Request an accelerator by type and count, with optional fallback types.
+   * Request an accelerator by type and count, with optional fallback types. Set class to cpu to deploy a cpu-capable managed recipe on CPU instances (type is then not required; the CPU resource floor comes from the recipe).
    */
-  accelerator: AcceleratorSpec;
+  accelerator: DeploymentAcceleratorSpec;
   id: string;
   node?: string | null | undefined;
   priceUsdHr?: number | undefined;
@@ -41,7 +41,7 @@ export type Replica = {
 /** @internal */
 export const Replica$inboundSchema: z.ZodType<Replica, z.ZodTypeDef, unknown> =
   z.object({
-    accelerator: AcceleratorSpec$inboundSchema,
+    accelerator: DeploymentAcceleratorSpec$inboundSchema,
     id: z.string(),
     node: z.nullable(z.string()).optional(),
     price_usd_hr: z.number().optional(),
@@ -59,7 +59,7 @@ export const Replica$inboundSchema: z.ZodType<Replica, z.ZodTypeDef, unknown> =
   });
 /** @internal */
 export type Replica$Outbound = {
-  accelerator: AcceleratorSpec$Outbound;
+  accelerator: DeploymentAcceleratorSpec$Outbound;
   id: string;
   node?: string | null | undefined;
   price_usd_hr?: number | undefined;
@@ -75,7 +75,7 @@ export const Replica$outboundSchema: z.ZodType<
   z.ZodTypeDef,
   Replica
 > = z.object({
-  accelerator: AcceleratorSpec$outboundSchema,
+  accelerator: DeploymentAcceleratorSpec$outboundSchema,
   id: z.string(),
   node: z.nullable(z.string()).optional(),
   priceUsdHr: z.number().optional(),