@meetkai/mka1 0.51.21 → 0.51.23

package/src/mcp-server/server.ts

@@ -300,64 +300,27 @@ import { tool$sandboxSetPricing } from "./tools/sandboxSetPricing.js";
 import { tool$sandboxTerminate } from "./tools/sandboxTerminate.js";
 import { tool$sandboxUploadArchive } from "./tools/sandboxUploadArchive.js";
 import { tool$sandboxUploadFile } from "./tools/sandboxUploadFile.js";
-import { tool$servingAcceleratorsAvailability } from "./tools/servingAcceleratorsAvailability.js";
 import { tool$servingAcceleratorsGet } from "./tools/servingAcceleratorsGet.js";
 import { tool$servingAcceleratorsList } from "./tools/servingAcceleratorsList.js";
-import { tool$servingDatasetsCommit } from "./tools/servingDatasetsCommit.js";
-import { tool$servingDatasetsCreate } from "./tools/servingDatasetsCreate.js";
-import { tool$servingDatasetsDelete } from "./tools/servingDatasetsDelete.js";
-import { tool$servingDatasetsGet } from "./tools/servingDatasetsGet.js";
-import { tool$servingDatasetsList } from "./tools/servingDatasetsList.js";
-import { tool$servingDatasetsUploadContent } from "./tools/servingDatasetsUploadContent.js";
 import { tool$servingDeploymentsCreate } from "./tools/servingDeploymentsCreate.js";
 import { tool$servingDeploymentsDelete } from "./tools/servingDeploymentsDelete.js";
-import { tool$servingDeploymentsDeregister } from "./tools/servingDeploymentsDeregister.js";
 import { tool$servingDeploymentsGet } from "./tools/servingDeploymentsGet.js";
 import { tool$servingDeploymentsGetLogs } from "./tools/servingDeploymentsGetLogs.js";
-import { tool$servingDeploymentsGetMetrics } from "./tools/servingDeploymentsGetMetrics.js";
 import { tool$servingDeploymentsList } from "./tools/servingDeploymentsList.js";
-import { tool$servingDeploymentsListReplicas } from "./tools/servingDeploymentsListReplicas.js";
-import { tool$servingDeploymentsListRevisions } from "./tools/servingDeploymentsListRevisions.js";
-import { tool$servingDeploymentsRegister } from "./tools/servingDeploymentsRegister.js";
-import { tool$servingDeploymentsRollback } from "./tools/servingDeploymentsRollback.js";
-import { tool$servingDeploymentsUpdate } from "./tools/servingDeploymentsUpdate.js";
-import { tool$servingDeploymentsUpdateScaling } from "./tools/servingDeploymentsUpdateScaling.js";
-import { tool$servingEligibilityPreview } from "./tools/servingEligibilityPreview.js";
 import { tool$servingFinetuneJobsCancel } from "./tools/servingFinetuneJobsCancel.js";
 import { tool$servingFinetuneJobsCreate } from "./tools/servingFinetuneJobsCreate.js";
 import { tool$servingFinetuneJobsGet } from "./tools/servingFinetuneJobsGet.js";
 import { tool$servingFinetuneJobsGetEvents } from "./tools/servingFinetuneJobsGetEvents.js";
 import { tool$servingFinetuneJobsGetLogs } from "./tools/servingFinetuneJobsGetLogs.js";
-import { tool$servingFinetuneJobsGetMetrics } from "./tools/servingFinetuneJobsGetMetrics.js";
 import { tool$servingFinetuneJobsList } from "./tools/servingFinetuneJobsList.js";
-import { tool$servingFinetuneJobsListCheckpoints } from "./tools/servingFinetuneJobsListCheckpoints.js";
-import { tool$servingImagesDelete } from "./tools/servingImagesDelete.js";
-import { tool$servingImagesGet } from "./tools/servingImagesGet.js";
-import { tool$servingImagesList } from "./tools/servingImagesList.js";
-import { tool$servingImagesRegister } from "./tools/servingImagesRegister.js";
-import { tool$servingModelsDelete } from "./tools/servingModelsDelete.js";
-import { tool$servingModelsGet } from "./tools/servingModelsGet.js";
-import { tool$servingModelsList } from "./tools/servingModelsList.js";
-import { tool$servingModelsRegister } from "./tools/servingModelsRegister.js";
-import { tool$servingRecipesDelete } from "./tools/servingRecipesDelete.js";
-import { tool$servingRecipesGet } from "./tools/servingRecipesGet.js";
-import { tool$servingRecipesList } from "./tools/servingRecipesList.js";
-import { tool$servingRecipesRegister } from "./tools/servingRecipesRegister.js";
-import { tool$servingRecipesSetEnabled } from "./tools/servingRecipesSetEnabled.js";
 import { tool$servingSecretsCreate } from "./tools/servingSecretsCreate.js";
 import { tool$servingSecretsDelete } from "./tools/servingSecretsDelete.js";
 import { tool$servingSecretsGet } from "./tools/servingSecretsGet.js";
 import { tool$servingSecretsList } from "./tools/servingSecretsList.js";
 import { tool$servingSpendGet } from "./tools/servingSpendGet.js";
-import { tool$servingTenantsDeleteOrganization } from "./tools/servingTenantsDeleteOrganization.js";
 import { tool$servingTenantsGet } from "./tools/servingTenantsGet.js";
-import { tool$servingTenantsGetOrganizationDeletion } from "./tools/servingTenantsGetOrganizationDeletion.js";
 import { tool$servingTenantsList } from "./tools/servingTenantsList.js";
 import { tool$servingTenantsSet } from "./tools/servingTenantsSet.js";
-import { tool$servingVolumesCreate } from "./tools/servingVolumesCreate.js";
-import { tool$servingVolumesDelete } from "./tools/servingVolumesDelete.js";
-import { tool$servingVolumesGet } from "./tools/servingVolumesGet.js";
-import { tool$servingVolumesList } from "./tools/servingVolumesList.js";
 import { tool$usageCosts } from "./tools/usageCosts.js";
 import { tool$usageSandbox } from "./tools/usageSandbox.js";
 
@@ -371,7 +334,7 @@ export function createMCPServer(deps: {
 }) {
   const server = new McpServer({
     name: "SDK",
-    version: "0.51.21",
+    version: "0.51.23",
   });
 
   const client = new SDKCore({
@@ -692,62 +655,25 @@ export function createMCPServer(deps: {
   tool(tool$authApiKeyGetJwtFromKey);
   tool(tool$servingAcceleratorsList);
   tool(tool$servingAcceleratorsGet);
-  tool(tool$servingAcceleratorsAvailability);
-  tool(tool$servingDatasetsList);
-  tool(tool$servingDatasetsCreate);
-  tool(tool$servingDatasetsDelete);
-  tool(tool$servingDatasetsGet);
-  tool(tool$servingDatasetsCommit);
-  tool(tool$servingDatasetsUploadContent);
   tool(tool$servingDeploymentsList);
   tool(tool$servingDeploymentsCreate);
   tool(tool$servingDeploymentsDelete);
   tool(tool$servingDeploymentsGet);
-  tool(tool$servingDeploymentsUpdate);
   tool(tool$servingDeploymentsGetLogs);
-  tool(tool$servingDeploymentsGetMetrics);
-  tool(tool$servingDeploymentsDeregister);
-  tool(tool$servingDeploymentsRegister);
-  tool(tool$servingDeploymentsListReplicas);
-  tool(tool$servingDeploymentsListRevisions);
-  tool(tool$servingDeploymentsRollback);
-  tool(tool$servingDeploymentsUpdateScaling);
-  tool(tool$servingEligibilityPreview);
   tool(tool$servingFinetuneJobsList);
   tool(tool$servingFinetuneJobsCreate);
   tool(tool$servingFinetuneJobsGet);
   tool(tool$servingFinetuneJobsCancel);
-  tool(tool$servingFinetuneJobsListCheckpoints);
   tool(tool$servingFinetuneJobsGetEvents);
   tool(tool$servingFinetuneJobsGetLogs);
-  tool(tool$servingFinetuneJobsGetMetrics);
-  tool(tool$servingImagesList);
-  tool(tool$servingImagesRegister);
-  tool(tool$servingImagesDelete);
-  tool(tool$servingImagesGet);
-  tool(tool$servingModelsList);
-  tool(tool$servingModelsRegister);
-  tool(tool$servingModelsDelete);
-  tool(tool$servingModelsGet);
-  tool(tool$servingRecipesList);
-  tool(tool$servingRecipesDelete);
-  tool(tool$servingRecipesGet);
-  tool(tool$servingRecipesSetEnabled);
-  tool(tool$servingRecipesRegister);
   tool(tool$servingSecretsList);
   tool(tool$servingSecretsCreate);
   tool(tool$servingSecretsDelete);
   tool(tool$servingSecretsGet);
-  tool(tool$servingSpendGet);
   tool(tool$servingTenantsList);
   tool(tool$servingTenantsGet);
   tool(tool$servingTenantsSet);
-  tool(tool$servingTenantsGetOrganizationDeletion);
-  tool(tool$servingTenantsDeleteOrganization);
-  tool(tool$servingVolumesList);
-  tool(tool$servingVolumesCreate);
-  tool(tool$servingVolumesDelete);
-  tool(tool$servingVolumesGet);
+  tool(tool$servingSpendGet);
 
   return server;
 }

package/src/mcp-server/tools/servingSpendGet.ts

@@ -7,7 +7,7 @@ import * as operations from "../../models/operations/index.js";
 import { formatResult, ToolDefinition } from "../tools.js";
 
 const args = {
-  request: operations.GetSpendRequest$inboundSchema,
+  request: operations.GetServingUsageRequest$inboundSchema,
 };
 
 export const tool$servingSpendGet: ToolDefinition<typeof args> = {

package/src/models/components/acceleratortype.ts

@@ -27,14 +27,14 @@ import {
 } from "./acquirability.js";
 
 /**
- * Coarse accelerator class. npu/tpu are reserved for future catalog entries.
+ * Coarse accelerator class. npu/tpu are reserved for future catalog entries. No cpu-class type is listed: CPU is a class, not an accelerator type, and cpu-class serving is requested through the deployment accelerator spec's class field rather than by naming a type here.
  */
 export const Class = {
   Gpu: "gpu",
   Cpu: "cpu",
 } as const;
 /**
- * Coarse accelerator class. npu/tpu are reserved for future catalog entries.
+ * Coarse accelerator class. npu/tpu are reserved for future catalog entries. No cpu-class type is listed: CPU is a class, not an accelerator type, and cpu-class serving is requested through the deployment accelerator spec's class field rather than by naming a type here.
  */
 export type Class = ClosedEnum<typeof Class>;
 
@@ -75,7 +75,7 @@ export type AcceleratorType = {
    */
   acquirableCount?: number | undefined;
   /**
-   * Coarse accelerator class. npu/tpu are reserved for future catalog entries.
+   * Coarse accelerator class. npu/tpu are reserved for future catalog entries. No cpu-class type is listed: CPU is a class, not an accelerator type, and cpu-class serving is requested through the deployment accelerator spec's class field rather than by naming a type here.
    */
   class: Class;
   /**
@@ -85,11 +85,11 @@ export type AcceleratorType = {
    */
   kind?: Kind | undefined;
   /**
-   * Max accelerators of this type per replica. For class cpu the unit is vCPUs: the maximum rentable vCPU count per replica.
+   * Max accelerators of this type per replica.
    */
   maxCount: number;
   /**
-   * Per-accelerator memory in GB. Absent for class cpu: CPU capacity is sized per request (min_vcpus/min_mem_gb), not by a fixed module.
+   * Per-accelerator memory in GB. Absent for cpu-class entries, which have no fixed memory module.
    */
   memoryGb?: number | undefined;
   name: string;

package/src/models/components/deployment.ts

@@ -9,99 +9,60 @@ import { ClosedEnum } from "../../types/enums.js";
 import { Result as SafeParseResult } from "../../types/fp.js";
 import { SDKValidationError } from "../errors/sdkvalidationerror.js";
 import {
-  DeploymentAcceleratorSpec,
-  DeploymentAcceleratorSpec$inboundSchema,
-  DeploymentAcceleratorSpec$Outbound,
-  DeploymentAcceleratorSpec$outboundSchema,
-} from "./deploymentacceleratorspec.js";
-import {
-  DeploymentEndpoint,
-  DeploymentEndpoint$inboundSchema,
-  DeploymentEndpoint$Outbound,
-  DeploymentEndpoint$outboundSchema,
-} from "./deploymentendpoint.js";
-import {
-  Engine,
-  Engine$inboundSchema,
-  Engine$outboundSchema,
-} from "./engine.js";
-import {
-  Scaling,
-  Scaling$inboundSchema,
-  Scaling$Outbound,
-  Scaling$outboundSchema,
-} from "./scaling.js";
+  RunSSH,
+  RunSSH$inboundSchema,
+  RunSSH$Outbound,
+  RunSSH$outboundSchema,
+} from "./runssh.js";
 
-export const DeploymentStatus = {
-  Pending: "pending",
+export const State = {
+  Requested: "requested",
   Provisioning: "provisioning",
-  Ready: "ready",
-  Scaling: "scaling",
-  Updating: "updating",
-  Degraded: "degraded",
+  Running: "running",
+  Succeeded: "succeeded",
   Failed: "failed",
-  Stopped: "stopped",
-  Deleted: "deleted",
+  Terminated: "terminated",
 } as const;
-export type DeploymentStatus = ClosedEnum<typeof DeploymentStatus>;
+export type State = ClosedEnum<typeof State>;
 
-/**
- * A deployed inference server.
- */
 export type Deployment = {
-  /**
-   * 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).
-   */
-  accelerator: DeploymentAcceleratorSpec;
+  accelerator?: string | undefined;
   createdAt: Date;
+  deadlineAt?: Date | undefined;
   /**
-   * Where and how to call the model; set while registered with the gateway
+   * Caller-facing URL. Its shape and transport differ by provider: one proxies over HTTPS, the others hand back IP:port over plain HTTP.
    */
-  endpoint?: DeploymentEndpoint | null | undefined;
+  endpoint?: string | undefined;
   /**
-   * Inference engine that backs a deployment.
+   * The workload's exit status. Provider status is not trusted for this: one provider reports the VM rather than the container.
    */
-  engine: Engine;
+  exitCode?: number | undefined;
+  gpuCount?: number | undefined;
   id: string;
-  image: string | null;
-  model: string;
-  name: string;
-  /**
-   * Free-vocabulary refinement of status explaining why the current status persists (e.g. `awaiting_scheduler`, `tenant_blocked`, `budget_blocked`, `acquiring_capacity`, `rolling_update`, `draining`). Additive: values may appear, change, or disappear at any time, so clients must render it verbatim and must not switch on it — branch on `status` only. Absent when the status needs no explanation.
-   */
-  phase?: string | null | undefined;
-  /**
-   * Human-readable, provider-neutral sentence expanding on `phase` for direct display. Not stable, not machine-parseable; absent whenever `phase` is absent.
-   */
-  phaseMessage?: string | null | undefined;
-  /**
-   * Current config revision
-   */
-  revision: number;
+  image?: string | undefined;
+  name?: string | undefined;
+  priceUsdHr?: number | undefined;
+  provider?: string | undefined;
   /**
-   * Autoscaling configuration for a deployment.
+   * Why the run is in this state, e.g. ssh_setup_failed or budget_exhausted.
    */
-  scaling: Scaling;
+  reason?: string | undefined;
   /**
-   * Model ID consumers pass to the gateway's inference endpoints; set while the deployment is registered with the gateway (see Register), null when unregistered
+   * Connection details. The login user is provider dependent and is read from the provider, never assumed.
    */
-  servedModelId?: string | null | undefined;
-  status: DeploymentStatus;
-  updatedAt: Date;
+  ssh?: RunSSH | undefined;
+  startedAt?: Date | undefined;
+  state: State;
+  terminalAt?: Date | undefined;
 };
 
 /** @internal */
-export const DeploymentStatus$inboundSchema: z.ZodNativeEnum<
-  typeof DeploymentStatus
-> = z.nativeEnum(DeploymentStatus);
+export const State$inboundSchema: z.ZodNativeEnum<typeof State> = z.nativeEnum(
+  State,
+);
 /** @internal */
-export const DeploymentStatus$outboundSchema: z.ZodNativeEnum<
-  typeof DeploymentStatus
-> = DeploymentStatus$inboundSchema;
+export const State$outboundSchema: z.ZodNativeEnum<typeof State> =
+  State$inboundSchema;
 
 /** @internal */
 export const Deployment$inboundSchema: z.ZodType<
@@ -109,46 +70,54 @@ export const Deployment$inboundSchema: z.ZodType<
   z.ZodTypeDef,
   unknown
 > = z.object({
-  accelerator: DeploymentAcceleratorSpec$inboundSchema,
+  accelerator: z.string().optional(),
   created_at: z.string().datetime({ offset: true }).transform(v => new Date(v)),
-  endpoint: z.nullable(DeploymentEndpoint$inboundSchema).optional(),
-  engine: Engine$inboundSchema,
+  deadline_at: z.string().datetime({ offset: true }).transform(v => new Date(v))
+    .optional(),
+  endpoint: z.string().optional(),
+  exit_code: z.number().int().optional(),
+  gpu_count: z.number().int().optional(),
   id: z.string(),
-  image: z.nullable(z.string()),
-  model: z.string(),
-  name: z.string(),
-  phase: z.nullable(z.string()).optional(),
-  phase_message: z.nullable(z.string()).optional(),
-  revision: z.number().int(),
-  scaling: Scaling$inboundSchema,
-  served_model_id: z.nullable(z.string()).optional(),
-  status: DeploymentStatus$inboundSchema,
-  updated_at: z.string().datetime({ offset: true }).transform(v => new Date(v)),
+  image: z.string().optional(),
+  name: z.string().optional(),
+  price_usd_hr: z.number().optional(),
+  provider: z.string().optional(),
+  reason: z.string().optional(),
+  ssh: RunSSH$inboundSchema.optional(),
+  started_at: z.string().datetime({ offset: true }).transform(v => new Date(v))
+    .optional(),
+  state: State$inboundSchema,
+  terminal_at: z.string().datetime({ offset: true }).transform(v => new Date(v))
+    .optional(),
 }).transform((v) => {
   return remap$(v, {
     "created_at": "createdAt",
-    "phase_message": "phaseMessage",
-    "served_model_id": "servedModelId",
-    "updated_at": "updatedAt",
+    "deadline_at": "deadlineAt",
+    "exit_code": "exitCode",
+    "gpu_count": "gpuCount",
+    "price_usd_hr": "priceUsdHr",
+    "started_at": "startedAt",
+    "terminal_at": "terminalAt",
   });
 });
 /** @internal */
 export type Deployment$Outbound = {
-  accelerator: DeploymentAcceleratorSpec$Outbound;
+  accelerator?: string | undefined;
   created_at: string;
-  endpoint?: DeploymentEndpoint$Outbound | null | undefined;
-  engine: string;
+  deadline_at?: string | undefined;
+  endpoint?: string | undefined;
+  exit_code?: number | undefined;
+  gpu_count?: number | undefined;
   id: string;
-  image: string | null;
-  model: string;
-  name: string;
-  phase?: string | null | undefined;
-  phase_message?: string | null | undefined;
-  revision: number;
-  scaling: Scaling$Outbound;
-  served_model_id?: string | null | undefined;
-  status: string;
-  updated_at: string;
+  image?: string | undefined;
+  name?: string | undefined;
+  price_usd_hr?: number | undefined;
+  provider?: string | undefined;
+  reason?: string | undefined;
+  ssh?: RunSSH$Outbound | undefined;
+  started_at?: string | undefined;
+  state: string;
+  terminal_at?: string | undefined;
 };
 
 /** @internal */
@@ -157,27 +126,31 @@ export const Deployment$outboundSchema: z.ZodType<
   z.ZodTypeDef,
   Deployment
 > = z.object({
-  accelerator: DeploymentAcceleratorSpec$outboundSchema,
+  accelerator: z.string().optional(),
   createdAt: z.date().transform(v => v.toISOString()),
-  endpoint: z.nullable(DeploymentEndpoint$outboundSchema).optional(),
-  engine: Engine$outboundSchema,
+  deadlineAt: z.date().transform(v => v.toISOString()).optional(),
+  endpoint: z.string().optional(),
+  exitCode: z.number().int().optional(),
+  gpuCount: z.number().int().optional(),
   id: z.string(),
-  image: z.nullable(z.string()),
-  model: z.string(),
-  name: z.string(),
-  phase: z.nullable(z.string()).optional(),
-  phaseMessage: z.nullable(z.string()).optional(),
-  revision: z.number().int(),
-  scaling: Scaling$outboundSchema,
-  servedModelId: z.nullable(z.string()).optional(),
-  status: DeploymentStatus$outboundSchema,
-  updatedAt: z.date().transform(v => v.toISOString()),
+  image: z.string().optional(),
+  name: z.string().optional(),
+  priceUsdHr: z.number().optional(),
+  provider: z.string().optional(),
+  reason: z.string().optional(),
+  ssh: RunSSH$outboundSchema.optional(),
+  startedAt: z.date().transform(v => v.toISOString()).optional(),
+  state: State$outboundSchema,
+  terminalAt: z.date().transform(v => v.toISOString()).optional(),
 }).transform((v) => {
   return remap$(v, {
     createdAt: "created_at",
-    phaseMessage: "phase_message",
-    servedModelId: "served_model_id",
-    updatedAt: "updated_at",
+    deadlineAt: "deadline_at",
+    exitCode: "exit_code",
+    gpuCount: "gpu_count",
+    priceUsdHr: "price_usd_hr",
+    startedAt: "started_at",
+    terminalAt: "terminal_at",
   });
 });