@gpuai/sdk 0.3.7 → 0.3.9

package/README.md

@@ -1,4 +1,4 @@
-# @gpuai/sdk@0.3.7
+# @gpuai/sdk@0.3.9
 
 A TypeScript SDK client for the api.gpu.ai API.
 
@@ -72,6 +72,7 @@ All URIs are relative to *https://api.gpu.ai/v1*
 *FineTuningApi* | [**getFineTuningJob**](docs/FineTuningApi.md#getfinetuningjob) | **GET** /fine_tuning/jobs/{id} | Retrieve a fine-tuning job (OpenAI-compatible)
 *FineTuningApi* | [**listFineTuningJobEvents**](docs/FineTuningApi.md#listfinetuningjobevents) | **GET** /fine_tuning/jobs/{id}/events | List fine-tuning job events (OpenAI-compatible)
 *FineTuningApi* | [**listFineTuningJobs**](docs/FineTuningApi.md#listfinetuningjobs) | **GET** /fine_tuning/jobs | List fine-tuning jobs (OpenAI-compatible)
+*FineTuningApi* | [**retrieveFile**](docs/FineTuningApi.md#retrievefile) | **GET** /files/{id} | Retrieve a file (OpenAI-compatible)
 *GpuTypesApi* | [**listGpuTypes**](docs/GpuTypesApi.md#listgputypes) | **GET** /gpu-types | List available GPU types (no auth required)
 *InferenceApi* | [**cancelVideo**](docs/InferenceApi.md#cancelvideo) | **POST** /videos/{id}/cancel | Cancel a video generation job
 *InferenceApi* | [**createChatCompletion**](docs/InferenceApi.md#createchatcompletion) | **POST** /chat/completions | Create a chat completion (OpenAI-compatible)
@@ -203,7 +204,7 @@ and is automatically generated by the
 [OpenAPI Generator](https://openapi-generator.tech) project:
 
 - API version: `1.0.0`
-- Package version: `0.3.7`
+- Package version: `0.3.9`
 - Generator version: `7.24.0`
 - Build package: `org.openapitools.codegen.languages.TypeScriptFetchClientCodegen`
 

package/dist/apis/FineTuningApi.d.ts

@@ -37,6 +37,9 @@ export interface ListFineTuningJobsRequest {
     after?: string;
     limit?: number;
 }
+export interface RetrieveFileRequest {
+    id: string;
+}
 /**
  *
  */
@@ -113,6 +116,20 @@ export declare class FineTuningApi extends runtime.BaseAPI {
      * List fine-tuning jobs (OpenAI-compatible)
      */
     listFineTuningJobs(requestParameters?: ListFineTuningJobsRequest, initOverrides?: RequestInit | runtime.InitOverrideFunction): Promise<FineTuningJobList>;
+    /**
+     * Creates request options for retrieveFile without sending the request
+     */
+    retrieveFileRequestOpts(requestParameters: RetrieveFileRequest): Promise<runtime.RequestOpts>;
+    /**
+     * Returns the metadata for one of your uploaded files. The lookup is scoped to the organization the API key belongs to: a file id that does not exist and one that belongs to another organization both return the same 404, so this route cannot be used to probe for other tenants\' ids.  Use it to check that a `training_file` is present and usable BEFORE creating a fine-tuning job — the create path performs the identical ownership check, so a file this route returns is a file that create will accept.
+     * Retrieve a file (OpenAI-compatible)
+     */
+    retrieveFileRaw(requestParameters: RetrieveFileRequest, initOverrides?: RequestInit | runtime.InitOverrideFunction): Promise<runtime.ApiResponse<FileObject>>;
+    /**
+     * Returns the metadata for one of your uploaded files. The lookup is scoped to the organization the API key belongs to: a file id that does not exist and one that belongs to another organization both return the same 404, so this route cannot be used to probe for other tenants\' ids.  Use it to check that a `training_file` is present and usable BEFORE creating a fine-tuning job — the create path performs the identical ownership check, so a file this route returns is a file that create will accept.
+     * Retrieve a file (OpenAI-compatible)
+     */
+    retrieveFile(requestParameters: RetrieveFileRequest, initOverrides?: RequestInit | runtime.InitOverrideFunction): Promise<FileObject>;
 }
 /**
  * @export

package/dist/apis/FineTuningApi.js

@@ -342,6 +342,54 @@ class FineTuningApi extends runtime.BaseAPI {
             return yield response.value();
         });
     }
+    /**
+     * Creates request options for retrieveFile without sending the request
+     */
+    retrieveFileRequestOpts(requestParameters) {
+        return __awaiter(this, void 0, void 0, function* () {
+            if (requestParameters['id'] == null) {
+                throw new runtime.RequiredError('id', 'Required parameter "id" was null or undefined when calling retrieveFile().');
+            }
+            const queryParameters = {};
+            const headerParameters = {};
+            if (this.configuration && this.configuration.accessToken) {
+                const token = this.configuration.accessToken;
+                const tokenString = yield token("bearerAuth", []);
+                if (tokenString) {
+                    headerParameters["Authorization"] = `Bearer ${tokenString}`;
+                }
+            }
+            let urlPath = `/files/{id}`;
+            urlPath = urlPath.replace('{id}', encodeURIComponent(String(requestParameters['id'])));
+            return {
+                path: urlPath,
+                method: 'GET',
+                headers: headerParameters,
+                query: queryParameters,
+            };
+        });
+    }
+    /**
+     * Returns the metadata for one of your uploaded files. The lookup is scoped to the organization the API key belongs to: a file id that does not exist and one that belongs to another organization both return the same 404, so this route cannot be used to probe for other tenants\' ids.  Use it to check that a `training_file` is present and usable BEFORE creating a fine-tuning job — the create path performs the identical ownership check, so a file this route returns is a file that create will accept.
+     * Retrieve a file (OpenAI-compatible)
+     */
+    retrieveFileRaw(requestParameters, initOverrides) {
+        return __awaiter(this, void 0, void 0, function* () {
+            const requestOptions = yield this.retrieveFileRequestOpts(requestParameters);
+            const response = yield this.request(requestOptions, initOverrides);
+            return new runtime.JSONApiResponse(response, (jsonValue) => (0, FileObject_1.FileObjectFromJSON)(jsonValue));
+        });
+    }
+    /**
+     * Returns the metadata for one of your uploaded files. The lookup is scoped to the organization the API key belongs to: a file id that does not exist and one that belongs to another organization both return the same 404, so this route cannot be used to probe for other tenants\' ids.  Use it to check that a `training_file` is present and usable BEFORE creating a fine-tuning job — the create path performs the identical ownership check, so a file this route returns is a file that create will accept.
+     * Retrieve a file (OpenAI-compatible)
+     */
+    retrieveFile(requestParameters, initOverrides) {
+        return __awaiter(this, void 0, void 0, function* () {
+            const response = yield this.retrieveFileRaw(requestParameters, initOverrides);
+            return yield response.value();
+        });
+    }
 }
 exports.FineTuningApi = FineTuningApi;
 /**

package/dist/esm/apis/FineTuningApi.d.ts

@@ -37,6 +37,9 @@ export interface ListFineTuningJobsRequest {
     after?: string;
     limit?: number;
 }
+export interface RetrieveFileRequest {
+    id: string;
+}
 /**
  *
  */
@@ -113,6 +116,20 @@ export declare class FineTuningApi extends runtime.BaseAPI {
      * List fine-tuning jobs (OpenAI-compatible)
      */
     listFineTuningJobs(requestParameters?: ListFineTuningJobsRequest, initOverrides?: RequestInit | runtime.InitOverrideFunction): Promise<FineTuningJobList>;
+    /**
+     * Creates request options for retrieveFile without sending the request
+     */
+    retrieveFileRequestOpts(requestParameters: RetrieveFileRequest): Promise<runtime.RequestOpts>;
+    /**
+     * Returns the metadata for one of your uploaded files. The lookup is scoped to the organization the API key belongs to: a file id that does not exist and one that belongs to another organization both return the same 404, so this route cannot be used to probe for other tenants\' ids.  Use it to check that a `training_file` is present and usable BEFORE creating a fine-tuning job — the create path performs the identical ownership check, so a file this route returns is a file that create will accept.
+     * Retrieve a file (OpenAI-compatible)
+     */
+    retrieveFileRaw(requestParameters: RetrieveFileRequest, initOverrides?: RequestInit | runtime.InitOverrideFunction): Promise<runtime.ApiResponse<FileObject>>;
+    /**
+     * Returns the metadata for one of your uploaded files. The lookup is scoped to the organization the API key belongs to: a file id that does not exist and one that belongs to another organization both return the same 404, so this route cannot be used to probe for other tenants\' ids.  Use it to check that a `training_file` is present and usable BEFORE creating a fine-tuning job — the create path performs the identical ownership check, so a file this route returns is a file that create will accept.
+     * Retrieve a file (OpenAI-compatible)
+     */
+    retrieveFile(requestParameters: RetrieveFileRequest, initOverrides?: RequestInit | runtime.InitOverrideFunction): Promise<FileObject>;
 }
 /**
  * @export

package/dist/esm/apis/FineTuningApi.js

@@ -339,6 +339,54 @@ export class FineTuningApi extends runtime.BaseAPI {
             return yield response.value();
         });
     }
+    /**
+     * Creates request options for retrieveFile without sending the request
+     */
+    retrieveFileRequestOpts(requestParameters) {
+        return __awaiter(this, void 0, void 0, function* () {
+            if (requestParameters['id'] == null) {
+                throw new runtime.RequiredError('id', 'Required parameter "id" was null or undefined when calling retrieveFile().');
+            }
+            const queryParameters = {};
+            const headerParameters = {};
+            if (this.configuration && this.configuration.accessToken) {
+                const token = this.configuration.accessToken;
+                const tokenString = yield token("bearerAuth", []);
+                if (tokenString) {
+                    headerParameters["Authorization"] = `Bearer ${tokenString}`;
+                }
+            }
+            let urlPath = `/files/{id}`;
+            urlPath = urlPath.replace('{id}', encodeURIComponent(String(requestParameters['id'])));
+            return {
+                path: urlPath,
+                method: 'GET',
+                headers: headerParameters,
+                query: queryParameters,
+            };
+        });
+    }
+    /**
+     * Returns the metadata for one of your uploaded files. The lookup is scoped to the organization the API key belongs to: a file id that does not exist and one that belongs to another organization both return the same 404, so this route cannot be used to probe for other tenants\' ids.  Use it to check that a `training_file` is present and usable BEFORE creating a fine-tuning job — the create path performs the identical ownership check, so a file this route returns is a file that create will accept.
+     * Retrieve a file (OpenAI-compatible)
+     */
+    retrieveFileRaw(requestParameters, initOverrides) {
+        return __awaiter(this, void 0, void 0, function* () {
+            const requestOptions = yield this.retrieveFileRequestOpts(requestParameters);
+            const response = yield this.request(requestOptions, initOverrides);
+            return new runtime.JSONApiResponse(response, (jsonValue) => FileObjectFromJSON(jsonValue));
+        });
+    }
+    /**
+     * Returns the metadata for one of your uploaded files. The lookup is scoped to the organization the API key belongs to: a file id that does not exist and one that belongs to another organization both return the same 404, so this route cannot be used to probe for other tenants\' ids.  Use it to check that a `training_file` is present and usable BEFORE creating a fine-tuning job — the create path performs the identical ownership check, so a file this route returns is a file that create will accept.
+     * Retrieve a file (OpenAI-compatible)
+     */
+    retrieveFile(requestParameters, initOverrides) {
+        return __awaiter(this, void 0, void 0, function* () {
+            const response = yield this.retrieveFileRaw(requestParameters, initOverrides);
+            return yield response.value();
+        });
+    }
 }
 /**
  * @export

package/dist/esm/models/CreateInstanceRequest.d.ts

@@ -88,6 +88,12 @@ export interface CreateInstanceRequest {
      * @memberof CreateInstanceRequest
      */
     diskGb?: number;
+    /**
+     * Terminate this instance automatically after this many hours of runtime. Omit for no limit (the default). The countdown starts when the instance reaches `running`, not when you call this — provisioning is neither billed nor counted against the limit — and the resolved deadline is returned as `auto_terminate_at` on every read. Enforcement is server-side and survives your process: this is the safety net for an unattended run whose client dies without cleaning up. When it fires the instance is terminated exactly as a DELETE would terminate it (billing stops, the `instance.terminated` webhook fires) and `status_reason` records that the limit was reached. Bounds: at least 1, at most 720 (30 days) — outside them returns 422 `validation_failed`. An explicit `0` is rejected rather than read as "no limit"; omit the field for that.
+     * @type {number}
+     * @memberof CreateInstanceRequest
+     */
+    autoTerminateHours?: number;
     /**
      * Environment variables. With `template_id`: per-deploy overrides — only keys the template marks user_overridable are accepted (others 422), and secret values are redacted from operation metadata. With `image`: plain container env vars — keys must be POSIX names and the platform's reserved namespace (`PUBLIC_KEY`, `GPUAI_*`) is refused (422).
      * @type {{ [key: string]: string; }}

package/dist/esm/models/CreateInstanceRequest.js

@@ -51,6 +51,7 @@ export function CreateInstanceRequestFromJSONTyped(json, ignoreDiscriminator) {
         'environment': json['environment'] == null ? undefined : json['environment'],
         'offeringId': json['offering_id'] == null ? undefined : json['offering_id'],
         'diskGb': json['disk_gb'] == null ? undefined : json['disk_gb'],
+        'autoTerminateHours': json['auto_terminate_hours'] == null ? undefined : json['auto_terminate_hours'],
         'env': json['env'] == null ? undefined : json['env'],
         'image': json['image'] == null ? undefined : json['image'],
         'registryCredentialId': json['registry_credential_id'] == null ? undefined : json['registry_credential_id'],
@@ -79,6 +80,7 @@ export function CreateInstanceRequestToJSONTyped(value, ignoreDiscriminator = fa
         'environment': value['environment'],
         'offering_id': value['offeringId'],
         'disk_gb': value['diskGb'],
+        'auto_terminate_hours': value['autoTerminateHours'],
         'env': value['env'],
         'image': value['image'],
         'registry_credential_id': value['registryCredentialId'],

package/dist/esm/models/Instance.d.ts

@@ -83,7 +83,19 @@ export interface Instance {
      */
     image?: string;
     /**
-     * Why the instance is in the error state, e.g. "provisioning stalled: the instance started but never became reachable". Present only on errored instances.
+     * The runtime limit this instance was launched with, in hours (the `auto_terminate_hours` you sent). Absent means the instance runs until something stops it. It stays present after the limit has fired, so a terminated instance still explains itself.
+     * @type {number}
+     * @memberof Instance
+     */
+    autoTerminateHours?: number;
+    /**
+     * When the runtime limit expires and the platform terminates this instance. Computed when the instance became running (ready_at + auto_terminate_hours), so the countdown covers usable runtime, not the time spent provisioning. Absent when no limit was set, when the instance is not running yet, or once the platform has begun acting on an elapsed limit.
+     * @type {Date}
+     * @memberof Instance
+     */
+    autoTerminateAt?: Date;
+    /**
+     * Why the instance stopped or failed, e.g. "provisioning stalled: the instance started but never became reachable", or — when a runtime limit fired — "Terminated: this instance reached the 8-hour runtime limit set at launch". Present only when the platform has something to explain.
      * @type {string}
      * @memberof Instance
      */

package/dist/esm/models/Instance.js

@@ -73,6 +73,8 @@ export function InstanceFromJSONTyped(json, ignoreDiscriminator) {
         'connection': json['connection'] == null ? undefined : InstanceConnectionFromJSON(json['connection']),
         'diskGb': json['disk_gb'] == null ? undefined : json['disk_gb'],
         'image': json['image'] == null ? undefined : json['image'],
+        'autoTerminateHours': json['auto_terminate_hours'] == null ? undefined : json['auto_terminate_hours'],
+        'autoTerminateAt': json['auto_terminate_at'] == null ? undefined : (new Date(json['auto_terminate_at'])),
         'statusReason': json['status_reason'] == null ? undefined : json['status_reason'],
         'createdAt': (new Date(json['created_at'])),
         'readyAt': json['ready_at'] == null ? undefined : (new Date(json['ready_at'])),
@@ -100,6 +102,8 @@ export function InstanceToJSONTyped(value, ignoreDiscriminator = false) {
         'connection': InstanceConnectionToJSON(value['connection']),
         'disk_gb': value['diskGb'],
         'image': value['image'],
+        'auto_terminate_hours': value['autoTerminateHours'],
+        'auto_terminate_at': value['autoTerminateAt'] == null ? value['autoTerminateAt'] : value['autoTerminateAt'].toISOString(),
         'status_reason': value['statusReason'],
         'created_at': value['createdAt'].toISOString(),
         'ready_at': value['readyAt'] == null ? value['readyAt'] : value['readyAt'].toISOString(),

package/dist/esm/models/Model.d.ts

@@ -89,6 +89,18 @@ export interface Model {
      * @memberof Model
      */
     aliases?: Array<string>;
+    /**
+     * Whether this model can be used as the base of a managed fine-tuning job. The tunable allowlist is operator-curated and much narrower than the serving catalog, so this is false for almost every model. Absent means false.
+     * @type {boolean}
+     * @memberof Model
+     */
+    fineTunable?: boolean;
+    /**
+     * The exact id to send as `model` when creating a fine-tuning job for this entry. Present only when `fine_tunable` is true, and deliberately distinct from `id`: the allowlist matches the customer-facing base id (e.g. `qwen2.5-7b-instruct`), which a served model normally carries as an alias rather than as its namespaced catalog id.
+     * @type {string}
+     * @memberof Model
+     */
+    fineTuneBaseModel?: string;
     /**
      *
      * @type {ModelStatusEnum}

package/dist/esm/models/Model.js

@@ -75,6 +75,8 @@ export function ModelFromJSONTyped(json, ignoreDiscriminator) {
         'parameters': json['parameters'] == null ? undefined : (json['parameters'].map(ModelParametersInnerFromJSON)),
         'pricing': ModelPricingFromJSON(json['pricing']),
         'aliases': json['aliases'] == null ? undefined : json['aliases'],
+        'fineTunable': json['fine_tunable'] == null ? undefined : json['fine_tunable'],
+        'fineTuneBaseModel': json['fine_tune_base_model'] == null ? undefined : json['fine_tune_base_model'],
         'status': json['status'],
     };
 }
@@ -98,6 +100,8 @@ export function ModelToJSONTyped(value, ignoreDiscriminator = false) {
         'parameters': value['parameters'] == null ? undefined : (value['parameters'].map(ModelParametersInnerToJSON)),
         'pricing': ModelPricingToJSON(value['pricing']),
         'aliases': value['aliases'],
+        'fine_tunable': value['fineTunable'],
+        'fine_tune_base_model': value['fineTuneBaseModel'],
         'status': value['status'],
     };
 }

package/dist/models/CreateInstanceRequest.d.ts

@@ -88,6 +88,12 @@ export interface CreateInstanceRequest {
      * @memberof CreateInstanceRequest
      */
     diskGb?: number;
+    /**
+     * Terminate this instance automatically after this many hours of runtime. Omit for no limit (the default). The countdown starts when the instance reaches `running`, not when you call this — provisioning is neither billed nor counted against the limit — and the resolved deadline is returned as `auto_terminate_at` on every read. Enforcement is server-side and survives your process: this is the safety net for an unattended run whose client dies without cleaning up. When it fires the instance is terminated exactly as a DELETE would terminate it (billing stops, the `instance.terminated` webhook fires) and `status_reason` records that the limit was reached. Bounds: at least 1, at most 720 (30 days) — outside them returns 422 `validation_failed`. An explicit `0` is rejected rather than read as "no limit"; omit the field for that.
+     * @type {number}
+     * @memberof CreateInstanceRequest
+     */
+    autoTerminateHours?: number;
     /**
      * Environment variables. With `template_id`: per-deploy overrides — only keys the template marks user_overridable are accepted (others 422), and secret values are redacted from operation metadata. With `image`: plain container env vars — keys must be POSIX names and the platform's reserved namespace (`PUBLIC_KEY`, `GPUAI_*`) is refused (422).
      * @type {{ [key: string]: string; }}

package/dist/models/CreateInstanceRequest.js

@@ -59,6 +59,7 @@ function CreateInstanceRequestFromJSONTyped(json, ignoreDiscriminator) {
         'environment': json['environment'] == null ? undefined : json['environment'],
         'offeringId': json['offering_id'] == null ? undefined : json['offering_id'],
         'diskGb': json['disk_gb'] == null ? undefined : json['disk_gb'],
+        'autoTerminateHours': json['auto_terminate_hours'] == null ? undefined : json['auto_terminate_hours'],
         'env': json['env'] == null ? undefined : json['env'],
         'image': json['image'] == null ? undefined : json['image'],
         'registryCredentialId': json['registry_credential_id'] == null ? undefined : json['registry_credential_id'],
@@ -87,6 +88,7 @@ function CreateInstanceRequestToJSONTyped(value, ignoreDiscriminator = false) {
         'environment': value['environment'],
         'offering_id': value['offeringId'],
         'disk_gb': value['diskGb'],
+        'auto_terminate_hours': value['autoTerminateHours'],
         'env': value['env'],
         'image': value['image'],
         'registry_credential_id': value['registryCredentialId'],

package/dist/models/Instance.d.ts

@@ -83,7 +83,19 @@ export interface Instance {
      */
     image?: string;
     /**
-     * Why the instance is in the error state, e.g. "provisioning stalled: the instance started but never became reachable". Present only on errored instances.
+     * The runtime limit this instance was launched with, in hours (the `auto_terminate_hours` you sent). Absent means the instance runs until something stops it. It stays present after the limit has fired, so a terminated instance still explains itself.
+     * @type {number}
+     * @memberof Instance
+     */
+    autoTerminateHours?: number;
+    /**
+     * When the runtime limit expires and the platform terminates this instance. Computed when the instance became running (ready_at + auto_terminate_hours), so the countdown covers usable runtime, not the time spent provisioning. Absent when no limit was set, when the instance is not running yet, or once the platform has begun acting on an elapsed limit.
+     * @type {Date}
+     * @memberof Instance
+     */
+    autoTerminateAt?: Date;
+    /**
+     * Why the instance stopped or failed, e.g. "provisioning stalled: the instance started but never became reachable", or — when a runtime limit fired — "Terminated: this instance reached the 8-hour runtime limit set at launch". Present only when the platform has something to explain.
      * @type {string}
      * @memberof Instance
      */

package/dist/models/Instance.js

@@ -81,6 +81,8 @@ function InstanceFromJSONTyped(json, ignoreDiscriminator) {
         'connection': json['connection'] == null ? undefined : (0, InstanceConnection_1.InstanceConnectionFromJSON)(json['connection']),
         'diskGb': json['disk_gb'] == null ? undefined : json['disk_gb'],
         'image': json['image'] == null ? undefined : json['image'],
+        'autoTerminateHours': json['auto_terminate_hours'] == null ? undefined : json['auto_terminate_hours'],
+        'autoTerminateAt': json['auto_terminate_at'] == null ? undefined : (new Date(json['auto_terminate_at'])),
         'statusReason': json['status_reason'] == null ? undefined : json['status_reason'],
         'createdAt': (new Date(json['created_at'])),
         'readyAt': json['ready_at'] == null ? undefined : (new Date(json['ready_at'])),
@@ -108,6 +110,8 @@ function InstanceToJSONTyped(value, ignoreDiscriminator = false) {
         'connection': (0, InstanceConnection_1.InstanceConnectionToJSON)(value['connection']),
         'disk_gb': value['diskGb'],
         'image': value['image'],
+        'auto_terminate_hours': value['autoTerminateHours'],
+        'auto_terminate_at': value['autoTerminateAt'] == null ? value['autoTerminateAt'] : value['autoTerminateAt'].toISOString(),
         'status_reason': value['statusReason'],
         'created_at': value['createdAt'].toISOString(),
         'ready_at': value['readyAt'] == null ? value['readyAt'] : value['readyAt'].toISOString(),

package/dist/models/Model.d.ts

@@ -89,6 +89,18 @@ export interface Model {
      * @memberof Model
      */
     aliases?: Array<string>;
+    /**
+     * Whether this model can be used as the base of a managed fine-tuning job. The tunable allowlist is operator-curated and much narrower than the serving catalog, so this is false for almost every model. Absent means false.
+     * @type {boolean}
+     * @memberof Model
+     */
+    fineTunable?: boolean;
+    /**
+     * The exact id to send as `model` when creating a fine-tuning job for this entry. Present only when `fine_tunable` is true, and deliberately distinct from `id`: the allowlist matches the customer-facing base id (e.g. `qwen2.5-7b-instruct`), which a served model normally carries as an alias rather than as its namespaced catalog id.
+     * @type {string}
+     * @memberof Model
+     */
+    fineTuneBaseModel?: string;
     /**
      *
      * @type {ModelStatusEnum}

package/dist/models/Model.js

@@ -83,6 +83,8 @@ function ModelFromJSONTyped(json, ignoreDiscriminator) {
         'parameters': json['parameters'] == null ? undefined : (json['parameters'].map(ModelParametersInner_1.ModelParametersInnerFromJSON)),
         'pricing': (0, ModelPricing_1.ModelPricingFromJSON)(json['pricing']),
         'aliases': json['aliases'] == null ? undefined : json['aliases'],
+        'fineTunable': json['fine_tunable'] == null ? undefined : json['fine_tunable'],
+        'fineTuneBaseModel': json['fine_tune_base_model'] == null ? undefined : json['fine_tune_base_model'],
         'status': json['status'],
     };
 }
@@ -106,6 +108,8 @@ function ModelToJSONTyped(value, ignoreDiscriminator = false) {
         'parameters': value['parameters'] == null ? undefined : (value['parameters'].map(ModelParametersInner_1.ModelParametersInnerToJSON)),
         'pricing': (0, ModelPricing_1.ModelPricingToJSON)(value['pricing']),
         'aliases': value['aliases'],
+        'fine_tunable': value['fineTunable'],
+        'fine_tune_base_model': value['fineTuneBaseModel'],
         'status': value['status'],
     };
 }

package/docs/CreateInstanceRequest.md

@@ -18,6 +18,7 @@ Name | Type
 `environment` | string
 `offeringId` | string
 `diskGb` | number
+`autoTerminateHours` | number
 `env` | { [key: string]: string; }
 `image` | string
 `registryCredentialId` | string
@@ -44,6 +45,7 @@ const example = {
   "environment": certified:pytorch@2.13,
   "offeringId": null,
   "diskGb": 200,
+  "autoTerminateHours": 8,
   "env": null,
   "image": ghcr.io/acme/trainer:v1,
   "registryCredentialId": null,

package/docs/FineTuningApi.md

@@ -10,6 +10,7 @@ All URIs are relative to *https://api.gpu.ai/v1*
 | [**getFineTuningJob**](FineTuningApi.md#getfinetuningjob) | **GET** /fine_tuning/jobs/{id} | Retrieve a fine-tuning job (OpenAI-compatible) |
 | [**listFineTuningJobEvents**](FineTuningApi.md#listfinetuningjobevents) | **GET** /fine_tuning/jobs/{id}/events | List fine-tuning job events (OpenAI-compatible) |
 | [**listFineTuningJobs**](FineTuningApi.md#listfinetuningjobs) | **GET** /fine_tuning/jobs | List fine-tuning jobs (OpenAI-compatible) |
+| [**retrieveFile**](FineTuningApi.md#retrievefile) | **GET** /files/{id} | Retrieve a file (OpenAI-compatible) |
 
 
 
@@ -466,3 +467,79 @@ example().catch(console.error);
 
 [[Back to top]](#) [[Back to API list]](../README.md#api-endpoints) [[Back to Model list]](../README.md#models) [[Back to README]](../README.md)
 
+
+## retrieveFile
+
+> FileObject retrieveFile(id)
+
+Retrieve a file (OpenAI-compatible)
+
+Returns the metadata for one of your uploaded files. The lookup is scoped to the organization the API key belongs to: a file id that does not exist and one that belongs to another organization both return the same 404, so this route cannot be used to probe for other tenants\' ids.  Use it to check that a `training_file` is present and usable BEFORE creating a fine-tuning job — the create path performs the identical ownership check, so a file this route returns is a file that create will accept.
+
+### Example
+
+```ts
+import {
+  Configuration,
+  FineTuningApi,
+} from '@gpuai/sdk';
+import type { RetrieveFileRequest } from '@gpuai/sdk';
+
+async function example() {
+  console.log("🚀 Testing @gpuai/sdk SDK...");
+  const config = new Configuration({ 
+    // Configure HTTP bearer authorization: bearerAuth
+    accessToken: "YOUR BEARER TOKEN",
+  });
+  const api = new FineTuningApi(config);
+
+  const body = {
+    // string | The file id returned by the upload, e.g. `file-9f2c1a...`.
+    id: id_example,
+  } satisfies RetrieveFileRequest;
+
+  try {
+    const data = await api.retrieveFile(body);
+    console.log(data);
+  } catch (error) {
+    console.error(error);
+  }
+}
+
+// Run the test
+example().catch(console.error);
+```
+
+### Parameters
+
+
+| Name | Type | Description  | Notes |
+|------------- | ------------- | ------------- | -------------|
+| **id** | `string` | The file id returned by the upload, e.g. `file-9f2c1a...`. | [Defaults to `undefined`] |
+
+### Return type
+
+[**FileObject**](FileObject.md)
+
+### Authorization
+
+[bearerAuth](../README.md#bearerAuth)
+
+### HTTP request headers
+
+- **Content-Type**: Not defined
+- **Accept**: `application/json`
+
+
+### HTTP response details
+| Status code | Description | Response headers |
+|-------------|-------------|------------------|
+| **200** | The file object. |  -  |
+| **401** | Missing or invalid API key (OpenAI error envelope). |  -  |
+| **403** | API key lacks the required scope (OpenAI error envelope). |  -  |
+| **404** | Model or resource not found (OpenAI error envelope). |  -  |
+| **429** | Rate limit exceeded (OpenAI error envelope). Retry-After header indicates seconds to wait. |  * Retry-After - Seconds the client should wait before retrying. <br>  |
+| **500** | Internal server error (OpenAI error envelope). |  -  |
+
+[[Back to top]](#) [[Back to API list]](../README.md#api-endpoints) [[Back to Model list]](../README.md#models) [[Back to README]](../README.md)
+

package/docs/Instance.md

@@ -17,6 +17,8 @@ Name | Type
 `connection` | [InstanceConnection](InstanceConnection.md)
 `diskGb` | number
 `image` | string
+`autoTerminateHours` | number
+`autoTerminateAt` | Date
 `statusReason` | string
 `createdAt` | Date
 `readyAt` | Date
@@ -42,6 +44,8 @@ const example = {
   "connection": null,
   "diskGb": null,
   "image": null,
+  "autoTerminateHours": 8,
+  "autoTerminateAt": null,
   "statusReason": null,
   "createdAt": null,
   "readyAt": null,

package/docs/Model.md

@@ -18,6 +18,8 @@ Name | Type
 `parameters` | [Array<ModelParametersInner>](ModelParametersInner.md)
 `pricing` | [ModelPricing](ModelPricing.md)
 `aliases` | Array<string>
+`fineTunable` | boolean
+`fineTuneBaseModel` | string
 `status` | string
 
 ## Example
@@ -39,6 +41,8 @@ const example = {
   "parameters": null,
   "pricing": null,
   "aliases": null,
+  "fineTunable": null,
+  "fineTuneBaseModel": null,
   "status": null,
 } satisfies Model
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@gpuai/sdk",
-  "version": "0.3.7",
+  "version": "0.3.9",
   "description": "TypeScript SDK for the GPU.ai Public Developer API",
   "author": "GPU.ai Support <support@gpu.ai>",
   "repository": {

package/src/apis/FineTuningApi.ts

@@ -72,6 +72,10 @@ export interface ListFineTuningJobsRequest {
     limit?: number;
 }
 
+export interface RetrieveFileRequest {
+    id: string;
+}
+
 /**
  * 
  */
@@ -436,6 +440,61 @@ export class FineTuningApi extends runtime.BaseAPI {
         return await response.value();
     }
 
+    /**
+     * Creates request options for retrieveFile without sending the request
+     */
+    async retrieveFileRequestOpts(requestParameters: RetrieveFileRequest): Promise<runtime.RequestOpts> {
+        if (requestParameters['id'] == null) {
+            throw new runtime.RequiredError(
+                'id',
+                'Required parameter "id" was null or undefined when calling retrieveFile().'
+            );
+        }
+
+        const queryParameters: any = {};
+
+        const headerParameters: runtime.HTTPHeaders = {};
+
+        if (this.configuration && this.configuration.accessToken) {
+            const token = this.configuration.accessToken;
+            const tokenString = await token("bearerAuth", []);
+
+            if (tokenString) {
+                headerParameters["Authorization"] = `Bearer ${tokenString}`;
+            }
+        }
+
+        let urlPath = `/files/{id}`;
+        urlPath = urlPath.replace('{id}', encodeURIComponent(String(requestParameters['id'])));
+
+        return {
+            path: urlPath,
+            method: 'GET',
+            headers: headerParameters,
+            query: queryParameters,
+        };
+    }
+
+    /**
+     * Returns the metadata for one of your uploaded files. The lookup is scoped to the organization the API key belongs to: a file id that does not exist and one that belongs to another organization both return the same 404, so this route cannot be used to probe for other tenants\' ids.  Use it to check that a `training_file` is present and usable BEFORE creating a fine-tuning job — the create path performs the identical ownership check, so a file this route returns is a file that create will accept.
+     * Retrieve a file (OpenAI-compatible)
+     */
+    async retrieveFileRaw(requestParameters: RetrieveFileRequest, initOverrides?: RequestInit | runtime.InitOverrideFunction): Promise<runtime.ApiResponse<FileObject>> {
+        const requestOptions = await this.retrieveFileRequestOpts(requestParameters);
+        const response = await this.request(requestOptions, initOverrides);
+
+        return new runtime.JSONApiResponse(response, (jsonValue) => FileObjectFromJSON(jsonValue));
+    }
+
+    /**
+     * Returns the metadata for one of your uploaded files. The lookup is scoped to the organization the API key belongs to: a file id that does not exist and one that belongs to another organization both return the same 404, so this route cannot be used to probe for other tenants\' ids.  Use it to check that a `training_file` is present and usable BEFORE creating a fine-tuning job — the create path performs the identical ownership check, so a file this route returns is a file that create will accept.
+     * Retrieve a file (OpenAI-compatible)
+     */
+    async retrieveFile(requestParameters: RetrieveFileRequest, initOverrides?: RequestInit | runtime.InitOverrideFunction): Promise<FileObject> {
+        const response = await this.retrieveFileRaw(requestParameters, initOverrides);
+        return await response.value();
+    }
+
 }
 
 /**

package/src/models/CreateInstanceRequest.ts

@@ -99,6 +99,12 @@ export interface CreateInstanceRequest {
      * @memberof CreateInstanceRequest
      */
     diskGb?: number;
+    /**
+     * Terminate this instance automatically after this many hours of runtime. Omit for no limit (the default). The countdown starts when the instance reaches `running`, not when you call this — provisioning is neither billed nor counted against the limit — and the resolved deadline is returned as `auto_terminate_at` on every read. Enforcement is server-side and survives your process: this is the safety net for an unattended run whose client dies without cleaning up. When it fires the instance is terminated exactly as a DELETE would terminate it (billing stops, the `instance.terminated` webhook fires) and `status_reason` records that the limit was reached. Bounds: at least 1, at most 720 (30 days) — outside them returns 422 `validation_failed`. An explicit `0` is rejected rather than read as "no limit"; omit the field for that.
+     * @type {number}
+     * @memberof CreateInstanceRequest
+     */
+    autoTerminateHours?: number;
     /**
      * Environment variables. With `template_id`: per-deploy overrides — only keys the template marks user_overridable are accepted (others 422), and secret values are redacted from operation metadata. With `image`: plain container env vars — keys must be POSIX names and the platform's reserved namespace (`PUBLIC_KEY`, `GPUAI_*`) is refused (422).
      * @type {{ [key: string]: string; }}
@@ -180,6 +186,7 @@ export function CreateInstanceRequestFromJSONTyped(json: any, ignoreDiscriminato
         'environment': json['environment'] == null ? undefined : json['environment'],
         'offeringId': json['offering_id'] == null ? undefined : json['offering_id'],
         'diskGb': json['disk_gb'] == null ? undefined : json['disk_gb'],
+        'autoTerminateHours': json['auto_terminate_hours'] == null ? undefined : json['auto_terminate_hours'],
         'env': json['env'] == null ? undefined : json['env'],
         'image': json['image'] == null ? undefined : json['image'],
         'registryCredentialId': json['registry_credential_id'] == null ? undefined : json['registry_credential_id'],
@@ -212,6 +219,7 @@ export function CreateInstanceRequestToJSONTyped(value?: CreateInstanceRequest |
         'environment': value['environment'],
         'offering_id': value['offeringId'],
         'disk_gb': value['diskGb'],
+        'auto_terminate_hours': value['autoTerminateHours'],
         'env': value['env'],
         'image': value['image'],
         'registry_credential_id': value['registryCredentialId'],

package/src/models/Instance.ts

@@ -94,7 +94,19 @@ export interface Instance {
      */
     image?: string;
     /**
-     * Why the instance is in the error state, e.g. "provisioning stalled: the instance started but never became reachable". Present only on errored instances.
+     * The runtime limit this instance was launched with, in hours (the `auto_terminate_hours` you sent). Absent means the instance runs until something stops it. It stays present after the limit has fired, so a terminated instance still explains itself.
+     * @type {number}
+     * @memberof Instance
+     */
+    autoTerminateHours?: number;
+    /**
+     * When the runtime limit expires and the platform terminates this instance. Computed when the instance became running (ready_at + auto_terminate_hours), so the countdown covers usable runtime, not the time spent provisioning. Absent when no limit was set, when the instance is not running yet, or once the platform has begun acting on an elapsed limit.
+     * @type {Date}
+     * @memberof Instance
+     */
+    autoTerminateAt?: Date;
+    /**
+     * Why the instance stopped or failed, e.g. "provisioning stalled: the instance started but never became reachable", or — when a runtime limit fired — "Terminated: this instance reached the 8-hour runtime limit set at launch". Present only when the platform has something to explain.
      * @type {string}
      * @memberof Instance
      */
@@ -193,6 +205,8 @@ export function InstanceFromJSONTyped(json: any, ignoreDiscriminator: boolean):
         'connection': json['connection'] == null ? undefined : InstanceConnectionFromJSON(json['connection']),
         'diskGb': json['disk_gb'] == null ? undefined : json['disk_gb'],
         'image': json['image'] == null ? undefined : json['image'],
+        'autoTerminateHours': json['auto_terminate_hours'] == null ? undefined : json['auto_terminate_hours'],
+        'autoTerminateAt': json['auto_terminate_at'] == null ? undefined : (new Date(json['auto_terminate_at'])),
         'statusReason': json['status_reason'] == null ? undefined : json['status_reason'],
         'createdAt': (new Date(json['created_at'])),
         'readyAt': json['ready_at'] == null ? undefined : (new Date(json['ready_at'])),
@@ -224,6 +238,8 @@ export function InstanceToJSONTyped(value?: Instance | null, ignoreDiscriminator
         'connection': InstanceConnectionToJSON(value['connection']),
         'disk_gb': value['diskGb'],
         'image': value['image'],
+        'auto_terminate_hours': value['autoTerminateHours'],
+        'auto_terminate_at': value['autoTerminateAt'] == null ? value['autoTerminateAt'] : value['autoTerminateAt'].toISOString(),
         'status_reason': value['statusReason'],
         'created_at': value['createdAt'].toISOString(),
         'ready_at': value['readyAt'] == null ? value['readyAt'] : value['readyAt'].toISOString(),

package/src/models/Model.ts

@@ -106,6 +106,18 @@ export interface Model {
      * @memberof Model
      */
     aliases?: Array<string>;
+    /**
+     * Whether this model can be used as the base of a managed fine-tuning job. The tunable allowlist is operator-curated and much narrower than the serving catalog, so this is false for almost every model. Absent means false.
+     * @type {boolean}
+     * @memberof Model
+     */
+    fineTunable?: boolean;
+    /**
+     * The exact id to send as `model` when creating a fine-tuning job for this entry. Present only when `fine_tunable` is true, and deliberately distinct from `id`: the allowlist matches the customer-facing base id (e.g. `qwen2.5-7b-instruct`), which a served model normally carries as an alias rather than as its namespaced catalog id.
+     * @type {string}
+     * @memberof Model
+     */
+    fineTuneBaseModel?: string;
     /**
      * 
      * @type {ModelStatusEnum}
@@ -181,6 +193,8 @@ export function ModelFromJSONTyped(json: any, ignoreDiscriminator: boolean): Mod
         'parameters': json['parameters'] == null ? undefined : ((json['parameters'] as Array<any>).map(ModelParametersInnerFromJSON)),
         'pricing': ModelPricingFromJSON(json['pricing']),
         'aliases': json['aliases'] == null ? undefined : json['aliases'],
+        'fineTunable': json['fine_tunable'] == null ? undefined : json['fine_tunable'],
+        'fineTuneBaseModel': json['fine_tune_base_model'] == null ? undefined : json['fine_tune_base_model'],
         'status': json['status'],
     };
 }
@@ -208,6 +222,8 @@ export function ModelToJSONTyped(value?: Model | null, ignoreDiscriminator: bool
         'parameters': value['parameters'] == null ? undefined : ((value['parameters'] as Array<any>).map(ModelParametersInnerToJSON)),
         'pricing': ModelPricingToJSON(value['pricing']),
         'aliases': value['aliases'],
+        'fine_tunable': value['fineTunable'],
+        'fine_tune_base_model': value['fineTuneBaseModel'],
         'status': value['status'],
     };
 }