@gpuai/sdk 0.3.1 → 0.3.3

package/README.md

@@ -1,4 +1,4 @@
-# @gpuai/sdk@0.3.1
+# @gpuai/sdk@0.3.3
 
 A TypeScript SDK client for the api.gpu.ai API.
 
@@ -196,7 +196,7 @@ and is automatically generated by the
 [OpenAPI Generator](https://openapi-generator.tech) project:
 
 - API version: `1.0.0`
-- Package version: `0.3.1`
+- Package version: `0.3.3`
 - Generator version: `7.24.0`
 - Build package: `org.openapitools.codegen.languages.TypeScriptFetchClientCodegen`
 

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

@@ -58,7 +58,13 @@ export interface CreateInstanceRequest {
      */
     maxPricePerHour?: number;
     /**
-     * Deploy a curated application template (e.g. comfyui, vllm). When set, gpu_type may be omitted — the cheapest eligible GPU is auto-selected. For templates serving a Hugging Face model via env.MODEL (vllm, sglang), the model's published weights size is checked against the GPU's VRAM at submit time: an oversized model on an explicit gpu_type returns 422 `model_too_large` (auto-select instead raises its VRAM floor to the estimate, and 422s only when no available GPU fits). Gated/private/unknown repos are never blocked — the deploy proceeds and the 202 carries a `warnings` entry.
+     * The retail $/hr the caller displayed to the user before launch. A catalog row is a quote, not a booking, and pinning an offering fixes which machine is placed but not its rate. If the launch-time price exceeds this by more than $0.01 the launch fails with `price_changed` carrying the new price instead of proceeding at the higher price; omit to accept the launch-time price (previous behavior). Refusal only — it is never read as consent, so it can neither raise what you are charged nor widen what may be bought on your behalf. Use max_price_per_hour to bound spend.
+     * @type {number}
+     * @memberof CreateInstanceRequest
+     */
+    viewedPricePerHour?: number;
+    /**
+     * Deploy a curated application template (e.g. comfyui, vllm). When set, gpu_type may be omitted — the cheapest eligible GPU is auto-selected. For templates serving a Hugging Face model via env.MODEL (vllm, sglang), the model's published weights size is checked against the GPU's VRAM at submit time: an oversized model on an explicit gpu_type returns 422 `model_too_large` (auto-select instead raises its VRAM floor to the estimate, and 422s only when no available GPU fits). Gated/private/unknown repos are never blocked — the deploy proceeds and the 202 carries a `warnings` entry. Separately, an explicit gpu_type whose VRAM is below the template's own `min_vram_gb` returns 422 `gpu_vram_too_small`, naming the floor, the type's VRAM and the compatible GPU types. Auto-select is unaffected — it already filters by the floor.
      * @type {string}
      * @memberof CreateInstanceRequest
      */

package/dist/esm/models/CreateInstanceRequest.js

@@ -45,6 +45,7 @@ export function CreateInstanceRequestFromJSONTyped(json, ignoreDiscriminator) {
         'sshKeyIds': json['ssh_key_ids'] == null ? undefined : json['ssh_key_ids'],
         'name': json['name'] == null ? undefined : json['name'],
         'maxPricePerHour': json['max_price_per_hour'] == null ? undefined : json['max_price_per_hour'],
+        'viewedPricePerHour': json['viewed_price_per_hour'] == null ? undefined : json['viewed_price_per_hour'],
         'templateId': json['template_id'] == null ? undefined : json['template_id'],
         'environment': json['environment'] == null ? undefined : json['environment'],
         'offeringId': json['offering_id'] == null ? undefined : json['offering_id'],
@@ -67,6 +68,7 @@ export function CreateInstanceRequestToJSONTyped(value, ignoreDiscriminator = fa
         'ssh_key_ids': value['sshKeyIds'],
         'name': value['name'],
         'max_price_per_hour': value['maxPricePerHour'],
+        'viewed_price_per_hour': value['viewedPricePerHour'],
         'template_id': value['templateId'],
         'environment': value['environment'],
         'offering_id': value['offeringId'],

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

@@ -67,6 +67,7 @@ export declare const ProblemCodeEnum: {
     readonly DiskUnavailable: "disk_unavailable";
     readonly DiskExceedsOffering: "disk_exceeds_offering";
     readonly ModelTooLarge: "model_too_large";
+    readonly GpuVramTooSmall: "gpu_vram_too_small";
     readonly PriceChanged: "price_changed";
     readonly InsufficientBalance: "insufficient_balance";
     readonly QuotaExceeded: "quota_exceeded";

package/dist/esm/models/Problem.js

@@ -26,6 +26,7 @@ export const ProblemCodeEnum = {
     DiskUnavailable: 'disk_unavailable',
     DiskExceedsOffering: 'disk_exceeds_offering',
     ModelTooLarge: 'model_too_large',
+    GpuVramTooSmall: 'gpu_vram_too_small',
     PriceChanged: 'price_changed',
     InsufficientBalance: 'insufficient_balance',
     QuotaExceeded: 'quota_exceeded',

package/dist/models/CreateInstanceRequest.d.ts

@@ -58,7 +58,13 @@ export interface CreateInstanceRequest {
      */
     maxPricePerHour?: number;
     /**
-     * Deploy a curated application template (e.g. comfyui, vllm). When set, gpu_type may be omitted — the cheapest eligible GPU is auto-selected. For templates serving a Hugging Face model via env.MODEL (vllm, sglang), the model's published weights size is checked against the GPU's VRAM at submit time: an oversized model on an explicit gpu_type returns 422 `model_too_large` (auto-select instead raises its VRAM floor to the estimate, and 422s only when no available GPU fits). Gated/private/unknown repos are never blocked — the deploy proceeds and the 202 carries a `warnings` entry.
+     * The retail $/hr the caller displayed to the user before launch. A catalog row is a quote, not a booking, and pinning an offering fixes which machine is placed but not its rate. If the launch-time price exceeds this by more than $0.01 the launch fails with `price_changed` carrying the new price instead of proceeding at the higher price; omit to accept the launch-time price (previous behavior). Refusal only — it is never read as consent, so it can neither raise what you are charged nor widen what may be bought on your behalf. Use max_price_per_hour to bound spend.
+     * @type {number}
+     * @memberof CreateInstanceRequest
+     */
+    viewedPricePerHour?: number;
+    /**
+     * Deploy a curated application template (e.g. comfyui, vllm). When set, gpu_type may be omitted — the cheapest eligible GPU is auto-selected. For templates serving a Hugging Face model via env.MODEL (vllm, sglang), the model's published weights size is checked against the GPU's VRAM at submit time: an oversized model on an explicit gpu_type returns 422 `model_too_large` (auto-select instead raises its VRAM floor to the estimate, and 422s only when no available GPU fits). Gated/private/unknown repos are never blocked — the deploy proceeds and the 202 carries a `warnings` entry. Separately, an explicit gpu_type whose VRAM is below the template's own `min_vram_gb` returns 422 `gpu_vram_too_small`, naming the floor, the type's VRAM and the compatible GPU types. Auto-select is unaffected — it already filters by the floor.
      * @type {string}
      * @memberof CreateInstanceRequest
      */

package/dist/models/CreateInstanceRequest.js

@@ -53,6 +53,7 @@ function CreateInstanceRequestFromJSONTyped(json, ignoreDiscriminator) {
         'sshKeyIds': json['ssh_key_ids'] == null ? undefined : json['ssh_key_ids'],
         'name': json['name'] == null ? undefined : json['name'],
         'maxPricePerHour': json['max_price_per_hour'] == null ? undefined : json['max_price_per_hour'],
+        'viewedPricePerHour': json['viewed_price_per_hour'] == null ? undefined : json['viewed_price_per_hour'],
         'templateId': json['template_id'] == null ? undefined : json['template_id'],
         'environment': json['environment'] == null ? undefined : json['environment'],
         'offeringId': json['offering_id'] == null ? undefined : json['offering_id'],
@@ -75,6 +76,7 @@ function CreateInstanceRequestToJSONTyped(value, ignoreDiscriminator = false) {
         'ssh_key_ids': value['sshKeyIds'],
         'name': value['name'],
         'max_price_per_hour': value['maxPricePerHour'],
+        'viewed_price_per_hour': value['viewedPricePerHour'],
         'template_id': value['templateId'],
         'environment': value['environment'],
         'offering_id': value['offeringId'],

package/dist/models/Problem.d.ts

@@ -67,6 +67,7 @@ export declare const ProblemCodeEnum: {
     readonly DiskUnavailable: "disk_unavailable";
     readonly DiskExceedsOffering: "disk_exceeds_offering";
     readonly ModelTooLarge: "model_too_large";
+    readonly GpuVramTooSmall: "gpu_vram_too_small";
     readonly PriceChanged: "price_changed";
     readonly InsufficientBalance: "insufficient_balance";
     readonly QuotaExceeded: "quota_exceeded";

package/dist/models/Problem.js

@@ -34,6 +34,7 @@ exports.ProblemCodeEnum = {
     DiskUnavailable: 'disk_unavailable',
     DiskExceedsOffering: 'disk_exceeds_offering',
     ModelTooLarge: 'model_too_large',
+    GpuVramTooSmall: 'gpu_vram_too_small',
     PriceChanged: 'price_changed',
     InsufficientBalance: 'insufficient_balance',
     QuotaExceeded: 'quota_exceeded',

package/docs/CreateInstanceRequest.md

@@ -13,6 +13,7 @@ Name | Type
 `sshKeyIds` | Array<string>
 `name` | string
 `maxPricePerHour` | number
+`viewedPricePerHour` | number
 `templateId` | string
 `environment` | string
 `offeringId` | string
@@ -33,6 +34,7 @@ const example = {
   "sshKeyIds": null,
   "name": null,
   "maxPricePerHour": null,
+  "viewedPricePerHour": null,
   "templateId": null,
   "environment": certified:pytorch@2.13,
   "offeringId": null,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@gpuai/sdk",
-  "version": "0.3.1",
+  "version": "0.3.3",
   "description": "TypeScript SDK for the GPU.ai Public Developer API",
   "author": "GPU.ai Support <support@gpu.ai>",
   "repository": {

package/src/models/CreateInstanceRequest.ts

@@ -62,7 +62,13 @@ export interface CreateInstanceRequest {
      */
     maxPricePerHour?: number;
     /**
-     * Deploy a curated application template (e.g. comfyui, vllm). When set, gpu_type may be omitted — the cheapest eligible GPU is auto-selected. For templates serving a Hugging Face model via env.MODEL (vllm, sglang), the model's published weights size is checked against the GPU's VRAM at submit time: an oversized model on an explicit gpu_type returns 422 `model_too_large` (auto-select instead raises its VRAM floor to the estimate, and 422s only when no available GPU fits). Gated/private/unknown repos are never blocked — the deploy proceeds and the 202 carries a `warnings` entry.
+     * The retail $/hr the caller displayed to the user before launch. A catalog row is a quote, not a booking, and pinning an offering fixes which machine is placed but not its rate. If the launch-time price exceeds this by more than $0.01 the launch fails with `price_changed` carrying the new price instead of proceeding at the higher price; omit to accept the launch-time price (previous behavior). Refusal only — it is never read as consent, so it can neither raise what you are charged nor widen what may be bought on your behalf. Use max_price_per_hour to bound spend.
+     * @type {number}
+     * @memberof CreateInstanceRequest
+     */
+    viewedPricePerHour?: number;
+    /**
+     * Deploy a curated application template (e.g. comfyui, vllm). When set, gpu_type may be omitted — the cheapest eligible GPU is auto-selected. For templates serving a Hugging Face model via env.MODEL (vllm, sglang), the model's published weights size is checked against the GPU's VRAM at submit time: an oversized model on an explicit gpu_type returns 422 `model_too_large` (auto-select instead raises its VRAM floor to the estimate, and 422s only when no available GPU fits). Gated/private/unknown repos are never blocked — the deploy proceeds and the 202 carries a `warnings` entry. Separately, an explicit gpu_type whose VRAM is below the template's own `min_vram_gb` returns 422 `gpu_vram_too_small`, naming the floor, the type's VRAM and the compatible GPU types. Auto-select is unaffected — it already filters by the floor.
      * @type {string}
      * @memberof CreateInstanceRequest
      */
@@ -131,6 +137,7 @@ export function CreateInstanceRequestFromJSONTyped(json: any, ignoreDiscriminato
         'sshKeyIds': json['ssh_key_ids'] == null ? undefined : json['ssh_key_ids'],
         'name': json['name'] == null ? undefined : json['name'],
         'maxPricePerHour': json['max_price_per_hour'] == null ? undefined : json['max_price_per_hour'],
+        'viewedPricePerHour': json['viewed_price_per_hour'] == null ? undefined : json['viewed_price_per_hour'],
         'templateId': json['template_id'] == null ? undefined : json['template_id'],
         'environment': json['environment'] == null ? undefined : json['environment'],
         'offeringId': json['offering_id'] == null ? undefined : json['offering_id'],
@@ -157,6 +164,7 @@ export function CreateInstanceRequestToJSONTyped(value?: CreateInstanceRequest |
         'ssh_key_ids': value['sshKeyIds'],
         'name': value['name'],
         'max_price_per_hour': value['maxPricePerHour'],
+        'viewed_price_per_hour': value['viewedPricePerHour'],
         'template_id': value['templateId'],
         'environment': value['environment'],
         'offering_id': value['offeringId'],

package/src/models/Problem.ts

@@ -73,6 +73,7 @@ export const ProblemCodeEnum = {
     DiskUnavailable: 'disk_unavailable',
     DiskExceedsOffering: 'disk_exceeds_offering',
     ModelTooLarge: 'model_too_large',
+    GpuVramTooSmall: 'gpu_vram_too_small',
     PriceChanged: 'price_changed',
     InsufficientBalance: 'insufficient_balance',
     QuotaExceeded: 'quota_exceeded',