Elastic Deployment API - Mint Starter Kit

Note: API operations are not supported for on-demand GPU instances. All API features are designed for Elastic Deployment only.

1. List Private Images

→ Request

POST /api/v1/dev/image/private/list

Request Body

{
  "page_index": 1,
  "page_size": 10
}

← Response

Response Parameters

Parameter	Type	Description
`code`	String	Response status. `Success` on success
`msg`	String	Error message. Empty when successful
`data.list`	List	List of image objects

response Object Fields

Parameter	Type	Description
`id`	Int	Image ID
`image_name`	String	Image name
`image_uuid`	String	Image UUID (used when creating deployments)

Response (on success)

{
    "code": "Success",
    "msg": ""
    "data": {
        "list": [
            {
                "id": 111,
                "created_at": "2022-01-20T18:34:08+08:00",
                "updated_at": "2022-01-20T18:34:08+08:00",
                "image_uuid": "image-db8346e037",
                "image_name": "image name",
                "status": "finished",
            }
        ],
        "page_index": 1,
        "page_size": 10,
        "max_page": 1,
        "offset": 0,
    },
}

Python

import requests
headers = {
    "Authorization": "your token",
    "Content-Type": "application/json"
}
url = "https://api.gpuhub.com/api/v1/dev/image/private/list"
body = {
    "page_index": 1,
    "page_size": 10,
}
response = requests.post(url, json=body, headers=headers)
print(response.content.decode())

2. Create Deployment

→ Request

POST /api/v1/dev/deployment

Place all parameters in the request body. Details are as follows:

Parameter	Type	Required	Description
`name`	String	Yes	Deployment name
`deployment_type`	String	Yes	Deployment type: `ReplicaSet`, `Job`, or `Container`
`replica_num`	Int	Required for ReplicaSet & Job	Desired number of container replicas
`parallelism_num`	Int	Required for Job only	Max number of containers running in parallel (for Job type)
`reuse_container`	Bool	No	Whether to reuse stopped containers (significantly speeds up creation). Default: `false`
`reuse_container_scope`	String	No	Scope of reuse: `all` (account-wide) or `deployment` (current deployment only). Default: `all`
`service_6006_port_protocol`	String	No	Protocol for port 6006 mapping: `http` or `tcp.` Default: `http`
`service_6008_port_protocol`	String	No	Protocol for port 6008 mapping: `http` or `tcp.` Default: `http`
`container_template`	Container Object	Yes	Container template (see below)

container_template Object

Parameter	Type	Required	Description
`dc_list`	`list<String>`	Yes	List of available regions (data centers). See Appendix for values
`cuda_v_from`	Int	Yes	Minimum CUDA version supported by driver (e.g., `118` = CUDA 11.8). See Appendix
`cuda_v_to`	Int	Yes	Maximum CUDA version (higher versions are backward compatible)
`gpu_name_set`	`list<String>`	Yes	Allowed GPU models (e.g., `["RTX 5090", "RTX Pro 6000"]`)
`gpu_num`	Int	Yes	Number of GPUs per container
`cpu_num_from` / `cpu_num_to`	Int	Yes	CPU core range (1 vCPU units)
`memory_size_from` / `memory_size_to`	Int	Yes	Memory range in GB
`price_from` / `price_to`	Int	Yes	Price range in USD * 1000 (e.g., USD0.10/hr → `100`, USD9.00/hr → `9000`)
`image_uuid`	String	Yes	UUID of your private image or public base image
`cmd`	String	Yes	Startup command (executed when container starts)
`cmd_before_shutdown`	String	No	Command executed before container stops (5-second timeout; forced stop on timeout)

Request Body

{
 "name": "api-auto-created",
 "deployment_type": "ReplicaSet",
 "replica_num": 2,
 "reuse_container": true,
 "container_template": {
   "dc_list": ["Singpore-A", "Singapore-B"],
   "gpu_name_set": ["RTX 5090", "RTX Pro 6000"],
   "cuda_v_from": 130,
   "cuda_v_to": 999,
   "gpu_num": 1,
   "cpu_num_from": 4,
   "cpu_num_to": 64,
   "memory_size_from": 16,
   "memory_size_to": 256,
   "cmd": "python main.py",
   "price_from": 100,    // base price: $0.10/hr
   "price_to": 9000,     // base price: $9.00/hr
   "image_uuid": "base-image-ec1e9vdbd3"
 }
}

← Response

Response (on success)

{
   "code": "Success",
   "msg": "",
   "data": {
       "deployment_uuid": "833f1cd5a764fa3"
   }
}

Python

import requests

headers = {
    "Authorization": "your token",
    "Content-Type": "application/json"
}

url = "https://api.gpuhub.com/api/v1/dev/deployment"

# Example 1: Create a ReplicaSet deployment (auto-scaling)
body = {
    "name": "my-inference-service",
    "deployment_type": "ReplicaSet",
    "replica_num": 2,
    "reuse_container": True,
    "container_template": {
        "dc_list": ["Singpore-A", "Singapore-B"],
        "gpu_name_set": ["RTX 5090", "RTX Pro 6000"],
        "gpu_num": 1,
        "cuda_v_from": 118,
        "cuda_v_to": 128,
        "cpu_num_from": 4,
        "cpu_num_to": 64,
        "memory_size_from": 16,
        "memory_size_to": 256,
        "cmd": "python main.py",
        "price_from": 100,   # $0.10/hr (price × 1000)
        "price_to": 9000,    # $9.00/hr (price × 1000)
        "image_uuid": "image-xxxxxxxx"
    }
}

response = requests.post(url, json=body, headers=headers)
print(response.content.decode())

# Example 2: Create a Job deployment (batch processing)
{
    "name": "batch-training-job",
    "deployment_type": "Job",
    "replica_num": 4,
    "parallelism_num": 1,
    "reuse_container": True,
    "container_template": {
        "dc_list": ["Singpore-A"],
        "gpu_name_set": ["RTX 5090"],
        "gpu_num": 1,
        "cuda_v_from": 118,
        "cuda_v_to": 128,
        "cpu_num_from": 8,
        "cpu_num_to": 64,
        "memory_size_from": 32,
        "memory_size_to": 256,
        "cmd": "python train.py",
        "price_from": 100,
        "price_to": 9000,
        "image_uuid": "image-xxxxxxxx"
    }
}

# Example 3: Create a single Container deployment
{
    "name": "debug-container",
    "deployment_type": "Container",
    "reuse_container": True,
    "container_template": {
        "dc_list": ["Singpore-A"],
        "gpu_name_set": ["RTX 5090"],
        "gpu_num": 1,
        "cuda_v": 118,
        "cpu_num_from": 4,
        "cpu_num_to": 32,
        "memory_size_from": 16,
        "memory_size_to": 128,
        "cmd": "sleep infinity",
        "price_from": 100,
        "price_to": 9000,
        "image_uuid": "image-xxxxxxxx"
    }
}

3. List Deployments

→ Request

POST /api/v1/dev/deployment/list

Place the following parameters in the request body:

Parameter	Type	Required	Description
`page_index`	Int	Yes	Page number (starting from 1)
`page_size`	Int	Yes	Number of items per page
`name`	String	No	Filter by exact deployment name (exact match only, no fuzzy search)
`status`	String	No	Filter by deployment status: `running` → active deployments `stopped` → stopped deployments → leave empty to return all
`deployment_uuid`	String	No	Filter by specific deployment UUID

Request Body

{
  "page_index": 1,
  "page_size": 10,
  "name": "my-api",           // optional: filter by exact deployment name
  "status": "running",        // optional: "running" | "stopped" | empty (all)
  "deployment_uuid": "xxx"    // optional: filter by deployment UUID
}

← Response

The fields returned in each deployment object have the same meaning as the parameters used when creating the deployment.

Response (on success)

{
  "code": "Success",
  "msg": "",
  "data": {
    "list": [
      {
        "id": 214,
        "uuid": "53a677bb3e281b8",
        "name": "xxxx",
        "deployment_type": "Container",
        "status": "stopped",
        "replica_num": 1,
        "parallelism_num": 1,
        "reuse_container": true,
        "starting_num": 0,
        "running_num": 0,
        "finished_num": 2,
        "image_uuid": "image-db8346e037",
        "template": {
          "dc_list": ["Singpore-A", "Singapore-B"],
          "gpu_name_set": ["RTX 5090"],
          "gpu_num": 1,
          "cpu_num_from": 1,
          "cpu_num_to": 100,
          "memory_size_from": 1073741824,
          "memory_size_to": 274877906944,
          "price_from": 100,
          "price_to": 9000,
          "cuda_v_from": 118,
          "cuda_v_to": 128,
          "cmd": "sleep 100"
        },
        "price_estimates": 0,
        "created_at": "2023-01-05T20:34:07+08:00",
        "updated_at": "2023-01-05T20:34:07+08:00",
        "stopped_at": null
      }
    ],
    "page_index": 1,
    "page_size": 10,
    "offset": 0,
    "max_page": 1,
    "result_total": 3
  }
}

4. Scale Replicas (ReplicaSet only)

→ Request

PUT /api/v1/dev/deployment/replica_num

Place the following parameters in the request body:

Parameter	Type	Required	Description
`deployment_uuid`	String	Yes	UUID of the deployment to scale replicas for
`replica_num`	Int	Yes	Desired number of replicas (must be greater than 0)

Request Body

{
  "deployment_uuid": "53a677bb3e281b8",
  "replica_num": 3
}

← Response

Parameter	Type	Description
`code`	String	Response status. `Success` on success
`msg`	String	Error message. Empty when successful
`data`	null	No data returned

Response (on success)

{
  "code": "Success",
  "msg": "",
  "data": null
}

Python

import requests

headers = {
    "Authorization": "your token",
    "Content-Type": "application/json"
}

url = "https://api.gpuhub.com/api/v1/dev/deployment/replica_num"

body = {
    "deployment_uuid": "5be3045703152b9",
    "replica_num": 16
}

response = requests.put(url, json=body, headers=headers)
print(response.json())

5. Stop Entire Deployment

→ Request

PUT /api/v1/dev/deployment/operate

Place the following parameters in the request body:

Parameter	Type	Required	Description
`deployment_uuid`	String	Yes	UUID of the deployment to operate
`operation`	String	Yes	Operation to perform: `stop` → stop the deployment `delete` → delete the deployment

Request Body

{
  "deployment_uuid": "53a677bb3e281b8",
  "operation": "stop"
}

← Response

Parameter	Type	Description
`code`	String	Response status. `Success` on success
`msg`	String	Error message. Empty when successful
`data`	null	No data returned

Response (on success)

{
  "code": "Success",
  "msg": "",
  "data": null
}

Python

import requests
headers = {
    "Authorization": "your token",
    "Content-Type": "application/json"
}
url = "https://api.gpuhub.com/api/v1/dev/deployment/operate"
body = {
    "deployment_uuid": "5be3045703152b9",
    "operate": "stop"
}
response = requests.put(url, json=body, headers=headers)
print(response.content.decode())

6. Delete Entire Deployment

→ Request

If the deployment is still running, the system will automatically stop it first, then delete it completely (including all containers and associated resources).

DELETE /api/v1/dev/deployment

Request Body

JSON{
  "deployment_uuid": "xxx"
}

← Response

Parameter	Type	Description
`code`	String	`Success` on success
`msg`	String	Empty when successful
`data`	null	No data returned

Response (on success)

JSON{
  "code": "Success",
  "msg": "",
  "data": null
}

Python

import requests
headers = {
    "Authorization": "your_token_here",
    "Content-Type": "application/json"
}

url = "https://api.gpuhub.com/api/v1/dev/deployment"
body = {"deployment_uuid": "5be3045703152b9"}

response = requests.delete(url, json=body, headers=headers)
print(response.json())

7. List Container Events

→ Request

POST /api/v1/dev/deployment/container/event/list

Place the following parameters in the request body:

Parameter	Type	Required	Description
`deployment_uuid`	String	Yes	Deployment UUID
`deployment_container_uuid`	String	No	Specific container UUID (leave empty for all)
`page_index`	Int	Yes	Page number (starting from 1)
`page_size`	Int	Yes	Number of items per page
`offset`	Int	No	Starting offset (used for polling new events)

Request Body

{
  "deployment_uuid": "da497aea1eb8343", 
  "deployment_container_uuid": "", 
  "page_index": 1, 
  "page_size": 10,
  "offset": 0
}

← Response

Response Parameters

Parameter	Type	Description
`code`	String	Response status. `Success` on success
`msg`	String	Error message. Empty when successful
`data`	Object	Container event data

Event Object Fields

Parameter	Type	Description
`deployment_container_uuid`	String	UUID of the container
`status`	String	Container state (e.g., `creating`, `starting`, `running`, `shutting_down`, `shutdown`)
`created_at`	String	Timestamp when the state change occurred

Response (on success)

{
    "code": "Success",
    "data": {
        "list": [
            {
                "deployment_container_uuid": "da497aea1eb8343-f94411a60c-1502e6e2",
                "status": "shutdown",
                "created_at": "2022-12-13T16:42:45+08:00"
            },
            {
                "deployment_container_uuid": "da497aea1eb8343-f94411a60c-1502e6e2",
                "status": "shutting_down",
                "created_at": "2022-12-13T16:42:40+08:00"
            },
            {
                "deployment_container_uuid": "da497aea1eb8343-f94411a60c-1502e6e2",
                "status": "running",
                "created_at": "2022-12-13T16:34:57+08:00"
            },
            {
                "deployment_container_uuid": "da497aea1eb8343-f94411a60c-1502e6e2",
                "status": "oss_merged",
                "created_at": "2022-12-13T16:34:55+08:00"
            },
            {
                "deployment_container_uuid": "da497aea1eb8343-f94411a60c-1502e6e2",
                "status": "starting",
                "created_at": "2022-12-13T16:34:55+08:00"
            },
            {
                "deployment_container_uuid": "da497aea1eb8343-f94411a60c-1502e6e2",
                "status": "created",
                "created_at": "2022-12-13T16:34:54+08:00"
            },
            {
                "deployment_container_uuid": "da497aea1eb8343-f94411a60c-1502e6e2",
                "status": "creating",
                "created_at": "2022-12-13T16:34:47+08:00"
            }
        ],
        "page_index": 1,
        "page_size": 10,
        "offset": 0,
        "max_page": 1,
    },
    "msg": ""
}

Python

import requests
headers = {
    "Authorization": "your_token_here",
    "Content-Type": "application/json"
}

url = "https://api.gpuhub.com/api/v1/dev/deployment/container/event/list"
body = {
    "deployment_uuid": "da497aea1eb8343", 
    "deployment_container_uuid": "", 
    "page_index": 1, 
    "page_size": 10,
    "offset": 0
}

response = requests.post(url, json=body, headers=headers)
print(response.json())

8. List Containers

→ Request

Inside the container, you can get the current container’s UUID via the environment variable: AutoDLContainerUUID

POST /api/v1/dev/deployment/container/list

Request Body Parameters

Parameter	Type	Required	Description
`deployment_uuid`	String	Yes	Deployment UUID
`container_uuid`	String	No	Filter by specific container UUID
`date_from` / `date_to`	String	No	Filter by container creation time range
`gpu_name`	String	No	Filter by GPU model
`cpu_num_from/to`	Int	No	CPU core range
`memory_size_from/to`	Int	No	Memory range (bytes)
`price_from/to`	Float	No	Base price range (USD × 1000)
`released`	Boolean	No	Include released containers?
`status`	`List<String>`	No	Filter by status: e.g., `["running"]`
`page_index`	Int	Yes	Page number (default: 1)
`page_size`	Int	Yes	Items per page (default: 10)
`offset`	Int	No	Starting offset

Request Body

{
    "deployment_uuid": "da497aea1eb8343", 
    "page_index": 1, 
    "page_size": 10
}

← Response

Response Fields

Parameter	Type	Description
`code`	String	Response status. `Success` on success
`msg`	String	Error message. Empty when successful
`data.list`	Array	List of container objects

Container Object

Field	Type	Description
`uuid`	String	Container UUID
`version`	String	Container version (auto-generated or user-specified on deployment update)
`data_center`	String	Region / data center code (e.g., `Singpore-A`)
`deployment_uuid`	String	Parent deployment UUID
`machine_id`	String	Physical host ID
`status`	String	Current status (`running`, `stopped`, `starting`, etc.)
`gpu_name`	String	GPU model (e.g., `RTX 5090`)
`gpu_num`	Int	Number of GPUs allocated
`cpu_num`	Int	Number of CPU cores
`memory_size`	Int	Memory size in bytes
`image_uuid`	String	Image UUID
`price`	Float	Base price in USD × 1000 (e.g., 1600 = $1.60/hr)
`info`	Object	Connection details (see below)
`started_at`	String	Time when container started running
`stopped_at`	String	Time when container stopped (null if running)
`created_at`	String	Creation timestamp
`updated_at`	String	Last update timestamp

info Object (Connection Details)

Field	Type	Description
`ssh_command`	String	Full SSH login command
`root_password`	String	Root password for SSH
`service_6006_port_url`	String	Public HTTPS URL for port 6006
`service_6008_port_url`	String	Public HTTP URL for port 6008
`service_url`, `proxy_host`, `custom_port`	—	Deprecated — use the two service URLs above

Response (on success)

{
  "code": "Success",
  "msg": "",
  "data": {
    "list": [
      {
        "uuid": "53a677bb3e281b8-f94411a60c-63c24009",
        "data_center": "Singpore-A",
        "deployment_uuid": "da497aea1eb8343",
        "status": "running",
        "gpu_name": "RTX 5090",
        "gpu_num": 1,
        "cpu_num": 16,
        "memory_size": 68719476736,
        "image_uuid": "image-xxxxxxxx",
        "price": 1600,
        "info": {
          "ssh_command": "ssh -p 22345 root@singpore-a.gpuhub.com",
          "root_password": "xxxxxxxxxx",
          "service_6006_port_url": "https://singpore-a.gpuhub.com:22346",
          "service_6008_port_url": "http://singpore-a.gpuhub.com:22348"
        },
        "started_at": "2025-11-19T10:43:03+08:00",
        "created_at": "2025-11-19T10:42:50+08:00"
      }
    ],
    "page_index": 1,
    "page_size": 10,
    "max_page": 1
  }
}

Python

import requests

headers = {
    "Authorization": "your_token_here",
    "Content-Type": "application/json"
}

url = "https://api.gpuhub.com/api/v1/dev/deployment/container/list"

body = {
    "deployment_uuid": "da497aea1eb8343",
    "status": ["running"],
    "page_index": 1,
    "page_size": 10
}

response = requests.post(url, json=body, headers=headers)
print(response.json())

9. Stop a Container

→ Request

PUT /api/v1/dev/deployment/container/stop

Place the following parameters in the request body:

Parameter	Type	Required	Description
`deployment_container_uuid`	String	Yes	UUID of the container to stop
`decrease_one_replica_num`	Boolean	No	ReplicaSet only: if `true`, decreases the desired replica count by 1 after stopping (default: `false`)
`no_cache`	Boolean	No	If `true`, prevents the stopped container from being cached for reuse (default: `false`)
`cmd_before_shutdown`	String	No	Command to run before shutdown (5-second timeout; force-kills on timeout). Overrides any `cmd_before_shutdown` set during deployment creation (very rare chance both run)

Request Body

{
     "deployment_container_uuid": "da497aea1eb8343-f94411a60c-a394fb30",
     "decrease_one_replica_num": false,
     "no_cache": false,
     "cmd_before_shutdown": "sleep 5"
}

← Response

Response Parameters

Parameter	Type	Description
`code`	String	Response status. `Success` on success
`msg`	String	Error message. Empty when successful
`data`	null	No data returned

Response (on success)

{
    "code": "Success",
    "msg": "",
    "data": null
}

Python

import requests
headers = {
    "Authorization": "your token",
    "Content-Type": "application/json"
}

url = "https://api.gpuhub.com/api/v1/dev/deployment/container/event/stop"
body = {
    "deployment_container_uuid": "da497aea1eb8343-f94411a60c-a394fb30",
    "decrease_one_replica_num": false,
    "no_cache": false,
    "cmd_before_shutdown": "sleep 5"
}

response = requests.put(url, json=body, headers=headers)
print(response.content.decode())

10. Host Blacklist

→ Request

To avoid bad machines: if a container experiences unknown issues (e.g., slow startup, crashes), you can blacklist its host to prevent future scheduling on that machine.

POST /api/v1/dev/deployment/blacklist

Parameter	Type	Required	Description
`deployment_container_uuid`	String	Yes	UUID of the problematic container
`expire_in_minutes`	Int	No	Blacklist duration in minutes. Default: 1440 (24 hours). Max: 43200 (30 days)
`comment`	String	No	Optional note (e.g., “Slow boot”)

Request Body

{
  "deployment_container_uuid": "da497aea1eb8343-f94411a60c-1502e6e2",
  "expire_in_minutes": 1440,
  "comment": "Slow startup — avoid this host"
}

← Response

Response Parameters

Parameter	Type	Description
`code`	String	Response status. `Success` on success
`msg`	String	Error message. Empty when successful
`data`	null	No data returned

Response (on success)

{
    "code": "Success",
    "msg": "",
    "data": null
}

Python

import requests

headers = {"Authorization": "your token", "Content-Type": "application/json"}
url = "https://api.gpuhub.com/api/v1/dev/deployment/blacklist"

body = {
    "deployment_container_uuid": "da497aea1eb8343-f94411a60c-1502e6e2",
    "expire_in_minutes": 1440,
    "comment": "Slow startup — avoid this host"
}

response = requests.post(url, json=body, headers=headers)
print(response.json())

11. Get Active Blacklist

→ Request

GET /api/v1/dev/deployment/blacklist

No body required.

← Response

Response Parameters

Parameter	Type	Description
`code`	String	Response status. `Success` on success
`msg`	String	Error message. Empty when successful
`data`	List	List of active blacklist entries

Blacklist Entry Object

Field	Type	Description
`created_at`	String	Timestamp when the blacklist was first created
`updated_at`	String	Timestamp of the last update (updates when expiry time is extended)
`data_center`	String	Region / data center code (e.g., `Singpore-A`)
`expired_time`	String	Timestamp when the blacklist entry will automatically expire
`machine_id`	String	Physical host ID that is blacklisted
`msg`	String	Optional comment provided when creating the blacklist

Success Response

{
  "code": "Success",
  "msg": "",
  "data": [
    {
      "created_at": "2025-03-25T17:42:55+08:00",
      "updated_at": "2025-03-25T17:48:11+08:00",
      "data_center": "Singpore-A",
      "machine_id": "24fb4ca36a",
      "expired_time": "2025-03-26T17:48:11+08:00",
      "msg": "Slow startup — avoid this host"
    }
  ]
}

Python

import requests

headers = {"Authorization": "your token"}
url = "https://api.gpuhub.com/api/v1/dev/deployment/blacklist"

response = requests.get(url, headers=headers)
print(response.json())

12. Real-time GPU Stock by Region

→ Request

POST /api/v1/dev/machine/region/gpu_stock

Use this endpoint before creating deployments to check real-time availability and avoid scheduling failures. Stock is calculated assuming 1 GPU per container. Even if a host has 2 idle GPUs, they may be on different machines — a container requiring 2 GPUs might still fail to schedule. Request Body Parameters

Parameter	Type	Required	Description
`region_sign`	String	Yes	Region code (see Appendix)
`cuda_v_from`	Int	No	Minimum supported CUDA version (e.g., `118` = CUDA 11.8)
`cuda_v_to`	Int	No	Maximum supported CUDA version
`gpu_name_set`	`List<String>`	No	Filter by specific GPU models
`memory_size_from`	Int	No	Minimum memory size in GB
`memory_size_to`	Int	No	Maximum memory size in GB
`cpu_num_from`	Int	No	Minimum CPU cores
`cpu_num_to`	Int	No	Maximum CPU cores
`price_from`	Int	No	Minimum price (USD * 1000, e.g., $0.10/hr → `100`)
`price_to`	Int	No	Maximum price (USD * 1000)

Request Body

{
  "region_sign": "Singpore-A",
  "cuda_v_from": 118,
  "cuda_v_to": 128
}

← Response

Response Parameters

Parameter	Type	Description
`code`	String	Response status. `Success` on success
`msg`	String	Error message. Empty when successful
`data`	List	List of GPU inventory objects

GPU Inventory Object

Field	Type	Description
`{GPU_MODEL}`	Object	Key is the GPU model name
`idle_gpu_num`	Int	Number of currently idle GPUs
`total_gpu_num`	Int	Total number of GPUs

Response (on success)

{
  "code": "Success",
  "msg": "",
  "data": [
    {
      "RTX 5090": {
        "idle_gpu_num": 312,
        "total_gpu_num": 2850
      }
    },
    {
      "RTX Pro 6000": {
        "idle_gpu_num": 48,
        "total_gpu_num": 640
      }
    }
  ]
}

Python

import requests

headers = {
    "Authorization": "your_token_here",
    "Content-Type": "application/json"
}

url = "https://api.gpuhub.com/api/v1/dev/machine/region/gpu_stock"

body = {
    "region_sign": "Singpore-A",
    "cuda_v_from": 118,
    "cuda_v_to": 138
}

response = requests.post(url, json=body, headers=headers)
print(response.json())

13. Check Duration Package Balance

→ Request

GET /api/v1/dev/deployment/ddp/overview

Request Parameters (Query String)

Parameter	Type	Required	Description
`deployment_uuid`	String	Yes	UUID of the deployment

← Response

Response Parameters

Parameter	Type	Description
`code`	String	Response status. `Success` on success
`msg`	String	Error message. Empty when successful
`data`	List	List of prepaid package objects

Prepaid Package Object

Field	Type	Description
`gpu_type`	String	GPU model (e.g., `RTX 5090`)
`total`	Int	Total remaining seconds across all unused prepaid packages (seconds)
`balance`	Int	Currently remaining seconds from unused packages (seconds)
`dc_list`	String	Regions covered by the package (comma-separated if multiple, e.g., `singpore-a,singapore-b`)

← Response

Response (on success)

{
    "code": "Success",
    "data": [
        {
            "gpu_type": "RTX 5090",
            "total": 86400,
            "balance": 83829,
            "dc_list": "Singpore-A,Singapore-B"
        }
    ],
    "msg": ""
}

Python

import requests

headers = {"Authorization": "your_token_here"}
url = "https://api.gpuhub.com/api/v1/dev/deployment/ddp/overview"
params = {"deployment_uuid": "your_deployment_uuid_here"}

response = requests.get(url, params=params, headers=headers)
print(response.json())

14. Appendix

region_sign & dc_list

Values for dc_list (preferred) or the deprecated region_sign parameter when creating a deployment, After a container starts, the current region is also available inside the container via the environment variable.

Region Name	`dc_list` Value	Old `region_sign`
Singapore-B	`Singpore-A`	`Singpore-A`
Singapore-A	`Singapore-B`	`Singapore-B`

Public Image UUID

Framework	Framework Version	Image UUID	Description
PyTorch	2.8.0 (Blackwell)	`base-image-ec1e9vdbd3`	CUDA 12.8 · Ubuntu 22.04
PyTorch	2.7.0	`base-image-d13e3vu5d3`	CUDA 12.8 · Ubuntu 22.04
PyTorch	2.5.1	`base-image-b2343du52k`	CUDA 12.4 · Ubuntu 22.04
PyTorch	2.5.1	`base-image-c1333ve5d3`	CUDA 8.1.rc1 · Ubuntu 22.04
PyTorch	2.3.0	`base-image-321232d5d4`	CUDA 12.1 · Ubuntu 22.04
PyTorch	2.1.2	`base-image-lbdbb183fk`	CUDA 11.8 · Ubuntu 22.04
PyTorch	2.1.0	`base-image-323d32u534`	CUDA 12.1 · Ubuntu 22.04
PyTorch	2.1.0	`base-image-pi743du52k`	CUDA 8.0 · Ubuntu 22.04
PyTorch	2.0.0	`base-image-l2t43iu6uk`	CUDA 11.8 · Ubuntu 20.04
PyTorch	1.11.0	`base-image-l374uiucui`	CUDA 11.3 · Ubuntu 20.04
PyTorch	1.10.0	`base-image-tbk3g58lhs`	CUDA 11.3 · Ubuntu 20.04
PyTorch	1.9.0	`base-image-12be412037`	CUDA 11.1 · Ubuntu 18.04
PyTorch	1.8.1	`base-image-db83412037`	CUDA 11.1 · Ubuntu 18.04
PyTorch	1.7.0	`base-image-db8346e037`	CUDA 11 · Ubuntu 18.04
PyTorch	1.6.0	`base-image-ldnbxc4by8`	CUDA 10.1 · Ubuntu 18.04
PyTorch	1.5.1	`base-image-qyil80d3s4`	CUDA 10.1 · Ubuntu 18.04
PyTorch	1.1.0	`base-image-0x7zgbw1ji`	CUDA 10 · Ubuntu 18.04
TensorFlow	2.9.0	`base-image-uxeklgirir`	CUDA 11.2 · Ubuntu 20.04
TensorFlow	2.5.0	`base-image-0gxqmciyth`	CUDA 11.2 · Ubuntu 18.04
TensorFlow	1.15.5	`base-image-4bpg0tt88l`	CUDA 11.4 · Ubuntu 20.04
Miniconda	—	`base-image-8hkyyugih5`	CUDA 11.8 · Ubuntu 22.04
Miniconda	—	`base-image-m26c71lgul`	CUDA 11.8 · Ubuntu 20.04
Miniconda	—	`base-image-mbr2n4urrc`	CUDA 11.6 · Ubuntu 20.04
Miniconda	—	`base-image-zsd8dwuwus`	CUDA 11.3(cudagl) · Ubuntu 20.04
Miniconda	—	`base-image-7bn8iqhkb5`	CUDA 11.3 · Ubuntu 18.04
Miniconda	—	`base-image-h041hn36yt`	CUDA 11.1 · Ubuntu 18.04
Miniconda	—	`base-image-hek323ev8b`	CUDA 10.1 · Ubuntu 18.04
Miniconda	—	`base-image-qkkhitpik5`	CUDA 10.2 · Ubuntu 18.04
Miniconda	—	`base-image-1e5uf433g1`	CUDA 10 · Ubuntu 18.04
Miniconda	—	`base-image-k0vep6kyq8`	CUDA 9 · Ubuntu 16.04
Miniconda	—	`base-image-b23jugu52k`	CUDA 8.0 · Ubuntu 22.04
TensorRT	8.6.1	`base-image-l2193iu12k`	CUDA 11.8 · Ubuntu 20.04
TensorRT	8.5.1	`base-image-l2843iu23k`	CUDA 11.8 · Ubuntu 20.04

More images are added regularly: contact support for the latest list or request a custom base image.

CUDA Version values

CUDA Version	Value to use in `cuda_v_from` / `cuda_v_to`
11.8	`118`
12.0	`120`
12.1	`121`
12.2	`122`

If your framework needs CUDA 11.5 (or any unlisted version), choose the lowest available version ≥ your requirement (e.g., 11.8 → 118). Higher drivers are backward-compatible, but picking a version that’s too high reduces available GPUs. Always select the smallest compatible value to maximize scheduling success.

Container Environment Variables

Variable Name	ODescription
AutoDLContainerUUID	Unique ID of the current container
AutoDLDeploymentUUID	UUID of the parent deployment
AutoDLDataCenter	Current region / data center code

​1. List Private Images

​→ Request

​← Response

​2. Create Deployment

​→ Request

​← Response

​3. List Deployments

​→ Request

​← Response

​4. Scale Replicas (ReplicaSet only)

​→ Request

​← Response

​5. Stop Entire Deployment

​→ Request

​← Response

​6. Delete Entire Deployment

​→ Request

​← Response

​7. List Container Events

​→ Request

​← Response

​8. List Containers

​→ Request

​← Response

​9. Stop a Container

​→ Request

​← Response

​10. Host Blacklist

​→ Request

​← Response

​11. Get Active Blacklist

​→ Request

​← Response

​12. Real-time GPU Stock by Region

​→ Request

​← Response

​13. Check Duration Package Balance

​→ Request

​← Response

​← Response

​14. Appendix

​region_sign & dc_list

​Public Image UUID

​CUDA Version values

​Container Environment Variables

1. List Private Images

→ Request

← Response

2. Create Deployment

→ Request

← Response

3. List Deployments

→ Request

← Response

4. Scale Replicas (ReplicaSet only)

→ Request

← Response

5. Stop Entire Deployment

→ Request

← Response

6. Delete Entire Deployment

→ Request

← Response

7. List Container Events

→ Request

← Response

8. List Containers

→ Request

← Response

9. Stop a Container

→ Request

← Response

10. Host Blacklist

→ Request

← Response

11. Get Active Blacklist

→ Request

← Response

12. Real-time GPU Stock by Region

→ Request

← Response

13. Check Duration Package Balance

→ Request

← Response

← Response

14. Appendix

region_sign & dc_list

Public Image UUID

CUDA Version values

Container Environment Variables