Creating and Using Templates with API

Introduction

A template in the Vast.ai API is a configuration bundle that stores default settings for instance creation. Instead of specifying every parameter each time you create an instance, you can reference a template by its hash_id and optionally override specific values. Templates are useful for:

Standardization: Ensure all team members launch instances with consistent configurations
Convenience: Avoid repeating the same parameters across multiple API calls
Sharing: Share configurations via template hash ID

For information about managing templates in the web interface, see Templates Introduction.

Template Fields Reference

When creating a template, the following fields can be configured:

Field	Type	Description
`name`	string	Required. Human-readable name for the template
`image`	string	Required. Docker image path (e.g., `vastai/pytorch`)
`tag`	string	Docker image tag. Defaults to `latest`
`desc`	string	Description of the template
`readme`	string	Longer documentation/readme content
`env`	string	Environment variables and port mappings in Docker flag format (e.g., `"-e VAR=val -p 8000:8000"`)
`onstart`	string	Shell commands to run when instance starts
`runtype`	string	Launch mode: `ssh`, `jupyter`, or `args`. Defaults to `args`
`args_str`	string	Replaces the image’s Docker `CMD`. If the image defines an `ENTRYPOINT`, this is passed as arguments to it
`ssh_direct`	boolean	Enable direct SSH connection (recommended with `runtype: "ssh"`)
`use_ssh`	boolean	Enable SSH access
`jup_direct`	boolean	Enable direct Jupyter connection
`jupyter_dir`	string	Directory to launch Jupyter from
`use_jupyter_lab`	boolean	Use JupyterLab instead of Jupyter Notebook
`docker_login_repo`	string	Required (use `""` if not needed). Private Docker registry URL
`docker_login_user`	string	Required (use `""` if not needed). Username for private registry
`docker_login_pass`	string	Required (use `""` if not needed). Access token for private registry
`href`	string	Link to Docker Hub or image documentation
`repo`	string	Repository identifier (e.g., `library/ubuntu`)
`extra_filters`	object	Default machine search filters (e.g., `{"cuda_max_good": {"gte": 12.6}}`)
`recommended_disk_space`	number	Recommended disk space in GB (default: 8)
`private`	boolean	Whether the template is private
`volume_info`	object	UI hint for volume configuration (not used for actual instance creation)

Template Identifiers

Templates have two primary identifiers:

Identifier	Type	Description
`id`	integer	Numeric identifier. Used for deleting templates
`hash_id`	string	Content-based hash. Used for creating instances and editing templates

Usage by operation:

Create instance: Use template_hash_id
Edit template: Use hash_id (via PUT)
Delete template: Use numeric id

Precedence Rules

When you create an instance with both a template and additional parameters, the following precedence rules apply:

Field Type	Behavior
Scalar fields (image, disk, runtype, etc.)	Request value overrides template value
`env` (dict)	Merged. Template values retained, request values added. Conflicting keys use request value
`extra_filters` (dict)	Merged by key. Request values win on conflicts

Example: Environment Variables

When creating an instance, the env field is passed as a JSON object (dict). When you provide env in your request, it is merged with the template’s env — existing values are retained and new values are added.

Templates store env as a Docker flag string (e.g., "-e VAR=val -p 8000:8000"), but instance creation uses a dict format. The API handles the conversion automatically when merging.

Template configuration (string format):

{
  "env": "-e MODEL_ID=deepseek-ai/DeepSeek-R1-Distill-Llama-8B -e MAX_TOKENS=4096"
}

Instance creation request (dict format):

{
  "template_hash_id": "4e17788f74f075dd9aab7d0d4427968f",
  "env": {
    "MODEL_ID": "mistralai/Mistral-7B-v0.1",
    "HF_TOKEN": "hf_xxx"
  }
}

Resulting instance environment:

MODEL_ID=mistralai/Mistral-7B-v0.1 (request overrides template)
MAX_TOKENS=4096 (retained from template)
HF_TOKEN=hf_xxx (added from request)

Cookbook Examples

Search for Templates

Search for templates using select_filters with comparison operators. Query Syntax:

select_filters = { "field": { "op": value } }

Operators: eq, neq, lt, lte, gt, gte, in, notin Available Fields:

Field	Type	Description
`creator_id`	int	ID of creator
`created_at`	float	Time of initial template creation (UTC epoch)
`count_created`	int	Number of instances created (popularity)
`default_tag`	string	Image default tag
`docker_login_repo`	string	Image docker repository
`id`	int	Template unique ID
`image`	string	Image used for template
`jup_direct`	bool	Supports jupyter direct
`hash_id`	string	Unique hash ID of template
`name`	string	Displayable name
`recent_create_date`	float	Last time of instance creation (UTC epoch)
`recommended_disk_space`	float	Min disk space required
`recommended`	bool	Is template on recommended list
`ssh_direct`	bool	Supports SSH direct
`tag`	string	Image tag
`use_ssh`	bool	Supports SSH (direct or proxy)

# Search for recommended templates with SSH support
curl -G "https://console.vast.ai/api/v0/template/" \
  -H "Authorization: Bearer $VAST_API_KEY" \
  --data-urlencode 'select_filters={"use_ssh":{"eq":true},"recommended":{"eq":true}}'

# Search for popular templates (more than 100 instances created)
curl -G "https://console.vast.ai/api/v0/template/" \
  -H "Authorization: Bearer $VAST_API_KEY" \
  --data-urlencode 'select_filters={"count_created":{"gt":100}}'

Create a New Template

Create a reusable template. This example shows a recommended configuration with SSH direct access.

curl -X POST "https://console.vast.ai/api/v0/template/" \
  -H "Authorization: Bearer $VAST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Example Template",
    "desc": "Template for running vLLM inference server",
    "image": "vllm/vllm-openai",
    "tag": "latest",
    "env": "-e MODEL_ID=deepseek-ai/DeepSeek-R1-Distill-Llama-8B -p 8000:8000",
    "onstart": "echo \"Starting vLLM server\"; vllm serve $MODEL_ID",
    "runtype": "ssh",
    "ssh_direct": true,
    "use_ssh": true,
    "docker_login_repo": "",
    "docker_login_user": "",
    "docker_login_pass": "",
    "recommended_disk_space": 50,
    "private": true
  }'

Full Template Example

Here’s a complete template creation request with all common fields:

{
  "name": "Example Template",
  "desc": "Description of what this template does",
  "readme": "Longer documentation\nwith multiple lines",
  "image": "library/ubuntu",
  "tag": "22.04",
  "repo": "library/ubuntu",
  "href": "https://hub.docker.com/r/library/ubuntu/",
  "env": "-e ENV1=val1 -p 8000:8000",
  "onstart": "echo \"hello\"",
  "args_str": "",
  "runtype": "ssh",
  "ssh_direct": true,
  "use_ssh": true,
  "jup_direct": true,
  "jupyter_dir": null,
  "use_jupyter_lab": false,
  "docker_login_repo": "",
  "docker_login_user": "",
  "docker_login_pass": "",
  "extra_filters": {"cuda_max_good": {"gte": 12.6}},
  "recommended_disk_space": 8,
  "private": true,
  "volume_info": null
}

Edit a Template

Edit an existing template using its hash_id. You only need to include the fields you want to change - unchanged fields retain their existing values.

curl -X PUT "https://console.vast.ai/api/v0/template/" \
  -H "Authorization: Bearer $VAST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "hash_id": "5915f1dc1ce881defb572015eb9d8178",
    "desc": "Updated description",
    "recommended_disk_space": 16
  }'

The hash_id will change after editing since it is derived from the template content.

Delete a Template

Delete a template by passing its numeric id (not hash_id) in the request body.

curl -X DELETE "https://console.vast.ai/api/v0/template/" \
  -H "Authorization: Bearer $VAST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"template_id": 334548}'

Create Instance from Template

Launch an instance using a template. No need to specify image as the template provides it.

curl -X PUT "https://console.vast.ai/api/v0/asks/12345678/" \
  -H "Authorization: Bearer $VAST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "template_hash_id": "4e17788f74f075dd9aab7d0d4427968f"
  }'

Create Instance with Image Override

Use a template but specify a different Docker image.

curl -X PUT "https://console.vast.ai/api/v0/asks/12345678/" \
  -H "Authorization: Bearer $VAST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "template_hash_id": "4e17788f74f075dd9aab7d0d4427968f",
    "image": "library/ubuntu:22.04"
  }'

Override Environment Variables

Override template environment variables with new values. Instance creation uses the dict format for env.

curl -X PUT "https://console.vast.ai/api/v0/asks/12345678/" \
  -H "Authorization: Bearer $VAST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "template_hash_id": "4e17788f74f075dd9aab7d0d4427968f",
    "env": {"MODEL_ID": "mistralai/Mistral-7B-Instruct-v0.2", "HF_TOKEN": "hf_xxxYourTokenHere", "-p 8000:8000": "1"}
  }'

Create Instance with Volume

Attach a volume when creating an instance. You can either link an existing volume or create a new one.

The volume_info field stored in templates is a UI hint only. To actually attach a volume, you must include volume_info in the instance creation request.

Link an existing volume:

curl -X PUT "https://console.vast.ai/api/v0/asks/12345678/" \
  -H "Authorization: Bearer $VAST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "template_hash_id": "4e17788f74f075dd9aab7d0d4427968f",
    "volume_info": {
      "create_new": false,
      "volume_id": 12345,
      "mount_path": "/workspace"
    }
  }'

Create a new volume:

curl -X PUT "https://console.vast.ai/api/v0/asks/12345678/" \
  -H "Authorization: Bearer $VAST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "template_hash_id": "4e17788f74f075dd9aab7d0d4427968f",
    "volume_info": {
      "create_new": true,
      "volume_id": 28908979,
      "size": 10,
      "mount_path": "/workspace"
    }
  }'

CLI equivalent:

# Link existing volume
vastai create instance 12345678 --template_hash abc123 --link-volume 12345 --mount-path /workspace

# Create new volume
vastai create instance 12345678 --template_hash abc123 --create-volume 28908979 --volume-size 10 --mount-path /workspace

CLI Reference

The Vast.ai CLI provides commands for template management:

Command	Description
`vastai create template --name <name> --image <image> [options]`	Create a new template
`vastai search templates [filters]`	Search for templates
`vastai update template <hash_id> [options]`	Update a template (uses `hash_id`)
`vastai delete template --template-id <id>`	Delete a template (uses numeric `id`)

Create template options:

--name NAME - Template name
--image IMAGE - Docker image
--image_tag TAG - Image tag
--env ENV - Docker options (env vars and ports)
--ssh - Launch as SSH instance
--jupyter - Launch as Jupyter instance
--direct - Use direct connections
--onstart-cmd CMD - Onstart script
--disk_space GB - Disk space in GB
--desc DESC - Description
--readme README - Readme content
--public - Make template public

The CLI update template command takes hash_id as its argument, while delete template uses the numeric id.

Runtype and Connection Options

The runtype field controls the launch mode of your instance:

Runtype	Description
`args`	Default. `args_str` replaces the image’s `CMD` and is passed to its `ENTRYPOINT` if one is defined
`ssh`	SSH access enabled. Recommended with `ssh_direct: true`
`jupyter`	Jupyter notebook/lab access

Recommendation: Use runtype: "ssh" with ssh_direct: true and use_ssh: true for reliable SSH access to your instances. The args_str field is used when runtype is args. It replaces the image’s Docker CMD — if the image defines an ENTRYPOINT, args_str is passed as arguments to it. If the image has no ENTRYPOINT (only CMD), args_str replaces the command entirely.

{
  "runtype": "args",
  "args_str": "--model deepseek-ai/DeepSeek-R1-Distill-Llama-8B --port 8000"
}

Common Pitfalls

I'm getting an error using template_id for instance creation

Instance creation requires template_hash_id, not template_id. The numeric id is only used for deleting templates. Use the hash_id returned when you create or search for templates:

{
  "template_hash_id": "4e17788f74f075dd9aab7d0d4427968f"
}

I passed template + image and got the wrong image

When you specify both template_hash_id and image, the request’s image overrides the template’s image. If you want to use the template’s image, omit the image field from your request.

My environment variable didn't apply

Template creation and instance creation use different formats for the env field:

Templates: Docker flag string format — "-e VAR1=value1 -e VAR2=value2 -p 8000:8000"
Instance creation: Dict format — {"VAR1": "value1", "VAR2": "value2", "-p 8000:8000": "1"}

When creating an instance with a template, the request env (dict) is merged with the template env — existing keys are overwritten, new keys are added.

Ports aren't open on my instance

When creating instances, port mappings are specified in the env dict using the -p syntax as keys:

{
  "env": {
    "-p 8000:8000": "1",
    "-p 8080:8080": "1"
  }
}

For SSH access, use runtype: "ssh" with ssh_direct: true.

Volume didn't mount

Volume mounting uses the volume_info structure in the instance creation request. Note that volume_info in templates is just a UI hint and doesn’t affect instance creation.To link an existing volume:

{
  "volume_info": {
    "create_new": false,
    "volume_id": 12345,
    "mount_path": "/workspace"
  }
}

To create a new volume:

{
  "volume_info": {
    "create_new": true,
    "volume_id": 28908979,
    "size": 10,
    "mount_path": "/workspace"
  }
}

Where:

volume_id is either an existing volume ID (from show volumes) or a volume offer ID (from search volumes)
size is only used when create_new is true
mount_path is where the volume mounts inside the container

Template search returns no results

Template search uses select_filters with comparison operators, not free-text search:

Use the correct filter syntax: {"field": {"op": value}}
Valid operators: eq, neq, lt, lte, gt, gte, in, notin
Verify your API key has user_read permissions
Check available fields in the search documentation above

Templates Introduction

Web interface guide for templates

Create Instance API

Full API reference for instance creation

CLI Commands

Command-line interface for templates

Search Offers API

Find available machines to rent

API Reference

Endpoints

Introduction

Template Fields Reference

Template Identifiers

Precedence Rules

Example: Environment Variables

Cookbook Examples

Search for Templates

Create a New Template

Full Template Example

Edit a Template

Delete a Template

Create Instance from Template

Create Instance with Image Override

Override Environment Variables

Create Instance with Volume

CLI Reference

Runtype and Connection Options

Common Pitfalls

Templates Introduction

Create Instance API

CLI Commands

Search Offers API

API Reference

Endpoints

​Introduction

​Template Fields Reference

​Template Identifiers

​Precedence Rules

​Example: Environment Variables

​Cookbook Examples

​Search for Templates

​Create a New Template

​Full Template Example

​Edit a Template

​Delete a Template

​Create Instance from Template

​Create Instance with Image Override

​Override Environment Variables

​Create Instance with Volume

​CLI Reference

​Runtype and Connection Options

​Common Pitfalls

​Related Resources

Templates Introduction

Create Instance API

CLI Commands

Search Offers API

Introduction

Template Fields Reference

Template Identifiers

Precedence Rules

Example: Environment Variables

Cookbook Examples

Search for Templates

Create a New Template

Full Template Example

Edit a Template

Delete a Template

Create Instance from Template

Create Instance with Image Override

Override Environment Variables

Create Instance with Volume

CLI Reference

Runtype and Connection Options

Common Pitfalls

Related Resources