Overview
Container templates let you create custom environments by layering your base Docker image with in10nt agent infrastructure.
Container templates create custom Docker images. For cloning existing instances with their workspace/files, use instances.createFromTemplate() instead
Create Template
Create a custom environment from a Docker image:
const template = await client.templates.create({
name: 'python-ml', // Required: Human-readable name
description: 'Python ML env', // Optional: Description
image: 'python:3.11-slim', // Required: Base Docker image
registryType: 'dockerhub', // Required: dockerhub | private | gcr
registryAuth: { // Optional: For private registries
username: 'user',
password: 'pass'
}
});
console.log(`Template ID: ${template.templateId}`);
console.log(`Status: ${template.status}`); // "building"
Parameters
| Parameter | Type | Required | Description |
|---|
name | string | Yes | Human-readable template name |
image | string | Yes | Base Docker image (e.g., python:3.11-slim) |
registryType | string | Yes | Registry type: dockerhub, private, or gcr |
description | string | No | Template description |
registryAuth | object | No | Authentication for private registries |
Registry Types
dockerhub: Public Docker Hub imagesprivate: Private Docker registriesgcr: Google Container Registry
Response
{
templateId: "tmpl_1704588123456",
name: "python-ml",
image: "python:3.11-slim",
status: "building", // building → ready | failed
createdAt: "2025-01-06T..."
}
Wait Until Ready
Wait for template to finish building:
const template = await client.templates.create({
name: 'ml-env',
image: 'nvidia/cuda:12.2-runtime',
registryType: 'dockerhub'
});
await template.waitUntilReady({
timeout: 300000, // 5 minutes (default)
pollInterval: 5000 // Check every 5 seconds (default)
});
console.log(`Ready! Platform image: ${template.platformImage}`);
Options
| Option | Type | Default | Description |
|---|
timeout | number | 300000 | Maximum wait time in ms |
pollInterval | number | 5000 | Polling interval in ms |
List Templates
Get all your templates:
const templates = await client.templates.list();
templates.forEach(template => {
console.log(`${template.name}: ${template.status}`);
});
Get Template Status
Check template build status:
const template = await client.templates.get('tmpl_1704588123456');
console.log(`Status: ${template.status}`);
if (template.status === 'failed') {
console.error(`Error: ${template.error}`);
}
Status Values
building: Template is being builtready: Template is ready to usefailed: Build failed (check error field)
Delete Template
Delete a template:
Or by ID:
const template = await client.templates.get('tmpl_123');
await template.delete();
Use Template with Instance
Once a template is ready, use it to create instances:
// Create and wait for template
const template = await client.templates.create({
name: 'ml-env',
image: 'python:3.11-slim',
registryType: 'dockerhub'
});
await template.waitUntilReady();
// Create instance with template
const instance = await client.instances.create({
name: 'ml-agent',
templateId: template.templateId
});
const result = await instance.run('Train a classifier on data.csv');
Private Registry Authentication
For private registries, provide authentication:
const template = await client.templates.create({
name: 'private-env',
image: 'registry.example.com/myimage:latest',
registryType: 'private',
registryAuth: {
username: process.env.REGISTRY_USER,
password: process.env.REGISTRY_PASS
}
});
Never hardcode registry credentials. Use environment variables or secret management
Common Use Cases
Python ML Environment
const mlTemplate = await client.templates.create({
name: 'python-ml',
image: 'python:3.11-slim',
registryType: 'dockerhub',
description: 'Python with ML libraries'
});
await mlTemplate.waitUntilReady();
Node.js Development
const nodeTemplate = await client.templates.create({
name: 'node-dev',
image: 'node:20-alpine',
registryType: 'dockerhub',
description: 'Node.js 20 development'
});
await nodeTemplate.waitUntilReady();
CUDA/GPU Environment
const gpuTemplate = await client.templates.create({
name: 'cuda-env',
image: 'nvidia/cuda:12.2-runtime',
registryType: 'dockerhub',
description: 'NVIDIA CUDA 12.2'
});
await gpuTemplate.waitUntilReady();
Error Handling
try {
const template = await client.templates.create({
name: 'my-env',
image: 'python:3.11-slim',
registryType: 'dockerhub'
});
await template.waitUntilReady({ timeout: 300000 });
} catch (error) {
if (error.message.includes('Template build failed')) {
console.error('Build failed - check image and credentials');
} else if (error.message.includes('timeout')) {
console.error('Build took too long');
} else {
console.error('Unexpected error:', error.message);
}
}
Template vs Instance Cloning
| Feature | Container Templates | Instance Cloning |
|---|
| What it does | Layers base Docker image with agent infrastructure | Clones existing instance workspace |
| Use case | Custom base environments (Python, Node, etc.) | Reuse configured workspaces |
| Method | client.templates.create() | client.instances.createFromTemplate() |
| Build time | 2-5 minutes | Instant |
| Files included | Only base image files | All workspace files |
