Template Base Images

The base image is the foundation of a sandbox template, determining the operating system and pre-installed software.

Creating a Template

When creating a template with Template(), you can configure file context options:

import { Template } from '@e2b/code-interpreter'

const template = Template({
  fileContextPath: './my-context',  // Custom file context path (default: ".")
  fileIgnorePatterns: ['*.log', 'node_modules/**']  // File patterns to exclude
})

Rules from a .dockerignore file are automatically integrated; matching files are not uploaded or included in hash calculations.

Defining a Base Image

Predefined Images

Use predefined image methods to quickly specify common environments:

import { Template } from '@e2b/code-interpreter'

// Ubuntu
const ubuntuTemplate = Template().fromUbuntuImage('jammy')  // or '22.04'

// Debian
const debianTemplate = Template().fromDebianImage('stable-slim')  // or 'bullseye'

// Python
const pythonTemplate = Template().fromPythonImage('3.13')  // or '3.11'

// Node.js
const nodeTemplate = Template().fromNodeImage('lts')  // or '20'

// Bun
const bunTemplate = Template().fromBunImage('1.3')

These predefined methods pull images from the corresponding official registry. If you are using public registries like Docker Hub in a network environment with restrictions, consider using fromImage() and specifying the full path via a mirror proxy.

Custom Images

Use images from any Docker Registry:

import { Template } from '@e2b/code-interpreter'

const customTemplate = Template().fromImage(
  'crp-internal.sufycs.com/registry-1.docker.io/library/python:3.13'
)

Default Base Image

If no image is specified, the E2B default base image is used:

import { Template } from '@e2b/code-interpreter'

const defaultTemplate = Template()  // Uses e2bdev/base

Based on an Existing Template

Create a new template from an existing one:

import { Template } from '@e2b/code-interpreter'

// Using a template alias
const template = Template().fromTemplate('my-existing-template')

// Using a namespaced template
const nsTemplate = Template().fromTemplate('namespace/template-name')

Parsing from a Dockerfile

Convert existing Dockerfile content into a template:

import { Template } from '@e2b/code-interpreter'

const template = Template().fromDockerfile(`
  FROM crp-internal.sufycs.com/registry-1.docker.io/library/python:3.11
  RUN pip install numpy pandas
  WORKDIR /app
  COPY . .
  ENV DEBUG=true
  CMD ["python", "app.py"]
`)
  .pipInstall(['matplotlib'])  // Can continue chaining configuration

Supported Dockerfile Instructions

Instruction	Status	Conversion Method
FROM	Supported	Converted to fromImage()
RUN	Supported	Converted to runCmd()
COPY/ADD	Supported	Converted to copy()
WORKDIR	Supported	Converted to setWorkdir()
USER	Supported	Converted to setUser()
ENV	Supported	Converted to setEnvs()
CMD/ENTRYPOINT	Supported	Converted to setStartCmd()
EXPOSE	Not supported	Skipped
VOLUME	Not supported	Skipped

Important Limitations

Base image methods can only be called once: Each template can only set one base image.

// Error: Cannot set base image multiple times
const template = Template()
  .fromImage('crp-internal.sufycs.com/registry-1.docker.io/library/python:3.11')
  .fromImage('crp-internal.sufycs.com/registry-1.docker.io/library/node:20')  // Error!

// Correct: Set only once
const template = Template().fromImage(
  'crp-internal.sufycs.com/registry-1.docker.io/library/python:3.11'
)

Multi-stage builds are not supported: Dockerfile parsing does not support multi-stage builds.