One-liner Usage¶

This guide covers how to use datamodel-code-generator with pipes and clipboard tools for quick, one-time code generation without permanent installation.

Note

The package name is datamodel-code-generator, and the CLI command is datamodel-codegen.

One-liner Execution with uvx/pipx¶

You don't need to install datamodel-code-generator permanently. Use uvx or pipx run to execute it directly.

Using uvx (Recommended)¶

uvx runs Python tools without installation:

# Basic usage
uvx datamodel-codegen --input schema.json --output model.py

# With extras (e.g., HTTP support)
uvx --from 'datamodel-code-generator[http]' datamodel-codegen --url https://example.com/api.yaml --output model.py

# With GraphQL support
uvx --from 'datamodel-code-generator[graphql]' datamodel-codegen --input schema.graphql --output model.py

Using pipx run¶

pipx also supports one-shot execution:

pipx run datamodel-code-generator --input schema.json --output model.py

Reading from stdin¶

datamodel-code-generator can read schema input from stdin, enabling powerful pipeline workflows:

# Pipe JSON directly
echo '{"type": "object", "properties": {"name": {"type": "string"}}}' | \
  uvx datamodel-codegen --input-file-type jsonschema

# Pipe from another command
curl -s https://example.com/schema.json | \
  uvx datamodel-codegen --input-file-type jsonschema --output model.py

Clipboard Integration¶

Combine stdin support with clipboard tools to quickly generate models from copied schema definitions.

macOS (pbpaste/pbcopy)¶

# Generate from clipboard and print to stdout
pbpaste | uvx datamodel-codegen --input-file-type jsonschema

# Generate from clipboard and save to file
pbpaste | uvx datamodel-codegen --input-file-type jsonschema --output model.py

# Generate from clipboard and copy result back to clipboard
pbpaste | uvx datamodel-codegen --input-file-type jsonschema | pbcopy

Linux (xclip/xsel)¶

xclipxsel

# Generate from clipboard and print to stdout
xclip -selection clipboard -o | uvx datamodel-codegen --input-file-type jsonschema

# Generate and copy result to clipboard
xclip -selection clipboard -o | uvx datamodel-codegen --input-file-type jsonschema | xclip -selection clipboard

# Generate from clipboard and print to stdout
xsel --clipboard --output | uvx datamodel-codegen --input-file-type jsonschema

# Generate and copy result to clipboard
xsel --clipboard --output | uvx datamodel-codegen --input-file-type jsonschema | xsel --clipboard --input

Installing clipboard tools on Linux

# Debian/Ubuntu
sudo apt install xclip
# or
sudo apt install xsel

# Fedora
sudo dnf install xclip
# or
sudo dnf install xsel

Windows (clip/PowerShell)¶

PowerShellCommand Prompt

# Generate from clipboard and print to stdout
Get-Clipboard | uvx datamodel-codegen --input-file-type jsonschema

# Generate and copy result to clipboard
Get-Clipboard | uvx datamodel-codegen --input-file-type jsonschema | Set-Clipboard

REM Windows Command Prompt doesn't have a built-in paste command
REM Use PowerShell from cmd:
powershell -command "Get-Clipboard" | uvx datamodel-codegen --input-file-type jsonschema

Windows clip command

The clip command on Windows only supports copying to the clipboard, not reading from it. Use PowerShell's Get-Clipboard and Set-Clipboard for full clipboard integration.

Practical Examples¶

Quick model generation workflow¶

Copy a JSON Schema from documentation or an API response
Run the generator from clipboard:

# macOS
pbpaste | uvx datamodel-codegen --input-file-type jsonschema --output-model-type pydantic_v2.BaseModel

# Linux
xclip -selection clipboard -o | uvx datamodel-codegen --input-file-type jsonschema --output-model-type pydantic_v2.BaseModel

# Windows PowerShell
Get-Clipboard | uvx datamodel-codegen --input-file-type jsonschema --output-model-type pydantic_v2.BaseModel

Generating from API documentation¶

# Fetch OpenAPI spec and generate models
curl -s https://petstore3.swagger.io/api/v3/openapi.json | \
  uvx datamodel-codegen --input-file-type openapi --output-model-type pydantic_v2.BaseModel --output petstore.py

Generating from GitHub raw URLs¶

You can directly fetch schemas from GitHub repositories using raw URLs:

# Fetch JSON Schema from GitHub and generate models
curl -s https://raw.githubusercontent.com/OAI/OpenAPI-Specification/main/schemas/v3.0/schema.json | \
  uvx datamodel-codegen --input-file-type jsonschema --output-model-type pydantic_v2.BaseModel

# Fetch OpenAPI spec from a GitHub repository
curl -s https://raw.githubusercontent.com/github/rest-api-description/main/descriptions/api.github.com/api.github.com.json | \
  uvx datamodel-codegen --input-file-type openapi --output-model-type pydantic_v2.BaseModel --output github_api.py

Using the --url option

If you have the http extra installed, you can use --url directly without curl:

uvx --from 'datamodel-code-generator[http]' datamodel-codegen \
  --url https://raw.githubusercontent.com/OAI/OpenAPI-Specification/main/schemas/v3.0/schema.json \
  --input-file-type jsonschema \
  --output-model-type pydantic_v2.BaseModel

Using with jq for JSON manipulation¶

# Extract a specific schema definition and generate a model
cat openapi.yaml | yq '.components.schemas.User' | \
  uvx datamodel-codegen --input-file-type jsonschema

Comparison: Installation Methods¶

Method	Command	Use Case
uvx	`uvx datamodel-codegen`	One-liner usage, no installation
pipx run	`pipx run datamodel-code-generator`	One-liner usage, alternative to uvx
pipx install	`pipx install datamodel-code-generator`	Global installation, frequent usage
uv add	`uv add datamodel-code-generator`	Project dependency
pip install	`pip install datamodel-code-generator`	Traditional installation

When to use each method

uvx/pipx run: Quick one-liner generation, testing different versions
pipx install: Frequent CLI usage across multiple projects
uv add/pip install: Project dependency, CI/CD pipelines, programmatic usage