erdify 0.3.1

Generate PlantUML ERD diagrams from SQLModel, SQLAlchemy, Pydantic and dataclass models

🗃️ erdify

🚀 Generate beautiful PlantUML Entity Relationship Diagrams from your SQLModel, SQLAlchemy, Pydantic and dataclass models automatically!

erdify parses your model files using AST (Abstract Syntax Tree) and generates comprehensive ERD diagrams in PlantUML format. It supports SQLModel, SQLAlchemy 2.0, Pydantic and standard-library dataclasses. No database connection required!

✨ Features

• 📊 Automatic ERD Generation - Parse your models and generate PlantUML diagrams
• 🧬 4 Frameworks - SQLModel, SQLAlchemy 2.0 (Mapped[...]/mapped_column()), Pydantic and dataclasses
• 🔍 AST-Based Parsing - No imports needed, works with any valid Python code
• 🎯 Zero Runtime Dependencies - Uses only Python standard library
• 🔗 Relationship Detection - Automatically detects foreign keys and relationships
• 🔑 Key Inference - Optional --infer-keys derives PK/FK from field names for keyless models
• 🚫 Exclude Patterns - Filter out entities by class or table name with glob patterns
• 📦 Inheritance Support - Correctly resolves fields from base classes and mixins
• 🏷️ Enum Support - Includes enum definitions in the diagram
• 🔄 Link Table Detection - Identifies many-to-many relationship tables
• 🎨 Beautiful Output - Clean, readable PlantUML with proper styling

📦 Installation

Using pip

pip install erdify

Using uv

uv add erdify

Using pipx (recommended for CLI usage)

pipx install erdify

Using uvx (no installation needed)

# Run directly without installing
uvx erdify ./src/database -o erd.puml

🚀 Quick Start

Command Line

# Generate ERD from a models directory
erdify ./src/database -o erd.puml

# With custom title
erdify ./src/models --title "My Database Schema" -o schema.puml

# Output to stdout
erdify ./src/database

Python API

from pathlib import Path
from erdify import parse_models_directory, generate_plantuml

# Parse your models
entities, enums = parse_models_directory(Path("./src/database"))

# Generate PlantUML
diagram = generate_plantuml(
    entities=entities,
    enums=enums,
    title="My Database ERD"
)

# Save or use the diagram
Path("erd.puml").write_text(diagram)

📖 Usage

CLI Options

usage: erdify [-h] [-o OUTPUT] [--title TITLE] [--exclude [PATTERN ...]]
                    [--infer-keys] [--no-enums] [--no-relationships] [-v]
                    input

Generate PlantUML ERD diagrams from SQLModel, SQLAlchemy, Pydantic and dataclass models

positional arguments:
  input                 Directory containing model files (searches for models.py recursively)

options:
  -h, --help            show this help message and exit
  -o OUTPUT, --output OUTPUT
                        Output .puml file (default: stdout)
  --title TITLE         Diagram title (default: 'Database ERD')
  --exclude [PATTERN ...]
                        Glob patterns (case-sensitive) to exclude entities by
                        class name or table name, e.g. --exclude '*Link' audit_log
  --infer-keys          For keyless models (Pydantic/dataclass), infer a primary
                        key from a field named 'id' and a foreign key from '<x>_id'
  --no-enums            Skip enum definitions in output
  --no-relationships    Skip relationship lines in output
  -v, --version         show program's version number and exit

Excluding Entities

Use --exclude to drop tables/entities from the diagram. Each pattern is a case-sensitive glob tested against both the class name and the table name — an entity is excluded if either matches. Any relationships pointing at an excluded entity are dropped too, so no dangling lines remain.

# Exclude all link tables (class names ending in "Link")
erdify ./src/database --exclude '*Link'

# Exclude by table name, with multiple patterns
erdify ./src/database --exclude audit_log 'tmp_*' Session

💡 Quote patterns containing * so your shell doesn't expand them.

Running as Module

python -m erdify ./src/database -o erd.puml

Example Models

Given these SQLModel definitions:

from enum import Enum
from sqlmodel import SQLModel, Field, Relationship

class UserRole(Enum):
    ADMIN = "admin"
    USER = "user"

class User(SQLModel, table=True):
    __tablename__: str = "user"

    id: int = Field(primary_key=True)
    name: str
    email: str = Field(index=True)
    role: UserRole = Field(default=UserRole.USER)

    orders: list["Order"] = Relationship(back_populates="user")

class Order(SQLModel, table=True):
    __tablename__: str = "order"

    id: int = Field(primary_key=True)
    user_id: int = Field(foreign_key="user.id")
    total: float

    user: "User" = Relationship(back_populates="orders")

The tool generates:

with following code:

@startuml Database ERD
!define primary_key(x) <b><color:#b8861b><&key></color> x</b>
!define foreign_key(x) <color:#aaaaaa><&key></color> x
!define column(x) <color:#efefef><&media-record></color> x

skinparam linetype ortho

' Enums
enum UserRole << (E,#FFCC00) >> {
  ADMIN
  USER
}

' Entities
entity "user" as User {
  primary_key(id) : int
  column(name) : str
  column(email) : str
  column(role) : UserRole = USER
}

entity "order" as Order {
  primary_key(id) : int
  foreign_key(user_id) : int
  column(total) : float
}

' Relationships
Order }o--|| User : "user_id"

@enduml

🧬 One Schema, Four Frameworks

erdify supports four model frameworks. The snippets below all describe the same User / Order schema — only the syntax differs. Each one produces the identical diagram:

ℹ️ The SQLModel and SQLAlchemy versions declare keys explicitly. Pydantic and dataclasses have no key concept, so they are rendered with --infer-keys (id → PK, <x>_id → FK) to match. The runnable sources live in docs/examples/.

SQLModel	SQLAlchemy 2.0
from sqlmodel import ( Field, Relationship, SQLModel, ) class User(SQLModel, table=True): __tablename__: str = "user" id: int = Field(primary_key=True) name: str email: str orders: list["Order"] = Relationship( back_populates="user") class Order(SQLModel, table=True): __tablename__: str = "order" id: int = Field(primary_key=True) user_id: int = Field( foreign_key="user.id") total: float user: "User" = Relationship( back_populates="orders")	from sqlalchemy import ForeignKey from sqlalchemy.orm import ( DeclarativeBase, Mapped, mapped_column, relationship, ) class Base(DeclarativeBase): ... class User(Base): __tablename__ = "user" id: Mapped[int] = mapped_column( primary_key=True) name: Mapped[str] = mapped_column() email: Mapped[str] = mapped_column() orders: Mapped[list["Order"]] = ( relationship(back_populates="user")) class Order(Base): __tablename__ = "order" id: Mapped[int] = mapped_column( primary_key=True) user_id: Mapped[int] = mapped_column( ForeignKey("user.id")) total: Mapped[float] = mapped_column() user: Mapped["User"] = relationship( back_populates="orders")
Pydantic --infer-keys	Dataclass --infer-keys
from pydantic import BaseModel class User(BaseModel): id: int name: str email: str orders: list["Order"] = [] class Order(BaseModel): id: int user_id: int total: float user: "User"	from dataclasses import dataclass, field @dataclass class User: id: int name: str email: str orders: list["Order"] = field( default_factory=list) @dataclass class Order: id: int user_id: int total: float user: "User" = None

How each framework is detected & parsed:

Framework	Detected by	Keys	Relationships
SQLModel	table=True	Field(primary_key=…, foreign_key=…)	Relationship()
SQLAlchemy 2.0	__tablename__ + Mapped[...] columns	mapped_column(primary_key=…), ForeignKey(...)	relationship() (lowercase)
Pydantic	BaseModel subclass (incl. transitive)	--infer-keys only	nested model refs (user: User, list["Order"])
Dataclass	@dataclass decorator	--infer-keys only	nested model refs

ℹ️ Mixins / abstract bases (e.g. a SQLAlchemy mixin without __tablename__) are not drawn as tables, but their columns are inherited into concrete entities. Imports aliased to other names (e.g. mapped_column as mc) are not detected.

Inferring keys (--infer-keys)

Pydantic models and dataclasses have no database key concept. By default all fields are rendered as plain columns and relationships come only from nested model references. If your models follow a database-like naming convention, pass --infer-keys to derive keys from field names:

• a field named id → primary key
• a field named <x>_id → foreign key targeting table <x>

# Plain columns (default)
erdify ./src/schemas

# Infer PK/FK from id / <x>_id naming
erdify ./src/schemas --infer-keys

ℹ️ --infer-keys only affects Pydantic/dataclass models. SQLModel and SQLAlchemy keys are always read from the explicit definitions and never overridden.

🎨 Viewing the Diagram

Online

1. Copy the generated .puml content
2. Paste at PlantUML Web Server

Local with PlantUML

# Install PlantUML (macOS)
brew install plantuml

# Generate PNG
plantuml erd.puml

# Generate SVG
plantuml -tsvg erd.puml

VS Code Extension

Install the PlantUML extension for live preview.

🔧 Advanced Usage

Programmatic Access

from erdify import (
    ASTDatabaseParser,
    PlantUMLGenerator,
    EntityInfo,
    FieldInfo,
    EnumInfo,
)

# Low-level parser access
parser = ASTDatabaseParser(Path("./models"))
entities, enums = parser.parse_all_models()

# Access entity details
for name, entity in entities.items():
    print(f"Table: {entity.table_name}")
    for field in entity.fields:
        if field.is_primary_key:
            print(f"  PK: {field.name}")
        elif field.is_foreign_key:
            print(f"  FK: {field.name} -> {field.foreign_table}")

# Custom generator with options
generator = PlantUMLGenerator(
    entities=entities,
    enums=enums,
    title="Custom ERD"
)
output = generator.generate()

Integration with CI/CD

# .github/workflows/docs.yml
name: Generate ERD

on:
  push:
    paths:
      - 'src/database/**'

jobs:
  generate-erd:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Set up Python
        uses: actions/setup-python@v5
        with:
          python-version: '3.12'

      - name: Install erdify
        run: pip install erdify

      - name: Generate ERD
        run: erdify ./src/database --title "Database Schema" -o docs/erd.puml

      - name: Generate PNG
        run: |
          sudo apt-get install -y plantuml
          plantuml docs/erd.puml

      - name: Commit changes
        uses: stefanzweifel/git-auto-commit-action@v5
        with:
          commit_message: "docs: update ERD diagram"
          file_pattern: "docs/erd.*"

Integration with pre-commit hooks

Keep your ERD diagrams automatically updated on every commit using pre-commit:

# .pre-commit-config.yaml
repos:
  - repo: local
    hooks:
      - id: generate-erd
        name: 🗃️ Generate ERD Diagram
        entry: erdify ./src/database --title "Database Schema" -o docs/erd.puml
        language: system
        files: ^src/database/.*\.py$
        pass_filenames: false

Or using uvx (no installation required):

# .pre-commit-config.yaml
repos:
  - repo: local
    hooks:
      - id: generate-erd
        name: 🗃️ Generate ERD Diagram
        entry: uvx erdify ./src/database --title "Database Schema" -o docs/erd.puml
        language: system
        files: ^src/database/.*\.py$
        pass_filenames: false

Setup:

# Install pre-commit
pip install pre-commit

# Install the hooks
pre-commit install

# Run manually on all files
pre-commit run generate-erd --all-files

How it works:

• 🔍 Only triggers when files in src/database/ change
• 📝 Automatically regenerates docs/erd.puml
• ✅ Stages the updated diagram with your commit
• 🚫 Fails if the diagram would change (ensuring docs stay in sync)

Tip: Add docs/erd.puml to your staged files before committing, or use the --all-files flag to regenerate.

📋 Supported Features

Feature	Status	Notes
Primary Keys	✅	Field(primary_key=True)
Foreign Keys	✅	Field(foreign_key="table.column")
Nullable Fields	✅	str | None or Optional[str]
Default Values	✅	Field(default=value)
Indexes	✅	Field(index=True)
Enums	✅	Python Enum classes
Relationships	✅	Relationship()
Inheritance	✅	Mixin classes supported
Link Tables	✅	Many-to-many detection
Custom Table Names	✅	__tablename__ attribute
Exclude Patterns	✅	--exclude glob on class/table name
Key Inference	✅	--infer-keys for Pydantic/dataclass (id, <x>_id)
SQLModel	✅	Field() / Relationship()
SQLAlchemy 2.0	✅	Mapped[...] / mapped_column()
Pydantic	✅	BaseModel subclasses, nested refs as relationships
Dataclass	✅	@dataclass, nested refs as relationships

🗺️ Roadmap

Recently shipped:

Feature	Status	Description
Exclude Option	✅ Done	Exclude tables or entities from ERD generation using glob patterns
SQLAlchemy Support	✅ Done	Native support for SQLAlchemy 2.0 (Mapped / mapped_column) models
Pydantic Support	✅ Done	Generate ERDs from Pydantic models, with optional --infer-keys
Dataclass Support	✅ Done	Support for standard Python dataclasses with type annotations

Have a feature request? Please open an issue on GitHub to discuss it!

🤝 Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

🔒 Security

For security concerns, please see SECURITY.md.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

erdify stands on the shoulders of great open-source projects:

Supported model frameworks

• SQLModel - The awesome SQL database library
• SQLAlchemy - The Python SQL toolkit and ORM
• Pydantic - Data validation using Python type hints
• dataclasses - Python standard-library data classes

Rendering & output

• PlantUML - For the diagram rendering
• Graphviz - Layout engine behind PlantUML's ER diagrams

Tooling & infrastructure

• uv - Packaging, builds and dependency management
• Ruff - Linting and formatting
• mypy - Static type checking
• pytest - Testing framework
• pre-commit - Git hook management

Community

• Contributor Covenant - Our Code of Conduct
• Keep a Changelog - Changelog format
• And everyone who contributes issues, ideas and pull requests 💜

---

Made with ❤️ by Devsuit GmbH