Full-Stack Entity Generator

Generate a complete FastAPI + SQLAlchemy + Jinja2 application from JSON entity definitions.

This project uses JSON files as a single source of truth and generates:

• Domain entities (dataclasses)
• Application DTOs (Pydantic)
• Repository ports and SQLAlchemy repository implementations
• SQLAlchemy models
• API routers (CRUD) with Bearer JWT authentication
• Frontend routers and HTML templates with custom CSS
• JWT authentication system (login, register, user management)
• i18n scaffolding
• App bootstrap files

Why this project exists

When building CRUD-heavy business apps, the same concepts are repeated in many layers:

• Domain models
• API schemas
• Persistence models
• Web forms and lists
• Authentication and authorization
• Basic translations

This generator removes repetitive boilerplate by deriving those layers from JSON entity specs.

Tech stack

• Python 3.14+ for the generator itself
• Jinja2 templates for code generation
• Generated project stack:
  • FastAPI
  • SQLAlchemy
  • Pydantic
  • Jinja2 templates
  • PyJWT + bcrypt (authentication)
  • Custom CSS (no Bootstrap dependency)
  • SQLite (by default)

Repository layout

fsgenerator/
	cli.py                  # CLI entrypoint with argparse
	parser.py               # Parses app config and entity JSON files
	resolver.py             # Sorts entities by relation dependencies
	type_mapping.py         # Maps JSON field types to Python/SQLAlchemy/Pydantic
	writer.py               # Writes generated files and __init__.py markers
	generators/             # Per-layer file generators
	templates/              # Jinja2 templates used by generators
	sample_entities/        # Starter JSON files for -x/--init scaffolding
json_entities/            # Input JSON entity files (example)
generate.py               # Convenience wrapper for fsgenerator CLI

How generation works

1. Load app config from app_config.json.
2. Load all entity JSON files in the target directory.
3. Validate relation targets.
4. Topologically sort entities (based on many_to_one and one_to_one dependencies).
5. Render templates for each entity and shared files.
6. Write output into a generated project folder named by app_name.

Quick start

1. Install dependencies

uv sync

2. Scaffold a starter project (optional)

uv run generate.py -x

This creates a json_entities/ directory in the current folder with sample JSON files and a README explaining the entity format.

3. Generate using the default input folder

uv run generate.py

4. Generate from a custom folder path

uv run generate.py -i ./json_entities

5. Generate into a custom destination parent directory

uv run generate.py -i ./json_entities -o ./output

This creates ./output/<app_name>.

You can pass any directory that contains:

• app_config.json
• one or more entity JSON files

CLI

usage: fsgenerator [-h] [-x] [-i INPUT_DIR] [-o OUTPUT_DIR]

options:
	-h, --help            show this help message and exit
	-x, --init            Create a json_entities/ directory with sample starter
	                      files, then exit
	-i INPUT_DIR, --input-dir INPUT_DIR
	                      Path to the JSON entities directory (default: ./json_entities)
	-o OUTPUT_DIR, --output-dir OUTPUT_DIR
	                      Parent directory for generated output (default: current directory)

If --input-dir is omitted, it defaults to ./json_entities. If --output-dir is omitted, it defaults to the current directory.

Input format

app_config.json

{
	"app_name": "my_app",
	"id_type": "integer"
}

Fields:

• app_name: Output directory name for the generated app
• id_type: Global ID type for primary and foreign keys
  • integer
  • uuid

Entity JSON example

{
	"name": "company",
	"representation": ["name"],
	"fields": {
		"name": {"type": "string", "maxLength": 80},
		"afm": {"type": "string_numeric", "length": 9}
	},
	"relations": {
		"parent": {"type": "many_to_one", "entity": "company"}
	},
	"unique": {
		"unique_name": ["name"],
		"unique_parent_name": ["parent", "name"]
	}
}

Entity schema details

Root keys

• name: Entity name (used for filenames, routes, table name)
• representation: Fields used for display labels in dropdowns and list views
• fields: Dictionary of scalar fields
• relations: Dictionary of relations
• unique: Named unique constraints

Field options

Option	Type	Default	Description
type	string	—	Required. One of the supported types below
maxLength	integer	—	Maximum string length
minLength	integer	—	Minimum string length validation
length	integer	—	Exact string length (sets both min and max)
isnull	boolean	false	Whether null is allowed (makes the field optional)
options	array	—	Required for select type. List of [value, "label"] pairs

Relation options

• type: many_to_one (currently fully supported in generation)
• entity: Target entity name (must match a JSON filename)

Supported scalar field types

Type	Python type	SQLAlchemy type	Notes
integer	int	Integer
string	str	String(N)	Use maxLength to set length
string_numeric	str	String(N)	Adds ^\d+$ regex validation
iso-date	datetime.date	Date
iso-datetime	datetime.datetime	DateTime
time	datetime.time	Time
value	Decimal	Numeric(10,2)	For monetary values
float	float	Float
percent	Decimal	Numeric(5,2)	For percentage values
key	int	Integer	Generic integer key
select	int	Integer	Requires options field

Mapping behavior is defined in fsgenerator/type_mapping.py.

Relationship behavior

• Relation definitions are parsed for all listed relation types.
• Code generation currently emits foreign key columns and object relationships for:
  • many_to_one
  • one_to_one
• Dependency sorting is based on many_to_one and one_to_one relations.
• Relation fields are placed before data fields in all generated files (entities, DTOs, models, templates).

Unique constraints

Named unique constraints are generated at SQLAlchemy table level.

When a unique constraint references a relation field name, the generator maps it to the corresponding foreign key column name (for example, parent -> parent_id).

Authentication

The generated application includes a complete JWT-based authentication system:

• Login/Register pages with bcrypt password hashing
• JWT tokens stored in HTTP-only cookies (60-minute expiry)
• Auth middleware that redirects unauthenticated users to the login page
• API Bearer auth — all REST API endpoints require a valid JWT in the Authorization header
• Admin user seeded on first startup (email: admin@admin.com, password: admin)
• User management — admin users can toggle active/admin status of other users
• Change password — authenticated users can change their own password

Public paths (login, register, static files, API docs) are accessible without authentication.

Generated output structure

For each entity named example, the generator produces:

• domain/entities/example.py
• application/dtos/example.py
• domain/ports/example.py
• application/use_cases/example.py
• infrastructure/persistence/sqlalchemy/models/example.py
• infrastructure/persistence/sqlalchemy/repositories/example.py
• infrastructure/web/fastapi/routers/example.py (API)
• infrastructure/web/fastapi/routers/example_frontend.py
• templates/example_list.html
• templates/example_form.html

Shared files:

• main.py (FastAPI app with auth middleware and admin seeding)
• pyproject.toml
• static/style.css (custom CSS design system)
• infrastructure/web/fastapi/dependencies.py
• infrastructure/web/fastapi/routers/i18n.py
• infrastructure/persistence/sqlalchemy/models/__init__.py
• infrastructure/persistence/sqlalchemy/models/auth_models.py
• auth_security.py (JWT + bcrypt utilities)
• infrastructure/web/fastapi/routers/auth.py
• templates/base.html
• templates/home.html
• templates/login.html
• templates/register.html
• templates/change_password.html
• templates/admin_users.html
• i18n_helper.py
• i18n/en.json

The generator also creates missing __init__.py files in generated package directories.

i18n

The generator creates i18n/en.json with:

• Common UI and error labels
• Per-entity labels
• Per-field labels
• Labels for many_to_one and one_to_one relation selectors
• Authentication labels (login, register, password, user management)

The generated app includes:

• Translation loading helper
• Language-switch router
• HTML templates that call translation keys

Running the generated app

After generation, change into the generated app folder (the folder named by app_name), install dependencies, and run with uvicorn.

Typical flow:

cd my_app
uv sync
uv run uvicorn main:app --reload

Default database is SQLite at app.db. An admin user is created on first startup.

Default admin credentials:

• Email: admin@admin.com
• Password: admin

Customizing output

Edit templates in fsgenerator/templates/ to change generated code shape:

• API behavior
• SQLAlchemy model patterns
• HTML layout and UX
• Authentication flow
• Dependency wiring
• CSS styling (static/style.css)
• i18n runtime helper

After template changes, run the generator again.

Known limitations

• The generator focuses on CRUD scaffolding.
• many_to_many and one_to_many relations are parsed but are not fully emitted as SQLAlchemy relationship collections in current templates.
• primary_key inside field definitions is parsed but global ID generation from app_config.id_type is used.
• No migration generation is included (for example, Alembic).

Troubleshooting

• Error about unknown relation target:
  • Ensure each relation entity points to an existing entity JSON file name.
• Missing app_config.json:
  • Ensure the target input folder contains app_config.json.
• Unexpected type behavior:
  • Check mappings in fsgenerator/type_mapping.py.
• Wrong output folder:
  • Check app_name in app_config.json.
• Delete errors showing raw JSON:
  • Entity delete endpoints display errors on the list page when a record has dependent relations.

Version

Current version: 0.5.0