FastAPI-Restly

Build maintainable REST APIs on FastAPI, SQLAlchemy 2.0, and Pydantic v2 — with real class-based views.

Status: 0.5.1 — public beta release.

After four years of internal development at two separate companies, Restly is finally ready for its first public release! Right now the goal is to see if the public API of Restly hits the right abstractions, and to stabilize the API for a 1.0.0 release. From 0.5.0 onwards, expect small breaking changes in naming and functionality on the deeper parts of the API surface. Feedback is always appreciated!

pip install "fastapi-restly[standard]"

Docs: https://rjprins.github.io/fastapi-restly/ · Changelog · Contributing · Security · Examples

Why FastAPI-Restly?

Restly helps building FastAPI apps faster, with consistent APIs.

It features class-based views that support inheritance, mixins, and method overrides.

Class-based views are essential for re-using code. The RestView and AsyncRestView provide full CRUD on top of a SQLAlchemy model with a single class declaration. It stays fully customizable by overriding endpoints, perform_* handlers, and other class methods.

• class-based views: group endpoints on real Python classes with inheritance and method overrides.
• REST endpoints in minutes: use View for custom resources, or AsyncRestView / RestView for generated CRUD.
• Incremental adoption: Restly doesn't get in your way; use it per resource and step out anytime. See Existing Projects.
• Class-level dependencies: Put dependencies that all endpoints need on the class level, and get them as attributes on self.
• Explicit override points: Call-chain allows for overriding at multiple levels.
• Filtering, pagination, sorting: Get fully-featured list routes specific to your Pydantic schema.
• Field control: ReadOnly / WriteOnly markers, plus foreign-key validation in Pydantic schemas via IDRef[...].
• React Admin ready: AsyncReactAdminView speaks the ra-data-simple-rest wire contract, no custom data provider needed.
• General app utilities: Things most FastAPI apps will need: SQLAlchemy engine and session setup, alembic test fixtures, etc.

Quickstart

FastAPI-Restly turns a SQLAlchemy model into a class-based CRUD resource:

import fastapi_restly as fr
from fastapi import FastAPI
from sqlalchemy.orm import DeclarativeBase, Mapped, mapped_column

app = FastAPI()

class Base(DeclarativeBase):
    pass

class User(Base):
    __tablename__ = "user"

    id: Mapped[int] = mapped_column(primary_key=True)
    name: Mapped[str]
    email: Mapped[str]

@fr.include_view(app)
class UserView(fr.AsyncRestView):
    prefix = "/users"
    model = User

That view exposes these HTTP routes:

GET    /users/       # list users, with filtering, sorting, and pagination
POST   /users/       # create a user
GET    /users/{id}   # read one user
PATCH  /users/{id}   # partially update one user
DELETE /users/{id}   # delete one user

Restly generates the Pydantic schemas automatically. For the full copy-paste app see Getting Started.

Installation (development)

git clone https://github.com/rjprins/fastapi-restly.git
cd fastapi-restly
uv sync

Main features

Manual schema definition

For custom validation, aliases, or stable public contracts, define an explicit read schema:

from datetime import datetime

class UserRead(fr.IDSchema):
    name: str
    email: str
    password: fr.WriteOnly[str]
    created_at: fr.ReadOnly[datetime]

@fr.include_view(app)
class UserView(fr.AsyncRestView):
    prefix = "/users"
    model = User
    schema = UserRead
    # create_schema = UserCreate  # auto-generated from UserRead
    # update_schema = UserUpdate  # auto-generated from UserCreat

Restly derives create and update schemas from UserRead by default. The UserCreate schema is created by omitting ReadOnly fields. The UserUpdate schema allows for partial updates by making all fields optional.

When you need full control over write payloads, declare them explicitly:

class UserCreate(fr.BaseSchema):
    name: str
    email: str

class UserUpdate(fr.BaseSchema):
    name: str | None = None
    email: str | None = None

@fr.include_view(app)
class UserView(fr.AsyncRestView):
    prefix = "/users"
    model = User
    schema = UserRead
    creation_schema = UserCreate
    update_schema = UserUpdate

Use auto-schema for prototypes and internal tools. Use an explicit schema when contract stability and validation control matter (public APIs, aliases, strict response shapes).

List endpoint query parameters

List endpoints expose a stable URL parameter dialect generated from the response schema:

GET /users/?name=John&age__gte=21
GET /users/?status=active,pending           # comma-separated → OR (IN)
GET /users/?status__ne=archived,deleted     # comma-separated → NOT IN
GET /users/?email__icontains=example
GET /users/?deleted_at__isnull=true
GET /users/?sort=-created_at,name
GET /users/?page=2&page_size=10

Parameter keys follow the response schema's public names end-to-end — including dotted relation paths. If ArticleRead.author has Field(alias="writer") and AuthorRead.name has Field(alias="authorName"), the URL key is writer.authorName. Aliased fields are only reachable by their alias; populate_by_name does not extend the URL surface with the Python field name.

Pagination is opt-in: omitting page_size returns every matching row. For public/production endpoints set default_page_size and max_page_size on the view class:

class UserView(fr.AsyncRestView):
    default_page_size = 25
    max_page_size = 200

See How-To: Filter, Sort, and Paginate Lists for the full operator surface, alias rules, and pagination guidance.

Read-only and write-only fields

IDSchema already provides a read-only id, so don't redeclare it unless you need to narrow the type.

class UserRead(fr.IDSchema):
    name: str
    email: str
    password: fr.WriteOnly[str]        # stripped by to_response_schema()
    created_at: fr.ReadOnly[datetime]  # skipped in `apply_schema()`

Relationship handling

Validate relationships on create and update using fr.IDRef[...]. SQLAlchemy init is handled smartly; init is with either the foreign key (customer_id) or the related object (Customer), whichever is in the signature of the SQLAlchemy mapper __init__.

class Order(fr.IDBase):
    customer_id: Mapped[int] = mapped_column(ForeignKey("customer.id"))
    customer: Mapped[Customer] = relationship()

class OrderRead(fr.IDSchema):
    customer_id: fr.IDRef[Customer]
    customer: fr.ReadOnly[CustomerRead]

Custom endpoints

Add custom routes using the same form of decorators you would use for regular FastAPI routes.

• @fr.get
• @fr.post
• @fr.put
• @fr.patch
• @fr.delete
• @fr.route

These simply forward all arguments to their standard FastAPI counterparts.

class UploadView(fr.AsyncRestView):
    prefix = "/uploads"
    model = Upload

    @fr.get(
        "/{id}/download",
        response_class=FileResponse,
        responses={200: {"content": {EXCEL_MIME_TYPE: {}}}},
    )
    async def download_excel(self, id: int):
        upload = await self.perform_get(id)
        return to_excel_response(upload)

React Admin integration

Use AsyncReactAdminView to get a backend that react-admin with ra-data-simple-rest connects to out of the box:

@fr.include_view(app)
class ProductView(fr.AsyncReactAdminView):
    prefix = "/products"
    model = Product
    schema = ProductRead

The view speaks the ra-data-simple-rest wire contract.

See React Admin Integration in the docs for CORS setup and customization.

Excluding built-in routes

@fr.include_view(app)
class UserView(fr.AsyncRestView):
    prefix = "/users"
    model = User
    exclude_routes = (fr.ViewRoute.DELETE,)

Pagination metadata

@fr.include_view(app)
class UserView(fr.AsyncRestView):
    prefix = "/users"
    model = User
    include_pagination_metadata = True
    # Response: {"items": [...], "total": N, "page": 1, "page_size": 100, "total_pages": N, ...}

Testing

fastapi_restly.pytest_fixtures provides namespaced pytest fixtures (restly_app, restly_client, restly_async_session, restly_session) for test clients and savepoint-based isolation. The testing extra installs a pytest plugin entry point, so pytest auto-loads these fixtures.

Install the testing extra when consuming FastAPI-Restly as a package:

pip install "fastapi-restly[testing]"

Configure Restly for your test database in conftest.py.

RestlyTestClient automatically asserts the expected HTTP status (200 for GET, 201 for POST, 204 for DELETE, ...) and raises a descriptive AssertionError with the response body on failure:

# test_users.py
def test_create_and_fetch_user(restly_client):
    # Raises AssertionError if status != 201
    response = restly_client.post("/users/", json={"name": "John", "email": "john@example.com"})
    user_id = response.json()["id"]

    # Raises AssertionError if status != 200
    data = restly_client.get(f"/users/{user_id}").json()
    assert data["name"] == "John"

Pass assert_status_code=None to skip the assertion and inspect the response yourself.

Configuration

# Async SQLite
fr.configure(async_database_url="sqlite+aiosqlite:///app.db")

# Async PostgreSQL
fr.configure(async_database_url="postgresql+asyncpg://user:pass@localhost/db")

# Sync SQLite
fr.configure(database_url="sqlite:///app.db")

Restly has one public process-wide configuration. For per-view databases, read replicas, or other custom session wiring, use a normal FastAPI dependency on that view; see the existing-project how-to in the documentation.

Documentation

• Getting Started — fast path from zero to a working API
• User Guide — tutorial walkthroughs and topic guides
• API Reference — complete API docs

Examples

Complete applications under example-projects/:

• Shop — e-commerce API with products, orders, customers
• Blog — minimal blog with a single Blog model
• SaaS — multi-tenant project management API

Contributing

Pull requests and issue discussions welcome. See CONTRIBUTING.md for setup, coding standards, and the test workflow. Security issues: see SECURITY.md.

License

MIT — see LICENSE.