Backend6 min read

FastAPI

High-performance Python API framework - automatic OpenAPI docs, Pydantic validation, async/await support, and type annotations that serve as the schema.

FastAPIPythonREST APIPydanticAsync

skill: FastAPI category: Backend tags: [FastAPI, Python, REST API, Pydantic, Async] readingTime: 6

FastAPI is the default choice for Python REST APIs - it's faster than Flask, generates automatic OpenAPI docs, and enforces types via Pydantic. For ML-serving specifically, FastAPI + Uvicorn is the standard stack. Django REST Framework is better for Django ORM-heavy apps.

Workflow

Set up FastAPI with UvicornUvicorn is the ASGI server that runs FastAPI - async-capable, production-grade, and the recommended server from the FastAPI docs.
Define request/response models with PydanticPydantic models are FastAPI's schema layer - they validate incoming data, generate OpenAPI docs, and provide type annotations in one declaration.
Use dependency injection for shared logicFastAPI's dependency system handles DB sessions, authentication, and pagination without repeating code in every endpoint.
Write async endpoints for I/O-bound tasksAsync endpoints don't block the event loop on network I/O - enabling FastAPI to handle many concurrent requests on a single process.
Run background tasksBackground tasks run after the response is sent - useful for sending emails, logging, or triggering webhooks without making the client wait.

Output Format

FastAPI app with Pydantic models, dependency injection, async endpoints, and background tasks.

app/
├── main.py           # FastAPI instance, router includes
├── api/
│   ├── users.py      # APIRouter for /users
│   └── auth.py
├── models/
│   ├── schemas.py    # Pydantic request/response models
│   └── db.py         # SQLAlchemy models
├── deps.py           # Shared dependencies
└── core/
    └── config.py

Examples

Input

Output

Edge Cases & Gotchas

Blocking calls in async endpoints - Using sync DB calls (psycopg2, sqlite3) in async endpoints blocks the thread pool. Use asyncpg or SQLAlchemy async session for true async I/O.
Pydantic v1 vs v2 - FastAPI 0.100+ uses Pydantic v2. Many tutorials show v1 syntax (.dict() vs .model_dump(), Config class vs model_config). Mixing them causes errors.
Response model leaking sensitive fields - If response_model is the DB model (not a separate schema), it may serialize password hashes or internal fields. Always use a dedicated response schema.
CORS for browser clients - Add CORSMiddleware from fastapi.middleware.cors before any other middleware. Without it, browser fetch to a different port fails silently.

When to Break These Rules

Use Django + DRF when the project is data-heavy with complex ORM relationships, admin panel requirements, or built-in auth and permissions.
Use Flask for very simple scripts or internal tools where FastAPI's structure is overkill.

Reference Files

https://fastapi.tiangolo.com/
https://docs.pydantic.dev/