Python (FastAPI)
fastapi.tiangolo.com- Category
- Developer Tools
- Rank
- No. 257Tools index
- Pricing
- Open Source
- Type
- TOOL
- Builder
- fastapi
- GitHub
- 102.4k stars
- Latest release
- 0.141.1
- Date
About
Modern Python async web framework. High-performance Python framework based on type hints and ASGI. Auto-generated OpenAPI docs. Common pick for AI/ML APIs.
What it does
FastAPI is a Python web framework that reads the type annotations on your handler functions and derives the rest from them: request parsing, validation, error responses, and an OpenAPI schema with interactive docs. Routing and the HTTP layer come from Starlette, data handling from Pydantic, so the package itself is mostly glue that turns annotated parameters into declared inputs. Those declarations cover query strings, path segments, headers, cookies, form fields and uploads. A companion command line tool launches an app under Uvicorn with auto reload.
Why it's ranked here
The type hint bet paid off. One annotation buys parsing, validation, the error response and the schema entry, and the package keeps that promise on five runtime dependencies. The engineering around it is disciplined: strict type checking is configured, the test settings turn every warning into an error, and each test is capped at twenty seconds. Two things keep this honest. The headline productivity numbers, 200 to 300 percent faster and 40 percent fewer errors, are footnoted as one internal team's estimate. And the packaging still declares beta status.
What's good
The import surface is small and flat: the application class, the router, the dependency marker and the parameter declarations all come from one place, with HTTP status codes re-exported from Starlette so you do not need a second import. Validation goes deep enough to cover self referential models that rebuild before the schema is produced. Documentation examples live in the repository as runnable applications, and the coverage configuration measures them, so the tutorials are executed code rather than prose. Linting, strict typing and a spell checker are all configured in the manifest.
Tradeoffs
The framework leans hard on two upstream projects, so a major shift in either lands on you. Python 3.10 is the floor. The command line is not in the base install: without the standard extra, the entry point prints an install hint and raises. That base install also leaves out the test client, templates, form and upload parsing, and email validation. The project front page carries a long sponsor block, including a keystone slot for a commercial cloud product run by the same team, which is worth knowing when you read the project's own framing.
How to use it well
Reach for this when you are putting a typed HTTP interface in front of Python code: model serving endpoints, internal services, anything where a generated schema saves you hand writing client docs. Install the standard extra from the start, since the runner, test client and form handling live there. Declare parameter and response types explicitly instead of returning loose dictionaries, because that is what feeds the schema. It gives you no database layer, no admin interface and no job queue: only short in request background work ships in the box.
Technical notes+
The package root re-exports the whole public surface from one module, fastapi/__init__.py, which pulls in FastAPI, APIRouter, BackgroundTasks, UploadFile, HTTPException, WebSocket, Request, Response, the parameter helpers Body, Query, Path, Header, Cookie, Form, File, Depends and Security, plus Starlette's status module. fastapi/cli.py is a thin shim that tries to import the external fastapi-cli package and, when that import fails, prints an install message and raises RuntimeError, while fastapi/__main__.py simply calls it; pyproject.toml wires a console script named fastapi to that shim. pyproject.toml also declares MIT, a Python 3.10 floor, the runtime dependencies starlette, pydantic, typing-extensions, typing-inspection and annotated-doc, extras named standard and all, mypy in strict mode with the pydantic plugin, ruff lint rules, pytest with warnings promoted to errors and a 20 second timeout, and coverage collected over the library, tests and doc examples. tests/main.py exercises path and query constraints such as min_length and numeric bounds, tests/test_validate_response_recursive/app.py covers self referential response models rebuilt before use, and docs_src/app_testing/app_b_an_py310/main.py shows Annotated header parameters with typed error paths. docs/language_names.yml and docs/missing-translation.md carry the translation setup, the latter announcing a move to automated translation.
Observed
- License
- MIT, declared in pyproject.toml
- Language
- Python, typed package built with the PDM backend
- Platform support
- Python 3.10 or newer, per requires-python in pyproject.toml
- Interfaces
- Importable library plus a fastapi console script entry point
- Install surface
- Base install plus optional extras named standard and all
- Core dependencies
- Starlette for the HTTP layer, Pydantic for validation
- Type checking
- mypy strict mode with the pydantic plugin, configured in pyproject.toml
- Test settings
- pytest with warnings treated as errors and a 20 second per test timeout
- Declared maturity
- Development Status 4 - Beta in the packaging classifiers
- Docs
- Language codes mapped in docs/language_names.yml, with an automated translation notice in docs/missing-translation.md
Read from README.md, pyproject.toml, fastapi/__init__.py, fastapi/__main__.py, fastapi/cli.py, docs/en/docs/index.md, tests/main.py, docs_src/app_testing/app_b_an_py310/main.py, docs_src/async_tests/app_a_py310/main.py, tests/test_validate_response_recursive/app.py, docs/language_names.yml, docs/missing-translation.md.
Tags
Tech Stack
Media
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.