FastAPI

Created by Sebastian Ramirez, FastAPI lets you define HTTP endpoints as plain Python functions whose annotated parameters drive request parsing, validation, serialization, and documentation at once. Path, query, header, cookie, form, and file inputs are declared in the function signature, request bodies are Pydantic models, and the framework generates an OpenAPI schema served as Swagger UI at /docs and ReDoc at /redoc. It sits on Starlette, so sync or async handlers, WebSockets, background tasks, CORS, and HTTPX-based testing come included, and a dependency injection system manages shared resources such as database sessions or OAuth2 and JWT auth. ML teams use it to put models behind REST or streaming endpoints, and servers such as TabbyAPI build on it. Install fastapi[standard] with pip or uv on Python 3.10 or later to get Uvicorn and the fastapi CLI, whose dev command auto-reloads and whose run command serves production. MIT licensed with no commercial strings; the team's commercial FastAPI Cloud hosting funds the project, but any host works. Best for Python developers turning models or data services into typed APIs.