03. FastAPI Basics

Chapter 3 of 18 · 20 min

KEY INSIGHT

FastAPI handles the plumbing that every web service needs: routing, request parsing, response serialization, and error handling. Learning to use its type system correctly eliminates an entire class of bugs where invalid data reaches your business logic. ### Project Structure A minimal FastAPI project requires an application instance, defined routes, and a run command. Organize code so that the FastAPI app object imports from separate modules rather than containing all logic inline. ``` app/ Γö£ΓöÇΓöÇ main.py Γö£ΓöÇΓöÇ routers/ Γöé Γö£ΓöÇΓöÇ completions.py Γöé ΓööΓöÇΓöÇ embeddings.py Γö£ΓöÇΓöÇ models/ Γöé Γö£ΓöÇΓöÇ requests.py Γöé ΓööΓöÇΓöÇ responses.py ΓööΓöÇΓöÇ services/ ΓööΓöÇΓöÇ inference.py ``` The `main.py` file creates the FastAPI instance, includes routers, and configures middleware. Routers group related endpoints together. Models define request and response schemas using Pydantic. Services contain the business logic that routers call. ### Defining a Route ```python from fastapi import FastAPI, HTTPException from pydantic import BaseModel app = FastAPI() class CompletionRequest(BaseModel): model: str messages: list[dict] temperature: float = 0.7 max_tokens: int | None = None @app.post("/v1/chat/completions") async def create_completion(request: CompletionRequest): if not request.messages: raise HTTPException(status_code=400, detail="messages cannot be empty") return {"choices": [{"message": {"content": "response"}}]} ``` The `async def` syntax allows the endpoint to handle concurrent requests without blocking. The Pydantic model validates the request body automatically. Invalid requests receive a 422 response without any custom validation code. ### Running the Server ```bash uvicorn app.main:app --host 0.0.0.0 --port 8000 --reload ``` The `--reload` flag enables auto-reload during development. For production, remove this flag and consider running behind a reverse proxy like nginx. ### Failure Modes Importing modules with circular dependencies crashes at startup with a cryptic error. Failing to install a dependency causes 500 errors at runtime instead of clear installation messages. Returning a raw Python dict from an endpoint bypasses Pydantic serialization, causing inconsistent response formats.

EXERCISE

Create a FastAPI application with two endpoints: GET /health returning {"status": "ok"} and POST /v1/models accepting a JSON body and returning it unchanged. Test both endpoints with curl.