FastAPI Backend Development Bootcamp · Leçon

Réponses en flux continu et requêtes par plages

Servez de gros fichiers avec StreamingResponse et prenez en charge les requêtes HTTP par plages pour permettre la reprise des téléchargements.

Réponses en flux continu et requêtes par plages est une leçon FastAPI Backend Development Bootcamp gratuite sur CoddyKit. Ceci est la leçon 2 sur 4. Tu peux lire la leçon complète ci-dessous gratuitement — puis la pratiquer en direct dans le navigateur avec un éditeur de code intégré et un tuteur IA 24/7. Elle fait partie du parcours d'apprentissage FastAPI Backend Development Bootcamp, et ta progression se synchronise sur le web et l'application CoddyKit. Le cours FastAPI Backend Development Bootcamp comprend 4 leçons au total.

Certaines parties de cette leçon n'ont pas encore été traduites et s'affichent en anglais.

Why Stream Responses?

By default, returning a file from FastAPI means loading the entire payload into memory before sending it. For a 2 GB video that is a disaster: memory spikes, slow first byte, and crashes under concurrency.

Streaming solves this by sending the body in small chunks as they become available. The server holds only one chunk at a time, and the client starts receiving data almost immediately.

StreamingResponse — wraps a generator/iterator that yields bytes.
FileResponse — a convenience for serving a file from disk efficiently.
Range requests — let clients fetch only part of a file (seeking, resuming).

This lesson builds all three, ending with resumable downloads.

A Generator That Yields Bytes

Streaming starts with an iterable of bytes. The cleanest source is a Python generator that reads a file in fixed-size chunks instead of all at once.

Here is the core idea, isolated from any framework. The generator yields 1 MB at a time, so peak memory stays tiny no matter how large the file is.

def file_chunks(path, chunk_size=1024 * 1024):
    with open(path, "rb") as f:
        while True:
            chunk = f.read(chunk_size)
            if not chunk:
                break
            yield chunk


if __name__ == "__main__":
    import os
    with open("sample.bin", "wb") as f:
        f.write(b"x" * (3 * 1024 * 1024 + 17))

    total = 0
    pieces = 0
    for chunk in file_chunks("sample.bin"):
        total += len(chunk)
        pieces += 1
    print("bytes:", total)
    print("chunks:", pieces)
    os.remove("sample.bin")

StreamingResponse Basics

StreamingResponse takes any sync or async iterable of bytes (or strings) as its first argument. You set the media_type so the browser knows how to handle the body.

Notice we pass the generator object itself, not its result — FastAPI iterates it lazily while sending.

from fastapi import FastAPI
from fastapi.responses import StreamingResponse

app = FastAPI()


def file_chunks(path, chunk_size=1024 * 1024):
    with open(path, "rb") as f:
        while chunk := f.read(chunk_size):
            yield chunk


@app.get("/download/report")
def download_report():
    return StreamingResponse(
        file_chunks("report.pdf"),
        media_type="application/pdf",
    )

Setting Content-Disposition

To make the browser download a file (instead of trying to display it) and choose a filename, send a Content-Disposition header.

attachment — force a download dialog.
inline — display in the browser if possible.
filename="..." — the suggested name.

Pass custom headers via the headers argument of StreamingResponse.

from fastapi import FastAPI
from fastapi.responses import StreamingResponse

app = FastAPI()


def file_chunks(path, chunk_size=1024 * 1024):
    with open(path, "rb") as f:
        while chunk := f.read(chunk_size):
            yield chunk


@app.get("/export/users.csv")
def export_users():
    headers = {
        "Content-Disposition": 'attachment; filename="users.csv"'
    }
    return StreamingResponse(
        file_chunks("users.csv"),
        media_type="text/csv",
        headers=headers,
    )

Streaming Generated Data On the Fly

Streaming is not limited to files on disk. You can generate the body incrementally — for example, exporting a huge CSV row by row from a database cursor without ever building the full string in memory.

The generator below yields one CSV line at a time. Each yield is flushed to the client as soon as it is produced.

import csv
import io


def csv_stream(rows):
    buffer = io.StringIO()
    writer = csv.writer(buffer)
    writer.writerow(["id", "name", "score"])
    yield buffer.getvalue()

    for row in rows:
        buffer.seek(0)
        buffer.truncate(0)
        writer.writerow(row)
        yield buffer.getvalue()


if __name__ == "__main__":
    data = [(i, f"user{i}", i * 10) for i in range(5)]
    output = "".join(csv_stream(data))
    print(output, end="")

FileResponse: The Easy Path

When you just need to serve an existing file from disk, FileResponse is simpler than wiring a generator. Starlette streams it efficiently and sets sensible headers for you.

Guesses Content-Type from the extension.
Sets Content-Length automatically.
Adds ETag and Last-Modified for caching.
Crucially, it already supports range requests out of the box.

For static, on-disk files, prefer FileResponse over a manual StreamingResponse.

from fastapi import FastAPI
from fastapi.responses import FileResponse

app = FastAPI()


@app.get("/media/{name}")
def serve_media(name: str):
    return FileResponse(
        path=f"media/{name}",
        filename=name,
        media_type="video/mp4",
    )

What Is an HTTP Range Request?

A range request lets a client ask for only part of a resource. The browser sends:

Range: bytes=1048576-2097151

The server replies with status 206 Partial Content and these headers:

Content-Range: bytes 1048576-2097151/5242880 — the slice and the total size.
Content-Length — the length of just this slice.
Accept-Ranges: bytes — advertises that ranges are supported.

This powers video seeking (jump to minute 5 without downloading minutes 0–4) and resumable downloads (continue from where a dropped connection stopped).

Parsing the Range Header

To support ranges manually, you must parse the Range header. The format is bytes=start-end where either side may be omitted:

bytes=500-999 — bytes 500 through 999.
bytes=500- — from 500 to the end.
bytes=-500 — the last 500 bytes (suffix range).

This standalone parser returns inclusive (start, end) offsets for a given file size.

def parse_range(header, file_size):
    units, _, rng = header.partition("=")
    if units.strip() != "bytes":
        raise ValueError("only byte ranges supported")
    start_s, _, end_s = rng.strip().partition("-")

    if start_s == "":
        # suffix range: last N bytes
        length = int(end_s)
        start = max(file_size - length, 0)
        end = file_size - 1
    else:
        start = int(start_s)
        end = int(end_s) if end_s else file_size - 1

    end = min(end, file_size - 1)
    if start > end:
        raise ValueError("unsatisfiable range")
    return start, end


if __name__ == "__main__":
    size = 5000
    print(parse_range("bytes=0-499", size))
    print(parse_range("bytes=4500-", size))
    print(parse_range("bytes=-100", size))

Reading Just the Requested Slice

Once you have (start, end), you must stream only that window. Use file.seek(start) to jump to the offset, then read in chunks while counting down the remaining bytes so you never overshoot end.

This generator yields exactly end - start + 1 bytes.

def ranged_chunks(path, start, end, chunk_size=1024 * 1024):
    remaining = end - start + 1
    with open(path, "rb") as f:
        f.seek(start)
        while remaining > 0:
            chunk = f.read(min(chunk_size, remaining))
            if not chunk:
                break
            remaining -= len(chunk)
            yield chunk


if __name__ == "__main__":
    import os
    with open("blob.bin", "wb") as f:
        f.write(bytes(range(256)) * 40)  # 10240 bytes

    got = b"".join(ranged_chunks("blob.bin", 100, 199, chunk_size=32))
    print("length:", len(got))
    print("first byte:", got[0])
    os.remove("blob.bin")

A Full Range-Aware Endpoint

Now we combine everything into one FastAPI endpoint that handles both full and partial downloads:

No Range header → stream the whole file with 200 OK.
Valid Range → stream the slice with 206 Partial Content plus Content-Range.
Unsatisfiable range → return 416 with a Content-Range: bytes */size header.

Always advertise Accept-Ranges: bytes so clients know seeking is allowed.

import os
from fastapi import FastAPI, Request
from fastapi.responses import StreamingResponse, Response

app = FastAPI()
VIDEO = "media/movie.mp4"


def ranged_chunks(path, start, end, chunk_size=1024 * 1024):
    remaining = end - start + 1
    with open(path, "rb") as f:
        f.seek(start)
        while remaining > 0 and (chunk := f.read(min(chunk_size, remaining))):
            remaining -= len(chunk)
            yield chunk


@app.get("/video")
def stream_video(request: Request):
    size = os.path.getsize(VIDEO)
    range_header = request.headers.get("range")

    if range_header is None:
        return StreamingResponse(
            ranged_chunks(VIDEO, 0, size - 1),
            media_type="video/mp4",
            headers={"Accept-Ranges": "bytes",
                     "Content-Length": str(size)},
        )

    start, end = parse_range(range_header, size)
    headers = {
        "Content-Range": f"bytes {start}-{end}/{size}",
        "Accept-Ranges": "bytes",
        "Content-Length": str(end - start + 1),
    }
    return StreamingResponse(
        ranged_chunks(VIDEO, start, end),
        status_code=206,
        media_type="video/mp4",
        headers=headers,
    )

Async Streaming and Cleanup

For non-blocking I/O under load, use an async generator. Reading the disk inside a thread pool keeps the event loop free; libraries like aiofiles do this for you.

Two important rules:

Streaming runs after your function returns, so resources opened inside the generator must be released in a finally block.
If the client disconnects mid-stream, FastAPI raises inside the generator — that finally still runs, so handles never leak.

import aiofiles
from fastapi import FastAPI
from fastapi.responses import StreamingResponse

app = FastAPI()


async def async_chunks(path, chunk_size=1024 * 1024):
    f = await aiofiles.open(path, "rb")
    try:
        while chunk := await f.read(chunk_size):
            yield chunk
    finally:
        await f.close()


@app.get("/async-download")
async def async_download():
    return StreamingResponse(
        async_chunks("big.bin"),
        media_type="application/octet-stream",
        headers={"Accept-Ranges": "bytes"},
    )

Quick Check

A client sends Range: bytes=2000-2999 for a 10000-byte file. Which status code and headers should your endpoint return for a correct partial download?

Recap

You can now serve large media efficiently and support resumable, seekable downloads.

StreamingResponse wraps a byte iterable so peak memory equals one chunk, not the whole file.
FileResponse is the easy path for on-disk files and already supports ranges plus caching headers.
A range request sends Range: bytes=start-end; reply with 206, Content-Range, a slice-sized Content-Length, and Accept-Ranges: bytes.
Parse the header (including suffix bytes=-N), seek(start), and read while tracking remaining bytes so you never overshoot.
Use async generators with a finally block to close handles even when clients disconnect mid-stream.

Questions Fréquemment Posées

La leçon « Réponses en flux continu et requêtes par plages » est-elle gratuite ?

Oui — le texte complet de « Réponses en flux continu et requêtes par plages » est gratuit à lire ici sur le web. Pour la pratiquer de manière interactive (un éditeur de code intégré et un tuteur IA 24/7) et déverrouiller le reste du cours FastAPI Backend Development Bootcamp, passe à CoddyKit PRO. Le cours FastAPI Backend Development Bootcamp comprend 4 leçons au total.

Qu'est-ce que j'apprendrai dans « Réponses en flux continu et requêtes par plages » ?

Servez de gros fichiers avec StreamingResponse et prenez en charge les requêtes HTTP par plages pour permettre la reprise des téléchargements. Tu pratiques FastAPI Backend Development Bootcamp avec du code pratique que tu exécutes directement dans le navigateur, et un tuteur IA 24/7 répond à tes questions au fur et à mesure que tu avances dans la leçon.

Dois-je avoir de l'expérience pour commencer FastAPI Backend Development Bootcamp ?

Aucune expérience préalable n'est requise. FastAPI Backend Development Bootcamp sur CoddyKit est structuré pour les débutants jusqu'aux apprenants avancés, donc tu peux commencer ici ou depuis le début et avancer à ton rythme. Ceci est la leçon 2 sur 4.

Combien de temps prend la leçon « Réponses en flux continu et requêtes par plages » ?

La plupart des leçons CoddyKit prennent environ 5–10 minutes. Chacune est courte et interactive, tu progresses régulièrement et tu repiques exactement où tu t'es arrêté sur le web et l'app.

Peux-tu écrire et exécuter du code dans cette leçon FastAPI Backend Development Bootcamp ?

Oui. Chaque leçon FastAPI Backend Development Bootcamp inclut un éditeur de code intégré, tu écris et exécutes du vrai code directement dans ton navigateur et tu reçois des retours IA instantanés — aucune configuration locale requise.

Toutes les leçons de ce cours

Téléversements multiparties et validation du contenu
Réponses en flux continu et requêtes par plages
Délégation du stockage vers des compartiments compatibles S3
Transformation asynchrone d’images et de documents

← Retour à FastAPI Backend Development Bootcamp