Stream and Schedule Downloads

Client.iter_download() yields ordered bytes and does not materialize the full file. It accepts normalized media, compatible raw media, or a miniproto file ID. The stream rejects resume and multi_session; use an explicit offset for a ranged continuation.

from pathlib import Path

from miniproto import Client


async def stream_to_file(client: Client, file_id: str, destination: Path) -> None:
    with destination.open("wb") as output:
        async for chunk in client.iter_download(file_id, concurrency=2, max_in_flight_bytes=4 * 1024 * 1024):
            output.write(chunk)

The stream closes outstanding part tasks when the generator is closed or its caller is cancelled. Its concurrency is a bounded part-request window, not a promise of throughput; choose it alongside the byte budget and observe server flood waits, local memory, and connection health.

For separate known media objects, schedule top-level downloads concurrently and let the client-level media scheduler apply its configured per-DC budgets:

import asyncio
from pathlib import Path

from miniproto import Client


async def download_pair(client: Client, first: object, second: object) -> None:
    async with asyncio.TaskGroup() as tasks:
        tasks.create_task(client.download_media(first, Path("first.bin"), concurrency=2))
        tasks.create_task(client.download_media(second, Path("second.bin"), concurrency=2))

download_media() can write to memory, a path, or a caller-owned destination; it may use multiple sessions only for complete, known-size bot downloads with sibling storage, and otherwise falls back to one session. See Media Primitives for range alignment, integrity checks, CDN redirects, destinations, progress callbacks, file IDs, and cancellation semantics.