Live sources

Anything with a .read(n) that only grows: a downloader's output, a pipe, an HTTP stream.

import urllib.request
import tgrtmp

handle = urllib.request.urlopen("http://example.org/live.ts")

player = tgrtmp.publish(handle, "rtmps://dc4-1.rtmp.t.me/s/", "KEY")

Live sources cannot be paused from the outside; they keep producing. tgrtmp offers two ways to handle that, picked by buffer_size.

Default: no buffering

buffer_size=0 (the default) drains nothing in advance: pause freezes the picture on the connection, and resume jumps straight to the live edge. Zero overhead: no reader thread, no memory.

Optional: the DVR window

Give buffer_size some bytes and the source is drained into a rolling window, so a pause holds its point while the live edge keeps filling it. Resume where you paused, like a DVR, as long as the pause fits the window. See the live window.

player = tgrtmp.publish(handle, url, key, buffer_size=48_000_000)

Which sources behave how

Source	Position	Pause (default)	Pause (with window)	EOF
Path or open file	exact	freezes the frame	same	holds last frame
Live .read(n) handle	your point	freezes, resumes at the live edge	holds its point	holds last frame

A path is opened and closed for you; an open file or handle is yours to close.