"""captions.py: live captions on every viewer's page, in Python. It reads a transcript as it is
spoken, one line at a time on its input, and publishes each line to a captions channel. It never
joins anything: only HTTP pushes, signed with a key that may publish there and nowhere else.

    your-transcriber | python captions.py

Each input line is plain text (a finished caption), or JSON from a speech-to-text stream such as
Transcribe Today: {"text": "…", "final": true}. Unfinished text ("final": false) is sent as
"partial", at most 4 times a second, so the page can show words as they come.
Python 3.10+. pip install -r requirements.txt (reads .env).

Its key (dashboard, New API key): right publish, kind public, channels: prefix captions-.
"""

import json
import sys
import time

from pikoman_bot import Rest, env

CHANNEL = env("CHANNEL", "captions-talk-1")
PARTIAL_EVERY = 0.25  # seconds: at most 4 partial updates a second

rest = Rest()  # PIKOMAN_APP and PIKOMAN_API_KEY; PIKOMAN_HTTP only for another server
sent = 0
last_partial = 0.0


def publish(event, text):
    global sent
    try:
        rest.publish(CHANNEL, event, {"text": text, "at": int(time.time() * 1000)})
        if event == "line":
            sent += 1
    except Exception as e:  # noqa: BLE001: the next line goes anyway
        print(f"publish {event}:", e, file=sys.stderr)


def read(line):
    """One input line: a caption as text, or {"text", "final"} from a speech-to-text stream."""
    line = line.strip()
    if not line:
        return None
    if line.startswith("{"):
        try:
            data = json.loads(line)
            text = data.get("text")
            if isinstance(text, str) and text.strip():
                return text.strip()[:500], data.get("final", True)
            return None
        except ValueError:
            pass  # not JSON after all: a caption that starts with {
    return line[:500], True


try:
    # Lines are sent in the order spoken, each as it comes. A partial is sent only when 250 ms have
    # passed since the last one: between them, newer words replace it.
    for raw in sys.stdin:
        caption = read(raw)
        if caption is None:
            continue
        text, final = caption
        if final:
            publish("line", text)
        elif time.monotonic() - last_partial >= PARTIAL_EVERY:
            last_partial = time.monotonic()
            publish("partial", text)
except KeyboardInterrupt:
    pass
print(f"Done: {sent} lines to {CHANNEL}")