services/tts.py

import os, uuid, base64, requests
from functools import lru_cache

TTS_API = os.getenv(
    "TTS_API_URL",
    "https://rahul7star-Chatterbox-Multilingual-TTS-API.hf.space/tts"
)

SESSION = requests.Session()

@lru_cache(maxsize=128)
def synthesize(text: str, lang: str) -> str:
    payload = {
        "text": text,
        "language_id": lang,
        "mode": "Speak 🗣️"
    }

    r = SESSION.post(TTS_API, json=payload, timeout=None)
    r.raise_for_status()

    path = f"/tmp/tts_{uuid.uuid4().hex}.wav"

    if r.headers.get("content-type", "").startswith("audio"):
        open(path, "wb").write(r.content)
        return path

    audio = r.json().get("audio") or r.json().get("audio_base64")
    open(path, "wb").write(base64.b64decode(audio))
    return path